PER DAMGAARD HUSTED

AI for CEOs
 Copyright © 2020 by Per Damgaard Husted

All rights reserved. No part of this publication may be reproduced, stored or

transmitted in any form or by any means, electronic, mechanical,
photocopying, recording, scanning, or otherwise without written
permission from the publisher. It is illegal to copy this book, post it to a
website, or distribute it by any other means without permission.

Per Damgaard Husted asserts the moral right to be identified as the author
of this work.

Per Damgaard Husted has no responsibility for the persistence or accuracy

of URLs for external or third-party Internet Websites referred to in this
publication and does not guarantee that any content on such Websites is, or
will remain, accurate or appropriate.

Designations used by companies to distinguish their products are often

claimed as trademarks. All brand names and product names used in this
book and on its cover are trade names, service marks, trademarks and
registered trademarks of their respective owners. The publishers and the
book are not associated with any product or vendor mentioned in this book.
None of the companies referenced within the book have endorsed the book.

The books cover is designed by Natalia Elichirigoity and the editor is Eve
Potts.

Second edition

ISBN: 978-87-972249-2-2

This book was professionally typeset on Reedsy.

Find out more at reedsy.com
 Contents

1 Introduction 1
Why I wrote this book 5
2 What is AI? 10
The reasoning behind Artificial Intelligence 10
The perspectives with AI 11
Artificial General Intelligence 31
AI as a Knowledge System 36
AI technologies 54
3 Operations and AI 67
How do you create value in your business? 70
How does AI create value for you? 78
Identification of your AI project candidates 85
4 Strategy and AI 121
Should you become employeeless? 121
How to create exponential growth 125
Scaling your business on feature, function, or corporate level? 143
AI as a competitive driver 175
Next steps 187
5 Leadership and AI 189
AI changes the basic premise of how you perform leadership 189
AI and leadership 191
AI or IT? Know how to make the right choice 203
How to design an AI project 209
The art of training your AI 214
How to work with AI as a co-worker 227
My personal experiences with management and AI 233
About the Author 239
 1

Introduc on

A
rtificial intelligence (AI). Do you use it in your business? Maybe the
technology is something you know of but have never used it. Perhaps
you already have some experience with your own AI projects, or with
services you use that are made with AI.
The media is full of stories about Silicon Valley companies using AI to create
something new and revolutionary almost every day. Philosophers discuss
whether artificial intelligence will be smarter than us at some point. And
there are continuous predictions about the Forth industrial revolution that
will be driven by AI1 . We read stories that 40% of the workforce’s jobs will
be taken over by AI and that it is now the lawyers, economists, and financial
professionals who can expect to lose their jobs to this new technology.2
So, there are lots of different people who have options on how AI can become
significant on many levels of our daily lives. But do you know exactly what AI
could mean for the way your business is run?
It may well be that the AI-driven changes will be insignificant in your

1
To understand more of the Fourth Industrial Revolution, I suggest the book “Shaping the
Fourth Industrial Revolution” by Founder and Executive Chairman of World Economic Forum,
Klaus Schwab
2
Kai Fu Lee, who wrote the bestseller “ A.I. Superpowers” is among the many who share this
view. See the interview with him as a reference: https://fortune.com/2019/01/10/automation-
replace-jobs/

1
 AI FOR CEOS

industry. That technology is not going to drive change on how your business
could be run. But how can you know if that’s right in your case? Imagine if
your industry is the next transformation case, without you realizing it and
therefore not acting on it.
We have seen technological change processes before. The Internet and the
Digital Transformation are the latest examples. The overall dynamics are
always the same. The premises for how you run your business are challenged.
Those who understand how to play with the new premises win—those who do
not, lose.
But when you dig a little deeper, we also know that not all companies feel
the changes the same way. The changes will happen in different places of your
business and will occur with different levels of strength.
Let’s take the product you are currently engaging with as an exam-
ple—namely this book.
You either read the book as a physical book or as an e-book. Or maybe you
don’t read it. You could have got it as an audiobook. You may have borrowed
the book from the library, purchased the book at the bookstore, online, or
maybe it was available to you as part of an online subscription service that you
have signed up for. And you probably got interested in the book because you
read a review of it on the web somewhere.
It is still possible to read a review of a book in a physical newspaper, then
go to a bookstore and buy it, as you would have done in the old days, before
the year 2000. It’s just not the most typical way to find and read new books
anymore.
So, the publishing industry is an excellent example of a sector that has been
transformed by digital technologies. Everything has changed. The way books
are produced, the way customers read them, the way books are marketed, how
they are distributed, and the underlying business models.
All this has happened because Internet technologies have changed the
premise of how knowledge in the form of books can be distributed. Every
other industry has also seen Internet-driven change. But that has happened
on different premises. Of course, some changes are similar across industries,
but many are also unique to their particular situation.

2
 INTRODUCTION

And now AI is coming.

I think many feel they are back to where the Internet journey began to
emerge. You have a clear sense that something is going to happen. And it has
the potential to be huge.
But what, and how, can be very abstract and difficult to grasp on a more
specific level. What should we do so that AI starts to become an asset for us?
What do we need to do, to understand the technology and work with it in ways
that can create new value for us and our customers?
This book is written for you if you have such considerations. It is written for
people who would like to understand how Artificial Intelligence can be used to
create growth in their businesses.
A part of your job is properly to make sure that your business is taking
full advantage of its technological potential. So, you can make the right
management decisions on how to use technology. And that includes AI.
You know Artificial Intelligence from the media’s success stories. But you
are looking for a basic understanding of what the technology is, and what it
can do for you and your business.
You want to make sure you don’t miss out on opportunities to make your
business better. So, you can find the areas that could benefit the most from AI.
I also imagine that you have been thinking about the possible strategic
changes that AI could drive. Is your industry the next publishing industry?
Is AI an area that redefines how you conduct your business, or is it merely a
sporadically used technology that you can use in a few operational tasks? You
want to understand the potential so that you can act with the right level of
leadership attention.
They are precisely the desires, reflections and considerations that the book
is written to help you to address.
The book seeks to give you an understanding of what AI is, why technology
can be used in your business, how you can use it, and where it makes the most
sense for you to have your focus. Not on a technical level but seen in a business
context.
The key here is that AI creates value. Your business creates value. You will
get an understanding of why the two things are interconnected and what it

3
 AI FOR CEOS

means to you and your business.

It is not a technical book, and it is not a quick fix book that gives 50 tips on
how to get started on an AI project tomorrow. It is a book that seeks to provide
you with the understanding you need to unite the core of your business with
the new AI brings to the table. The common denominator is value creation.
AI will challenge your value creation, and it can unfold on many levels and
many dimensions of your business. In operational terms, you can solve your
problems and your tasks differently with AI. Strategically, AI has the potential
to disrupt the way you create value and could redefine the goals you have with
your business.
The book helps you understand why you can work differently with AI and
how to uncover where it makes sense for you to start. Part of this means
getting tools to see where you can use the technology best in your business.
You cannot speak of such fundamental changes as AI has a potential to drive
without also addressing leadership. The potential depth of the change alone
requires a leadership effort, but as you will see, that is the least important
leadership aspect.
AI has the potential to change the premise for the way we work. The number
of tasks that need to be solved, how we work with technology, and how we
organize our business. All could change. This requires a revised view on
leadership that the book will provide you with perspectives on too.
You have to look at the knowledge you get from the book as a platform
on which you can build your AI efforts. It gives you the perspective and the
conceptual business understanding of the technology that you need to make
the right choices for how you will lead your business into the Age of AI.
The book’s four sections are designed to help you get a grasp of it all. The first
section lets you know what AI is, why it is different from any other technology
that you have come across.
Then, we zoom in on your business and why the way you create value can
be improved by the use of new AI technologies. The focus then shifts to
why AI could create new strategic opportunities for you, such as identifying
the prerequisite for creating exponential growth in your business and why
becoming employeeless could emerge as a path that you as a leader must form

4
 INTRODUCTION

an opinion on.
Finally, we address the leadership aspect of why running an AI-driven
organization will require a new mindset of all involved.
I’ll return to all that and much more soon. First, you briefly get my story
and my background for writing this book. So, you know a little more about
who I am and where my knowledge of AI in a business context comes from.
And then we get started on the first part of the journey that helps you
understand the concept of AI and why it could drive a profound change in
how you run your business.

Why I wrote this book

Before we move on to what AI can do for you, I would like to introduce myself
and to let you in on my reasons for writing this book. My name is Per Damgaard
Husted; I live in Copenhagen, Denmark. I’m the Founder and CEO of two AI-
based start-ups; Canecto and Cognifirm.
Before starting Canecto, I worked with IT management in major Danish
companies within the Finance, Consumer Products, and the Telecom indus-
tries. In my jobs, I have always worked in spheres that combines business and
technology.
In the last job I had before starting Canecto, I worked as an IT manager at a
bank where I was responsible for the operation of the bank’s online banking.
A part of my job was to work with users’ experiences of our online products
and services. Were user actions on our websites as expected? Did the customers
take the user journeys on our sites that we would like them to? Did our
customers book enough meetings with our sales staff? Did the customers
buy our new products online?
In other words, we had to ensure the customers’ quality experience in our
online bank was excellent and that we met the business requirements to serve
the customers professionally and to sell more to them.
One of my major frustrations in my job was web analytics or to be more
precise, the quality of the analytics that we had available. There was a lot of
focus on traffic in our web analytics. To help us get answers to questions like:

5
 AI FOR CEOS

• Where did the users come from?

• How long were they on our site?
• Have they visited the website before?

However, there was very little knowledge of what our customers liked about
our products in our analytics. What did they like, what didn’t work? Why did
they behave as they did on our website? The type of information where you
gain a more substantial understanding of the motivation behind user behavior.
I felt that we couldn’t get enough clear answers with the data and tools we
had available to us. And that it was valuable knowledge that we were missing.
Insights into users’ motives and motivations provide the basis for learning.
What works, what motivates our users, what should we do more of, what
should we do less of?
Relevant questions if you want to make the quality of our products and
services better, and at the same time, try to understand the motivation behind
users’ behavior.
However, my problem was that the tool that could do just that for us didn’t
exist. All web analysis tools were based on “old-fashioned” technology,
focusing on showing as many quantitative ways in which users used the bank’s
online services. How long were they on the site? How many pages did they
see? Which city did they come from?
A good way of looking at it is that the tools could describe users’ behavior,
not explain them. We got to know what users did, not why they made the
choices they actually did.
At the same time, I knew that this particular problem was something that
AI could be used to solve.
A thought began to grow in me. If you were to start over and build a web
analytics tool based on AI, then you could create a tool that would be able to
explain, do not describe the behavior of users. That could be great.
We should be able to understand the user behavior in a way where we would
understand why the users made the decisions they made. So, we could see the
motivation for their behavior. Such insights could help us create better online
solutions.

6
 INTRODUCTION

If you understand what motivates users to do their actions, then you have
an insight and a level of understanding that is far deeper than just looking
at numbers and data. You get closer to customers as people. This gives you
a much better starting point for being able to customize your product and
services so that you meet customers on their terms.
I knew that AI was really good at solving such problems. Because one of AI’s
strengths is to find patterns in complex information, AI can find the strongest
patterns, which in our case, best explain a user’s action on a website.
So that’s why we started Canecto. We based the company at the Technical
University of Denmark, so we could access the best AI experts. Our goal was to
do web analytics based 100% on Artificial Intelligence. We succeeded in doing
that.
Our product is an automated web analytics report that gives you specific
improvement suggestions on how your website can sell more products, how
to write better texts and where you should focus on to get more of the right
customers to your site.
To create our product, it required that we invented more than 25 different AI
algorithms. Our algorithms have an enormous academic span. We work with
algorithms that seek to understand what users’ interests are when navigating
a site, through advanced text analysis. Another example is that we also use AI
to predict which online initiatives are most likely to sell more products to our
customers.
Now, it’s almost three years since we started Canecto, and I want to share
what I’ve learned, but maybe not in the way you think.
Many of the questions I get from our customers and partners are not about
our products. They have been on how to work with Artificial Intelligence, how
to see the new strategic opportunities that the technology provides, and how
to develop solutions that solve problems in new ways.
In Canecto, we solved the challenge of web analytics differently. The
technology allowed us to create a new type of product that gave us a different
business starting point.
In my many conversations with clients and partners, it became clear to me
that our situation was not so unique, that others also considered how they

7
 AI FOR CEOS

could leverage AI, but they have a hard time getting a grip on it. Where to start
and in what order? It is the reflections on their frustrations that motivated
me to write this book.
If I could extract the sum of my experiences, then I could help others
get started on their AI journey by combing my managerial experience from
previous jobs, with my start-up knowledge. This combination makes me
understand the starting points our customers and partners were in. And I
know how to develop and work with AI on both technical and business levels.
So that is the sum of my previous leadership experiences and my knowledge
as an AI start-up CEO on which this book is built.
I believe that I understand the situation you are in as a leader in a company
when you need to evaluate the potential of new technologies and the decision-
making process you have to go through.
I have been there many times, both as a manager and as a Project Manager.
The potential must be uncovered, the opportunities must be described so that
others in the organization can see them too, and the project must be sold into
the management, anchored in the organization and implemented.
I also understand AI as a technology. I hold a patent of an AI infrastructure
design on how to work with AI. Just like three years of experience in developing
AI solutions also makes me feel relatively comfortable when the talk falls on
the technology’s applications, design, and its capabilities.
So, it is the sum of my AI knowledge and the questions that our customers
have asked me that have defined the content of this book.
I have sought to strike a balance in the book between theory and practice.
AI is so different an area that you cannot understand it without having a
theoretical substance. You get that by working with what I call a knowledge
system. It is called a knowledge system because AI in its essence is about a
new way of processing knowledge in your business.
My opinion is also that you first gain the value of a topic if you act on it.
That’s why you also get a number of tools to help you do just that in the book.
Among other things, you can uncover which AI project candidates there
could be in your company. Another example is a tool to help you assess AI
strategic potential for you.

8
 INTRODUCTION

However, before we get there, you need to know more about what kind of
technology AI is and how it works. AI is very different from everything you
have encountered in the past. There is a reason why experts claim that 40%
of all jobs are at risk of being taken over by AI3 , and the first part of the book
probably makes you see why this might just be right.

3
https://www.mckinsey.com/business-functions/mckinsey-digital/our-insights/where-
machines-could-replace-humans-and-where-they-cant-yet

9
 2

What is AI?

The reasoning behind Ar ficial Intelligence

I
n this part of the book, you will get answers on what AI is, and how you
can use the technology in your business.
Figuratively speaking, you should see AI as a new tool that you must
get to know. Even though AI is a technology, the tool you will get to know is
not a technical tool, it’s a new tool that you can use to lead and manage your
business - on brand new terms.
AI redefines the kind of value you can create in your business and the way
you create value. It is in this context you need to understand AI and this what
I aim for you to be able to do in this book. So, it is all about management and
leadership.
However, for you to understand the leadership part of AI, you need to know
how AI works. Otherwise, you will not know what tasks you can solve with AI.
This is what I aim to make you understand in this part of the book.
It means that this section is a bit more theoretical than the rest of the book.
However, you need to understand the dynamics behind the technology before
we can begin to really work with it and make it tangible for you and your
business.
Don’t get me wrong. It is not that the theory part in any way is boring. As

10
 WHAT IS AI?

technology, AI is super exciting, and the reasoning behind it will challenge

both your intellect and the way you see the world. I promise you that you will
see many things from a different perspective, once you know how a problem
can be solved with AI. Your view of the world will never be the same.
So now you need to prepare yourself to be provoked and to reflect on the
way you create value in your business. Just don’t expect to understand how
to solve all your business problems with AI after reading this section book
because you will not be able to do so (yet).
You are about to learn the concepts that define AI, and why it is so different
as a technology, AI is unlike anything else you may have come across. But
that’s the least exciting part. The fascinating point is that AI represents a new
way of working with knowledge. This opens up a spectrum of new ways in
which you can create value in your company.
That’s what this part of the book is about. And on the way to understanding
this, we will come across Stanford cancer research and why AI will radically
change the lives of photo models.

The perspec ves with AI

Everyone has heard of Artificial Intelligence, but it is probably difficult for

many to explain what the domain is all about.
The definition I like best is that which sees Artificial Intelligence as not
human intelligence—meaning, intelligence that originates from machines,
as opposed to natural intelligence that originates from humans4 .
The concept of Artificial Intelligence is often used to describe machines that
perform cognitive functions such as problem-solving or learning. I will get
back with much more on what this means in real life later. First, the definition
of AI needs to be unfolded a little bit more so you get a better understanding
of the idea behind it.
When you read articles on Artificial Intelligence, you often encounter
different words and concepts used to describe the field. The three that you

4
This definition was introduced by Max Tegmark in the Book Life 3.0

11
 AI FOR CEOS

will most frequently face are5 :

• AI
• Machine Learning
• Deep Learning
The short answer to the difference between them is that AI is a collective term
for all Artificial Intelligence technologies. Machine Learning is a (popular)
branch of Artificial Intelligence. Deep Learning is another branch of the
Machine Learning universe. So, mathematically, it is a combination of real
subsets, as shown in this illustration.

The significance of these three key concepts is:

• AI is the broadest term for Artificial Intelligence. It means that it covers

everything from conventional AI methods to advanced Machine Learning
and Deep Learning techniques.
• Machine Learning is a subgroup of AI that covers algorithms that learn
by being trained with data. Countless Machine Learning techniques exist.
If you come across words like linear regression, decision trees, then they
are all Machine Learning techniques.

5
This IBM article is one of many that covers the difference between the three types of
AI: https://www.ibm.com/blogs/systems/ai-machine-learning-and-deep-learning-whats-
the-difference/

12
 WHAT IS AI?

• Deep Learning refers to algorithms that can become self-learning if they

get enough data. A Deep Learning algorithm works in a way that is inspired
by the human brain. The way Deep learning is designed architecturally is
called neural networks.

The area that is growing the most by far is Deep Learning. Many of the new
revolutionary AI inventions that we almost daily hear of are made as Deep
Learning solutions. So, understanding what exactly Deep Learning can do for
you is significant, and I’ll return to that later in the book. First, I will give you
an example of how AI can make your daily life more convenient.

What makes AI smart?

From a business point of view, a very relevant question would be to ask: Why
is AI smart? How can it improve the way I run my business? And what exactly
makes AI so different from the IT I know and use today?
These are precisely the questions that this part of the book will provide
answers to. We’ll start by rolling out why AI is smart in a corporate context.
The short answer to the question is that technology enables you to automate
making decisions and solving tasks in your business. And that’s a big deal.
Whatever your business is involved in and what the goal of your company is,
a central condition will always exist. You make decisions and you solve tasks.
Today, some tasks are solved by your IT systems, some are solved as ad hoc
tasks, and in other contexts, you use processes as the way to ensure that tasks
are completed.
Let’s say your business sells curtains from a fleet of vans that visit your
customers at their homes. When your staff meets the customer, they take
measurements, choose the curtain materials with the customer, and hopefully
closes the sale. And two weeks later, you deliver customized curtains to your
customers.
Today, if a potential customer calls you, an employee must pick up the phone,
open a calendar program (e.g. Outlook), find a time when one of your vans will
be in the area where the customer lives. Once the appointment is confirmed,

13
 AI FOR CEOS

your employee books a time for a visit and ends the phone call.
You probably have a process for how such a conversation should be handled.
You have a process to guarantee that you have consistent customer experience
and to ensure that neither you nor the customer is missing information when
the conversation is finished. You have IT systems that enable you to book
the customer’s appointment. Your IT also makes it possible to communicate
with your curtain van driver so that he arrives at the arranged time at the
customer’s home.
So, it is a simple task involving people, processes, and IT systems. AI has
the potential to replace all three things. An AI bot can take the customer’s
phone call. Meaning, your AI will be having a dialogue on the phone with the
customer to ensure that the booking is done precisely as your employee would
have done it. And when the booking is confirmed, your AI will ensure that
the right next actions are taken, such as booking your van driver and sending
confirmation emails to your new potential customer.
In other words, your AI will be performing precisely the same task that used
to require efforts and time from your employees, access to an IT system, and
the use of a specific process flow.
AI replaces all of this, and once your AI system is installed, it can handle
anywhere from 1 to maybe 1000 calls per minute because there are no
dependencies of people in the design of the solution because such a setup
scales exponentially.
Google has already made a bot that can book appointments. It’s called
Duplex, and you can watch a YouTube video showing the principle at this link:
https://youtu.be/D5VN56jQMWM
A brief note. You should know that all links and external references are
available on the book’s website cognifirm.com/book, so you can always find
them and explore the content, even if you read the book as a physical book.
And back to the case.
The obvious learning is that such new technologies will change the opportu-
nities you have for running your business more efficiently. It will also enable
you to offer a range of new products and services that would not have been
possible before.

14
 WHAT IS AI?

The curtain van is just one example. As a case is it compelling, but you need
to know how to zoom in on AI and your business.
AI can change your business in five different areas with the effects that are
shown in this overview.

I’ll soon discuss what the five areas mean, so you can get a complete picture
of your overall opportunities on how to best work with the technology in ways
that would be most appropriate for your business needs.
To help you do this, you will get help from a set of frameworks that are
designed to help you identify where in your business AI can help you create
new ways of value.
However, before we get to that part, you have to understand two very
fundamental issues, namely how AI works and perspectives on how AI differs
from what you can do with IT.

15
 AI FOR CEOS

How does AI work?

The focus of this book is more on business than on technology, but there are a
few areas you need to have a more in-depth insight into for you to understand
what opportunities this new technology may have for you.
The first is a slightly more comprehensive explanation of what Deep
Learning is. Deep Learning is by far the hottest in Artificial Intelligence, and
is also the branch of development that could move your business the most6 .
What makes Deep Learning special is that you train the computer system.
Usually, you create IT by creating lines of code in the form of a computer
program or an app. That is not how you create Deep Learning applications.
In Deep Learning, you have several sub-components that look for patterns
that can explain a given behavior. Sometimes a subcomponent is of great
importance to the behavior. Other times it has none or little significance.
Conceptually, one often draws a Deep Learning system in the same way as a
brain that processes information. Here’s how:

6
If you are looking for a more technical approach to Deep Learning, the I suggest you read the
book Deep Learning (MIT Press Essential Knowledge series) by John D. Kelleher

16
 WHAT IS AI?

What is smart about processing data in this way is that the patterns that can
describe a user’s behavior the best arise in a learning process.
For example, let’s say that we would like to train a Deep Learning algorithm
to understand human handwriting.
You do this by showing such a system may be a thousand different ways in
which you can write the number “3” or the word “dog” with various people’s
handwriting. Then the system learns which patterns characterize whether a
number is “3” or what determines how the number “10” is written in hand.

If you ask a person why they can read the number “3” almost no matter how
it is written by hand, then very few people can articulate why they can identify
what they see as the number “3”.
But we do not doubt that what we see is the number “3” and not “6”. Deep

17
 AI FOR CEOS

Learning works with the same kind of logic. Humans recognize the number
“3” correctly because we use our intuition.
Intuition is defined as “the ability to understand something instinctively,
without the need for conscious reasoning”7 . As humans, we use our intuition
all the time.
Deep Learning works in ways that are similar to our intuition - It does not
use logic as we know it from the IT systems that we use every day.
Deep Learning algorithms work similarly to human intuition. There are no
rules in a Deep Learning system that determine how to define the number
three written as handwriting.
A Deep Learning system determines that a given number is “3” based on
the knowledge from an algorithm. This algorithm has gained this knowledge
by learning perhaps a thousand different ways in which the number three
can be written, and maybe 10,000 ways in which the number should not be
written. So the AI makes a deduction. This number is “3” when confronted
with a number “3”.

Deep Learning can predict human behavior

Another great strength of Deep Learning is that it is a technology that can be

brought much closer to our everyday lives than conventional IT can.
Conventional IT systems are designed as math and logic-based tools. For
them to work, they assume that mathematical formulas can describe the
environment they are looking at. It is a useful approach if you develop
spreadsheets or design a banking system.
But if you want to make services that interact with people on our terms,
that is, speech, text, and non-verbal communication, then the mathematical
problem-solving approach will give you problems. Deep Learning is already

7
You can find the full definition at this Wikipedia page: https://en.wikipedia.org/wiki/Intuition

18
 WHAT IS AI?

technologically superior to all conventional IT systems in such areas.8 The

consequences of this difference between AI and IT is a core area that we will
get back to later in the book.
Deep Learning can even be made so advanced that it could make predictions
about human intentions and rational behavior. Think about it and its
significance, that’s a big deal. There are various examples in the book of
how it would be practically possible to work with technology that way for you.
In the example of the curtain van, it could mean that the AI system could,
with a certain probability, predict the type of curtains the customer would end
up buying.
That knowledge could be based on information about what day the meeting
was held, the flow of the phone call, how long the conversation was, how many
clarifying questions the customer asked, at what point in the conversation
did the customer inquired information about the cost? All of the above is
information about the customer’s behavior.
Our AI can make predictions by looking at patterns from past customers’
buying behavior and comparing them with our new customer’s behavior.
The AI will look for significant patterns in how the new customer behaved.
The system will then make predictions based on this knowledge. Such forecasts
could include an expected output of your relationship with this customer, such
as how many visits you need before they buy from you, whether you can make
upsells and whether they are sensitive to price or not.
This means that we can create a set-up where we could be more knowledge-
able of a specific customer’s buying preferences than they are aware of. So,
we know in advance how they will react to the core elements of the meeting
with us.
Your employees would thus be able to know which products the specific
customer prefers without even talking to the customer, and the best ways to
close sales with him.

8
This research from the three Stanford professors Andrea Stevenson Won, Jeremy N. Bailenson,
Joris H. Janssen shows as an example of why Deep Learning is superior to IT technologies in
the most complex area, non-verbal communication. https://vhil.stanford.edu/mm/2014/won-
ieeeac-nonverbal-predicts-learning.pdf

19
 AI FOR CEOS

This level of knowledge will change the premises for the level of customer
service that you could provide.
For example, your Sales Reps will be able to zoom in to your client’s needs
quickly, and you will appear professional in the client’s eyes. Or maybe you
could offer a virtual sales experience that matched the level of service that an
actual Sales Rep could provide.
It is because of examples like this that some conceptually describe Deep
Learning as systems with intuition.9 The system cannot explain why it thinks
a given customer will be interested in a given product. However, if the system
has enough data to learn from, it will be able to predict a given customer’s
behavior with consistent accuracy.
The fact that AI can learn the meaning of behavior and make qualified
proposals for actions based on the knowledge acquired is a big thing. How you
could use this in practice is also a recurring theme in the book.
But before we open up to that talk, you need to understand the very
central premise of why AI is fundamentally different from anything you’ve
experienced in technical solutions before.
It starts a bit abstract in the way we deal with knowledge in general.
However, the description opens up the understanding of the most central
AI characteristic, namely that AI automates the processing of knowledge.
Knowledge can be described hierarchically in levels as a pyramid, and it
helps to understand what AI can do for us, and why it is different from IT.

The knowledge Pyramid

The conceptual difference between AI and IT is that AI is based on knowledge

while IT is based on data. So, AI can be said to be a knowledge-based
technology, and IT is a data-based technology. And what does that mean?

9
Carlos Perez describes the concept of Deep Learning and intuition well in his book “Artificial
Intuition: The Improbable Deep Learning Revolution.”

20
 WHAT IS AI?

The concept called the Knowledge Pyramid10 can help us gain perspective
on the difference between IT and AI.
The Knowledge Pyramid does not, as a concept, have anything to do with
technology. The concept explains levels of enriched knowledge. Specifically,
the pyramid tells us that knowledge can be enriched on four different levels.

As a thought example that can explain the four levels, we use a weather station.
Let’s say you’ve set up a digital weather station on the roof of your house.
It measures temperature, humidity, air pressure, wind direction, and wind
speed. You can see the weather information on a monitor in your house, and
perhaps also on an app on your Smartphone.
Your weather station probably works with all of the four levels of knowledge
in the Knowledge Pyramid.
The lower level of the pyramid is called data. Data is unstructured and
unprocessed information. In our example, it is the particular weather
measurements that the weather station records. So, data is the result of the
measurements in raw form. All these measurements are numbers. If you put
them in a spreadsheet, you would just get rows of numbers.
If you showed your weather measurement data to other people without

10
The concept is described on this Wikipedia page: https://en.wikipedia.org/wiki/DIKW_pyra-
mid

21
 AI FOR CEOS

telling them where they came from, then nobody would be able to use them for
anything. For anyone but you, they would just be numbers in a spreadsheet.
Maybe some could say that they were dealing with temperature or wind, but
that would be guesswork.
That is the essential characteristic of data in its raw form. Data in itself has
no value because looking at data alone does not make us understand what it is
used for. This is not possible for us, because the context of the data you are
looking at is not known to you.
That’s why we have information. For now, there is an understanding
level that gives data value. Now we put units on our weather data. So, our
spreadsheet now consists of data in a structure. For example, the information
is divided into time intervals. The temperature measurements are shown with
Celsius or Fahrenheit as the unit; we show the wind direction as degrees, etc.
What we do now is to create structure in our data, and thus we enrich it
so that it is no longer data. It’s information. In our new spreadsheet, other
people will now have a realistic chance to understand what the numbers in the
spreadsheet mean.
The preconditions for extracting knowledge from the spreadsheet are that
you know what Celsius and degrees are. People know this because they have
learned it. When you look at numbers with these units attached, you decide that
it is probably weather numbers because you have knowledge and learning that
make it possible to see a logical link from the numbers in the spreadsheet to
weather information. What you do is professionally called deductive learning.
In doing so, we move up to the third level of the knowledge pyramid. Here,
we find knowledge. By looking at our numbers in the spreadsheet, we know
that we are looking at weather observations. It is an understanding that comes
from a combination of the information we see in the spreadsheet and our
knowledge. We know, for example, that the number 25 degrees Celsius means
it is warm enough to wear a T-shirt if you need to get out of the house.
The next - and last - level is wisdom. Now we start acting on our knowledge.
If we see a rapid drop in air pressure and changes in wind direction, then
we know that there is a high risk of rain. So, we go out and remove the
cushions on the patio furniture. Such actions are guided by a reflection of what

22
 WHAT IS AI?

we can deduct from the weather information. So, we are doing something
because we are dealing with the consequences of the knowledge we get from
our spreadsheets.
Thus, the lowest level of the pyramid of knowledge is data, and the highest
is wisdom. The definitions of the four levels are:

• Data: A collection of facts in raw or in unorganized form.

• Information: Organized and structured data that has been cleared of
errors. It can, therefore, be measured, analyzed, and visualized.
• Knowledge: Learning is a central component of knowledge. Here you
learn based on insights and understanding of data and information.
• Wisdom: The highest level is wisdom. Here, reflection is the central
component, as well as being an action-oriented stage. You act based on
the understanding you have.

The key to knowledge is that you use it as a starting point for making decisions.
But there is a huge difference between a decision-making process based on
information or knowledge, especially when it comes to tech solutions.
Decisions that are based on information can only be mechanical, which
means that an information-based system reacts to data and information. If a
given situation occurs, then the system does something. Many of you know
this as an “if-then-else” scenario, such as “If the customer is about to leave
our website, then show them a pop-up message. If they are not, don’t show
the message.

23
 AI FOR CEOS

However, the system can only do what it is programmed to do. Hence the
mechanical part of it. All decisions are made by rules that are coded into the
system. The system does only actions that are defined by rules and the code
that specify the intentions of the rules.
This is not the case if decisions are made based on knowledge. Then, you
relate to what you see and decide on the best outcome. In the weather station
example, we interpreted what we saw as weather data and took actions based
on it, such as wearing a T-shirt and removing the garden cushions.
So, the starting point for making decisions is rooted in different places in
the knowledge pyramid. IT makes decisions based on the information layer,
and AI decisions originate from the knowledge layer.

The difference between the two approaches means that the question “Can I

24
 WHAT IS AI?

wear a T-shirt today?” would be answered with entirely different logics. Both
technologies use the same data and thus have the same underlying conditions.
But the way the two systems reach an answer to the question could not be
more different.
If an IT system was to answer that question, it would look at its data. What
is the temperature? 25 degrees. Is there a rule that describes the appropriate
intervals for T-shirt wearing? Yes, there is. There is a software code that says
T-shirts are ok between 20 and 29 degrees. So, the IT system would respond:
Yes, you could wear a T-shirt today.
The reasoning in AI is entirely different. It would seek to answer the question
by looking at the characteristics that apply when people wear T-shirts. From
that, you would then get the answer:
No, you shouldn’t wear a T-shirt today. A windbreaker is a better choice.
You wouldn’t know why it would answer that a windbreaker would be better.
But in this made-up example, we can say that it’s because your AI looks for
the most robust patterns that explain why people wear T-shirts.
Maybe it is the temperature. But it could be that it was also very windy today
and that the system has learned (via observations of people’s actions) that the
windbreakers beat T-shirts in such weather conditions.
The devil’s advocate would say that it’s just a matter of making the quality
and number of rules in an IT system good enough. Then, they would be able
to get to the same level of response as our AI did here. And that’s right; you
can if the world you look into can be described mathematically, and you know
all the possible outcomes. The problem is that it is rarely the case, and then
you can’t make the rules that are the prerequisite for your IT to solve the task
well enough.
The requirements and conditions for using either AI or IT will become a
recurring discussion as AI-based solutions gain momentum. You now know
the basic premise of choosing between the two, namely whether the starting
point should be an information-based system or a knowledge-based system.
This is a critical discussion that will be frequently addressed in this book,
especially in relation to the assessments of AI’s relevance as an alternative to
IT solutions or existing processes. We now finish off this talk with the logical

25
 AI FOR CEOS

question: Is AI IT?

IS AI IT?

The discussion of data, information, and knowledge naturally open up the

question:
Is AI IT?
IT stands for Information Technology, and we have just learned that AI is a
knowledge technology. So, where does that leave us?
I hope you got a sense when you read the opportunities that AI can create for
you, that it is something beyond what IT is capable of. But how is AI different
from conventional IT? You can choose three different approaches:

The three different angles are illustrated in the picture above. There is no
professional consensus on which of the three is the right approach.
But, hopefully, I’ve managed to convince you that AI is so different from
IT that you choose to look at it as an independent technological discipline,
because what one can do with AI professionally, the way AI is created, and the
philosophies behind the thinking are so different that you are best served by
viewing AI as something other than IT.

26
 WHAT IS AI?

So, that’s why AI and IT are treated as two different elements in this book.
You will see examples of IT and AI working together, but they do so on different
terms, such as AI making decisions on what an IT system should do, for
example, send out emails. So, the two technologies coexist, but they have to
be seen as different disciplines. AI’s potential is at a higher level because AI is
fundamentally based on automated processing of knowledge while IT is based
on automated processing of data and information.
And how far AI could come as technology is the subject of the next section,
which is about the future of Artificial Intelligence.

The perspec ves of Ar ficial Intelligence

A fair question that anybody could ask you as a manager is:

To which extent should we use Artificial Intelligence in our business?
You’ve probably thought of the question yourself, maybe the concern comes
from your employees, from your owners, or perhaps your Board.
To be able to answer such a question, or other variations of it, you must have
a sense of how far the possibilities of technology are.
So you can place your potential AI usage on a scale and set out what your
goals could be.
We will work much more structured with that issue, among other things, in
the strategy section of the book, but you are already now getting a framework
that can help you answer the part of the question that relates to how far the
technology possibly can evolve, and what that would mean for us as humans.
These are big topics, so what’s following now is a concentrated dose of the
future scenarios that the experts can agree on.
The Scientific community generally works with three levels of future
Artificial Intelligence11 .

11
Superintelligence by Nick Bostrom is one of the books that work with this perspective.

27
 AI FOR CEOS

1. Artificial Narrow Intelligence (ANI )

2. Artificial General Intelligence (AGI )
3. Artificial Super Intelligence (ASI)

What does it mean to be intelligent?

Before we go into the meaning of the three levels, we briefly return to the
definition of Artificial Intelligence, which goes thus:
Artificial Intelligence is not human intelligence. In other words, it is intelligence
made by machines.
That definition implicitly states that machines can reach the same level of
intelligence as humans have. The difference between artificial and human
intelligence is defined by the source, namely whether the origin of intelligence
is machine intelligence or human intelligence.
The question is, what is it that makes people intelligent, and how close can
computers get to the same kind of intelligence we as humans possess?
For example, a Mensa test says something about your logical and mathe-
matical skills. You get a score that places you on a normally distributed curve
with an average of 100. If you score over 130, then you can join the Mensa club.
That’s possible for less than 3% of the world’s population12 .
However, it is probably possible to program a computer with the same
computing power as your calculator to answer the Mensa questions faster
and better than any Mensa’s members can.
But this does not mean that your calculator is more intelligent than the
Mensa members are. It just means that it is better at decoding complex
information and that it is good at detecting patterns and connections.
If we are to consider what makes us intelligent as human beings, then we
must look much more nuanced at the concept of intelligence.
Let’s take the musician Madonna as an example. I think most people would
agree that Madonna is an intelligent performer. But what is it that makes her

12
You can read more on criteria for joining Mensa on Wikipedia: https://en.wikipedia.org/wiki/-
Mensa_International or on their website https://www.us.mensa.org/learn/about/

28
 WHAT IS AI?

wise?
At the risk of offending Madonna fans, it’s fair to say that she doesn’t have
the best vocals in the world. Not that she sings badly. There are just many
other vocalists that are better than she is. There are also better dancers than
Madonna; there are better songwriters, etc.
But few musicians have consistently been at the top for as many years as
her. It could be because it is the sum of what Madonna can do that makes her
intelligent. She may not be the best at any single discipline, but the sum of
what she can do makes her unique and intelligent.
It is wildly recognized that there are more than 100 dimensions on which
you as an individual have an intelligence13 . In reality, intelligence is perhaps
the wrong word to describe what we are looking for. Intellectual capacity is
perhaps an even better way of describing the capabilities that we seek.
The relevant questions in this context are: How well do you function socially?
Are you musical? Are you good at dancing? Do you have empathy?
We are not all equally good at all such disciplines. But it is the sum of what
we can that makes us intelligent. And not least, our ability to combine different
dimensions is what makes us intelligent. The more dimensions we can do well
in, and the greater the robustness we gain in our intelligence, the higher our
intellectual capacity becomes.
So, if the goal is to make a machine as good as us, then the machine has to
master intelligence on multiple levels simultaneously. Just as we do, otherwise,
you’ll never get a robot that becomes as good a performer as Madonna, if that’s
your goal.
The scientists working on the future artificial intelligence use human
intelligence as a benchmark. They thus define Artificial Intelligence as
something that can be below the human level. At the human level, or above.
That is why there are the three levels that I start by mentioning, namely:

13
Kevin Kelly is a futurist and best-selling author. He explains the concept well at this TED
talk. https://www.ted.com/talks/kevin_kelly_how_ai_can_bring_on_a_second_indus-
trial_revolution/transcript

29
 AI FOR CEOS

1. Artificial Narrow Intelligence ( ANI )

2. Artificial General Intelligence ( AGI )
3. Artificial Super Intelligence (ASI)

Ar ficial Narrow Intelligence

The fact that AI in the first stage is “Narrow” in its intelligence definition
means that technology has mastered few intelligence dimensions.14
For example, your GPS is better than you to navigate. Your calculator is better
at math than you are, etc. However, AI in this type of Artificial Intelligence
class does not master all the dimensions of intelligence that a human does.
It is this limitation that will keep the technology from being able to take over
and automate many types of human jobs and tasks. Although the technology
is too narrow in its intellectual capacity, there are some niches that technology
will master better than humans.
Take the example of the Google booking bot that could book meetings. It
is a case where Artificial Narrow Intelligence outperforms people, processes,
and IT systems in a particular task.
The more dimensions of intelligence that we include - or the more capabili-
ties that are equal to human capacity- the closer the machines come to be able
to perform tasks that would otherwise only be performed by humans.
It is not enough that AI is good at counting; AI must also be able to function
in a social context, be able to understand emotions, etc. In other words, build
layers of human-like traits in the way the AI works.
If you succeed in developing AI in that direction, then AI will at some point
catch up with humans.
The consequence will be that we will develop an AI that will be able to match
human intelligence. This stage is called Artificial General Intelligence.

14
The three concepts of intelligence are explained by Max Tegmark in the book Life 3.0

30
 WHAT IS AI?

Ar ficial General Intelligence

The level at which AI’s intelligence and human intelligence are at an equal
level is called Artificial General Intelligence (AGI). Machines at this stage have
the same intellectual capacity level as we humans have.15
So, all that Madonna can grasp or any other human being can, so can the
machines.
From a commercial perspective, this means that there are no tasks that
employees need to solve anymore. Machines can perform any tasks that a
human can. So, why let people do any of them if robots can make any task for
“free”?
The consequences of reaching such a stage are colossal. If machines are
capable of solving all the tasks that humans do today, why would anyone work?
All value and everything we need can be created for us automatically.
The topic is so huge that you can find university professors that specialize
in answering just this question.
If this aspect of AI has caught your interest, I have put together several
links to good books and blogs on the subject which you can find on the book’s
website: cognifirm.com/book.
Of course, researchers are debating whether or not we want to reach a level
where Artificial General Intelligence is possible or if it is technically possible.
For example, AI at this level will require robots to get creative in ways that
make it possible for them to write a better song than “Bohemian Rapsody”
and paint a better picture than Picasso’s “Femme Assise”. And can they do so
without having the emotion, soul, and life experience?
Some scientists believe that we will reach the Artificial General Intelligence
stage. They discuss the time frame, which they believe can range from 10 to
30 years from now. Of course, there is also a skeptical group that believes that
robots will be smarter, but that humans always will have dimensions that can

15
This Wikipedia page explains the concept in more detail and has links to more literature on
the subject. https://en.wikipedia.org/wiki/Artificial_general_intelligence

31
 AI FOR CEOS

never be copied by a machine.16

Ar ficial Super Intelligence

Let’s say we reach the Artificial General Intelligence stage. What happens
then? Robots can now do everything we can. But that doesn’t stop there. If we
can make a robot that is 100% as good as the best creative person, why not
make it 10% better?
You can ask that question on all the 100 intelligent dimensions. And the
answer will be that of course; we would make the robots better than humans
if we could. We do that now with the AI solutions that we have already created.
The difference is that they may only be great in 2 of the 100 intelligence
dimensions.
The consequence is that we get robots with characteristics that surpass
our own human intelligence. This stage is called Artificial Super Intelligence
(ASI)17
ASI is an intelligence that transcends the human level of intelligence. The
robots will continue their journey to become more intelligent with high speed.
As with all other digital technologies, the development is happening with an
exponentially growing curve.
All those who believe that we will achieve Artificial Super Intelligence
also agree on the statement that Artificial Intelligence will be the last great
invention that humanity will ever create.
Because when we reach the ASI level, Artificial Intelligence will be the source
of all future significant scientific progress.
Humans will be inferior to the technology, and therefore all technological

16
This Forbes article elaborates on what could drive us towards Artificial General Intelligence,
and what could inhibit us from reaching this stage, https://www.forbes.com/sites/cognitive-
world/2019/06/10/how-far-are-we-from-achieving-artificial-general-intelligence/
17
Vincent C. Müller is a Professor of Philosophy from Eindhoven University of Technology, and
Nick Bostrom from the University of Oxford made an interesting expert survey on when we
can expect to reach Artificial Super Intelligence. It is called “Future Progress in Artificial
Intelligence

32
 WHAT IS AI?

advances will be driven by ASI. There is simply nothing humans could do that
machines would not be able to do in a better, faster, and cheaper way.
The most recognized researcher in this field is the Swede Max Tegmark,
who in the book Life 3.0 describes the journey between these three stages. His
perspective is mostly philosophical. The book is highly recommended if you
want to get more into understanding this element of the AI technology.
It is clear that the implications of having a technology at the Artificial Super
Intelligence stage are enormous.
If we looked back at the Artificial General Intelligence stage, here people no
longer needed to work. All our tasks were solved for us. With political will, we
could eradicate poverty, all diseases, and all climate issues. So, the potential
implications are enormous.
However, reaching Artificial Super Intelligence means the machines are
starting to invent things we would never have been able to do ourselves. There
will be new forms of energy, new computers, and new materials that we
would not be able to understand how they function. The consequence is that
humanity is no longer the smartest on earth.
Although there is disagreement among scientists whether or not Artificial
General Intelligence will ever be reached, they almost all agree on how AI
development can be described. It is shown in this graph, which illustrates the
evolution of AI technology over time.

The way you should view the graph is that we are now on a journey toward
Artificial General Intelligence. If we reach Artificial General Intelligence, then

33
 AI FOR CEOS

there will be an exponential development toward Artificial Super Intelligence.

The theory behind the dynamic will drive the development in this: When we
reach Artificial General Intelligence, then the progress continues (very steep).
Therefore, Artificial Super Intelligence is achieved relatively shortly after that.
The development will not stop there either; it will increase at a very high rate
even after reaching Artificial General Intelligence.
My view on this development theory is that if we ever reach the Artificial
General Intelligence stage, then the Super Intelligence stage comes with
certainty. But the question is whether it is realistic to reach a level where
a machine can do everything a human can. And the question is also whether
this is what we actually want?
We don’t want a robotic version of Madonna. What we want technology to do
for us is to help solve tasks for us. The tasks must be increasing in complexity
and must also be as human-like as possible. Meaning, we want AI to solve as
difficult tasks for us as possible because this would create the most value for
us.
The thing is this. We don’t need to create a human-like intelligence to be
able to drive a car from A to B, the same way a human operates a car, using a
steering wheel and pedals. Or to create a human-like intelligence that operates
in precisely the same way a surgeon does it.
Driving a car and performing surgery are both tasks that potentially can be
solved by AI. The point is that technology should not copy the precise way we
solve and approach a task today. The technology needs to be able to achieve a
better result on the terms that are best for technology. The result of the task
that it performs, not the way, is essential.
From a business leadership point of view, the idea of AI and how you solve
tasks and the creation of value should be of core interest to you.
As you may remember from the introduction, I promised you a new frame-
work for how to use AI in your business. This framework will help you
understand where AI makes sense in your business, and how you can work
with the technology to create value with the new terms that AI can create for
you.
The foundation of this framework is based on how you create value in your

34
 WHAT IS AI?

organization because AI has its greatest significance in the value-creating

processes.
The two parts of the book that deals with strategy and management are
based on how you create value in your organization. However, before we get
to this, you need to have a new and different perspective on knowledge.
Twenty-five years ago, the word “landline phone” did not exist because
there was no reason to state that a phone was anything other than a landline
phone. At that time, there were no cell phones, so any phone was a landline
phone. It was only when there was an alternative that we began to point out
that there were landline phones and mobile phones.
Today, your company’s knowledge is the sum of your employees’ experience,
but not for long. Because soon, your company will work with two different
kinds of knowledge. The intelligence of your employees and your company’s
Artificial Intelligence or digital knowledge.
You have never had to think of your business as having intelligence and
knowledge before. Until now, the knowledge of your company has been the
same as your employee’s intelligence.
This will change with AI. You now have to reflect on how you work with
knowledge and intelligence in your business in a new way.
So, before we start with the frameworks that will help you get more specific
about how AI can be used to create value for you, you need to focus on one
more thing.
You should reflect on the consequences of having to deal with the fact the
intelligence in your business now also could be artificial.
For you to be able to do so, you must build a system that holds the knowledge
you create in your company just as a human being has a central brain that
holds all our knowledge.
Then, you have to think about how you want to use your company’s digital
knowledge. So, it can be activated best, in ways that it can solve as many and
as complex tasks for you as possible.
That’s what the next section is about, and here you are presented with the
foundation of what I call a Knowledge System and your company’s digital
intelligence.

35
 AI FOR CEOS

AI as a Knowledge System

This section is essential because this is where we start by looking into what it
means to work with AI in your business. Meaning, how do you actually use
digital knowledge to solve tasks for you?
The curtain van case from earlier, where a customer called and talked to
an AI robot can be used as an illustration of how a systematic way of working
with AI would be like.
If you look back at the various user interfaces that this AI had, you realize
that they span quite a lot from a function point of view.
There was not just one AI application that solved the task of handling
customer bookings. There was one AI application that was designed to
understand the spoken English language. One to decode the meaning of the
customers intent. Another application was used to construct the answers to
the customer and yet another one that could actually speak to the customer in
English.
This use of multiple AI functions is the typical case. AI functions rarely in a
stand-alone context. In the Curtain van case, AI becomes usable because AI
can engage in a dialogue with the customer and book a meeting with him. In
this case, it means that our AI must be able to:
• Understand human speech: The application can decode spoken English
into written English
• Decode the meaning of messages: Our AI knows how to understand the
intentions from a text
• Perform logical actions based on interpreted intentions: Book calendar
time
• Formulate responses based on interpretation of intentions: Give feedback
to the customer
• Speak English: Formulate answers in English that present the response to
the customer
Let’s return to the talk of the development of Artificial Intelligence over time.
Remember that this theory tells us that the more human-like dimensions AI
masters, the smarter it will be; because the more elements that are closer to

36
 WHAT IS AI?

human behavior (speech, language understanding, reflection, etc.), the more

meaningful tasks AI can solve for us.
This example where AI has to engage in a meaningful dialogue with a human
is a good example. AI only becomes useful when it works as a system to solve
a task. The system consists of several AI applications that work together in
solving the assignment. In this example, they handle a booking.
So, it is a goal for you to design your AI applications so they can work together
in a system to solve tasks. There are several great reasons for this.
First of all, there is a practical reason. If all your AI is organized in one system,
then you don’t have to invent a new voice component for your intranet if you
already have one on your website.
So, the components need to be designed so that they can be reused in
different contexts. It is smart and rational. But that is not the main reason
why you would like to see your AI as a coherent system.
The most important reason why you should wish to see your AI as a unified
system is to have the opportunity to work with your company’s overall digital
intelligence.
The idea to build a cumulative intelligence for your company is that it can
be activated in many different situations, in a way, where it follows the same
growth patterns as were illustrated in the previous chapter.
I name the system that gathers your knowledge in such a way as a Knowledge
System.
It is named a Knowledge System because the goal of the system is to gather
and systematize your overall digital knowledge. So, you can work and activate
your company’s knowledge through a coherent system.
So, when you start doing a new project similar to the curtain van case in
your company, then you solve a specific problem with the AI that you use in
that project. But at the same time, this new AI becomes an integrated part of
your companies total AI set up. So, each new project contributes to building
up your company’s sum of digital knowledge.
You want to make the knowledge and the components that the curtain van
has created available for other parts of your business too. It is convenient
because you then only have to build one chat robot. The next case where you

37
 AI FOR CEOS

need a chat robot can then base its solution on what you have already built.
But more importantly, the knowledge you create can form the basis for
new knowledge-based solutions. You are now building something new—your
company’s digital intelligence. The more digital intelligence you make, the
more you can use it for.
It’s very fundamental to think that way when working with AI. This thinking
can be traced back to the weather station example. You may remember that I
said that the AI version would recommend a windbreaker as a clothing choice,
not a T-shirt. It did so because it saw patterns in user behavior that indicated
a windbreaker was a better suggestion.
The fact is, we do not know what the basis for AI decisions is; but we know
that the more and better data AI has access to, the better conditions AI has to
make more qualified choices for us.
This is central when looking at AI in your business, to enable AI to handle
even more complex and complicated tasks. The more knowledge you have,
the more AI algorithms that can work together, the better solutions you get.
It is the sum of the knowledge that you create that will be a cornerstone
of your company’s overall digital intelligence. The sum of your company’s
knowledge in one system is what enables you to build your company’s digital
intelligence.
So, the rest of this section of the book is about how to build a Knowledge
System because the Knowledge System is the foundation of how well you can
use AI in your business. When that part is in place, we can focus on how to
create value with AI on a more practical level.

What is a Knowledge System?

You need to look at a Knowledge System as a way to build your company’s

digital intelligence.
You build Artificial Intelligence in precisely the same way as human intelli-
gence: as layers of competences. By doing so, you will create a system that
becomes gradually closer to human-like intelligence.
However, your goal isn’t to build an intelligence that matches what a human

38
 WHAT IS AI?

can do. The purpose of the Knowledge System is to create a foundation for
solving tasks in your business. The better your corporate digital intelligence
is, the more types of tasks it can solve for you.
So, precisely the same evolution patterns human follows when we grow up.
As a child, we learn math; we learn to read, we learn to write. They are basic
skills that enable us to solve ever-demanding tasks. The evaluation is possible
because skills are continually being built on top of what we are already capable
of doing.
The curtain van is an example of a solution that could have been designed
as a Knowledge System. After all, there wasn’t one AI feature that solved the
task. It was a series of interconnected AI applications that, together, allowed
customers to book a visit by talking to the company’s AI robot.
The way a Knowledge System is designed follows a little the same basic
principle as you probably remember from the Knowledge Pyramid. The system,
like the pyramid, has four layers:

1. Data
2. Business Logic
3. Intelligence as a Service
4. Function layer

A Knowledge System also starts with data and build more value on top of it.
The difference is that where the Knowledge Pyramid ends up having wisdom
in its upper level, you have a functional layer in a Knowledge System. It is the
features in the functional layer that solves your business problems. They do so
by using the AI components and the knowledge that your Knowledge System
consists of. Here is how.
It is in the functional layer that you can create chatbots that can write and
talk to your customers automatically. It is in the functional layer that you
can extract advanced analysis of your customers’ behavior, and it is in the
functional layer that you can get AI to manage how you most effectively handle
your inventory.
And every feature in the functional layer draws on a variety of AI features

39
 AI FOR CEOS

and data that is found in your Knowledge System. Some AI features will be so
general that they can be used in several different contexts, for example, the AI
application that can speak English. Others are designed for a specific purpose.
But they all exist in a coherent system.
That is why it is called the Knowledge System because that is exactly what it
is. It is a system that holds knowledge—your company’s digital intelligence.
The four components that make up the system are data, business logic,
Intelligence as a Service, and the functional layer. Each of them has a purpose.
To know how to work with AI in your corporation, you need to know a bit more
about how it works.

Data

At the center of the Knowledge System is data. Data is the source and fuel of all
AI learnings. What you need to keep in mind here is that AI looks at patterns
in behavior and looks at how data can be used to explain a given behavior.
Therefore, you would like a lot of data, but more importantly, you get data
from different sources. Keep in mind that the goal is to understand what can
best explain a given behavior, and the more angles in the form of various data
sources, the better your chances of detecting useful patterns in your data.
If you are a banker and an average customer has 100 account transactions
per account per year, then there is value in analyzing 100,000 customers’
accounts movements. But you would probably understand your customers
better if you have 10,000 customers account movement, 1000 customers’ geo-
locations from their mobile phone, and data from 1000 credit ratings. Because,
overall, it offers more opportunities to see the patterns that can best explain
the behavior you want to uncover.

Business logic

Business logic refers to the functions that AI can perform for you. You might
have heard of the term “AI applications”. Most of them can be found in this
layer.

40
 WHAT IS AI?

The business logic area is where AI solves specific tasks for you. Some of
the tasks will be generic, for example, translating handwriting into digital
text. Others are developed for a particular purpose. It could be to analyze your
product data.
So, this is where your tasks are solved. However, most of your business
problems are too complex to be solved by a single AI application. They almost
always require the use of multiple AI applications.
There are seven different types of AI business logics, or application types.
We will go through them each at the end of this section to give you an idea of
the specific tasks that can be performed by AI.

Intelligence as a Service

When developing software solutions, it is usually good design practice to create

features so that they can be reused in other contexts.
This is done so that you do not have to code the same function several times.
This principle gives you a more straightforward and more transparent IT
architecture, which is also cheaper to maintain, troubleshoot and develop.
The same arguments apply to AI applications in a Knowledge System. Here,
it is even more important to design solutions so that knowledge from your
different applications in the business areas can be reused in new contexts.
Your company’s digital intelligence is the sum of your data and the knowl-
edge that your AI applications in the business team have created, and you want
to make this as accessible as possible.
Sometimes, a market chatbot and an efficiency chatbot will need to request
the information in your Knowledge System. That is, they need to draw on the
same AI business logics and learnings.
That is why an Intelligence as a Service layer must be developed. This layer
allows different components to query for the same AI business logic in various
Use Case scenarios.
The four different layers can thus be illustrated as an onion with data in the
center and a higher degree of knowledge enrichment the further you get out
from the core.

41
 AI FOR CEOS

The feature layer

The feature layer is where the actual value is created for you. This is where
things come together, and your tasks are solved using all the relevant AI
functionality, data, and knowledge found in your Knowledge System.

You can create value with AI in the feature layer in three different ways:

• An Assistant: AI can be an assistant who tells you (proactively) issues

about your business that you were not necessarily aware of. The use of AI
in business intelligence is an example of an AI assistant.
• Automation: AI can take over the management of several functions in
your business. For example, sending mail, controlling production flow,
etc.
• Bots: Allows customers and employees to access knowledge in a commu-
nication format that is intuitively understandable to humans. This is done
either through a written dialogue as a chatbot, or by talking to you like
you know from e.g. Google Home.

The way you can use a Knowledge System is illustrated below. It shows how
the system, based on data, enables you to solve the tasks in the functional
layer.

42
 WHAT IS AI?

The three areas can either be used together or separately. The Curtain van case
is an example of how a voice message is combined with automation (mails
and calendar integration). As you may recall, you could call a bot that could
have a conversation with you, and book a time to measure for new curtains at
your home.

Assistant

An assistant works by giving you insights and knowledge that you would not
otherwise have the chance to acquire. The simplest type of assistants is known
from the Business Intelligence world. In Business Intelligence, you draw
insights from data sets by running several AI models, and then get results
about patterns in a population.

Assistants thus help us deal with complexity and see patterns we cannot see
ourselves. In this way, they can, for example, help us do analyzes and make
suggestions for product designs. Building products with AI might sound
strange to you, but I will tell you why this is a way of using AI as an assistant
later in the book.

43
 AI FOR CEOS

So, assistants are AI technology that works together with humans to accom-
plish specific tasks.

However, it is in a different way of work, compared to what you have been

used to until now. When AI works as an assistant, then you should really look
at it as a kind of virtual colleague with whom you work. An example explains
how and why.

In 2017, Google’s Deep Learning won the world’s best Go player. Go is not
so known outside Asia, but is the world’s most complex board game. There
are more gameplay combinations in Go, compared to Chess. It is therefore
considered the game that is intellectually the most difficult to master.

So, it was a big deal for the AI researchers as Google’s AI when they finally
could become the world’s best Go player.
As an example, this is interesting, but basically, it does not say anything
about how to work with assistants. The next part of the story, however, does.
The best Go players in the world today is not a Deep Learning application
from Google or any other tech giant. The best Go players are teams consisting
of both AI applications and excellent Go players working together.
The same is true in Chess. The best chess players today are not individuals or
AI applications. These are teams made up of both people and AI applications.
This view is also shared by the chess genius Garry Kasparov.

The extra exciting thing about the Chess example is that the world’s best chess
team does not actually consist of top chess players, nor does it use the best AI
algorithms available. The best team consists of outstanding chess players and
a good (but not the best) AI chess applications.
The lesson from the two cases is that working with AI and people in teams
is the winning combination.

Because such teams can draw on the best of the AI world with the best human
experience, this combination produces the best results. Because, overall, they

44
 WHAT IS AI?

understand how to use AI where it is great, and the human dimension, where
it works best.

The keyword here is the synergy effect. For that is precisely where the great
strength lies when you use AI as a team “colleague” who at some points will
be infinitely far ahead of you, and in others will fall short.

As technology, assistants are used in two very different contexts:

1. Analytics: Assistants are used for better decision making through

predictive analytics
2. Product Development: Assistants are used to make designs for new
products.

Analy cs Assistants

To understand why Assistants are good at analytics, we start by briefly

specifying what the goal of doing analytics is. Performing analysis is never
an end goal. The purpose of the analytics is always to provide insights so that
qualified decisions can be made.
The way you have traditionally worked with analytics follows a process that
could look like this:

45
 AI FOR CEOS

Let’s say the goal of your work is to find out what marketing efforts you need
to initiate to sell ten more shoes on your website in the next week.
The way you usually would solve such a task would be by letting an employee
use one or more analytics tools to look at your data (CRM, Google Analytics,
Facebook, etc.). They would do this to understand what drives your sales.
Based on the insights the tools give them, and their professional knowledge,
they come to what they think are the best ways to sell the extra pair of shoes.
Thus, an analysis-driven hypothesis.
You then act on their knowledge by following their recommendations. Then
you wait and see what happens with the sale. Maybe the ideas were successful;
maybe they weren’t. So, you have a learning loop. Then you repeat what
seemed to work, just as you may have been inspired to try new things.
Thus, you are working toward achieving your goals that were to sell ten
shoes more on your site. A large part of the success criteria for achieving your
goals is thus tied to your employees’ competencies, that is, their ability to
understand the data they see, their ability to put knowledge into actual actions,
and their ability to learn what your results are tied to.
That’s not how AI-based analytics work. AI analytics follows a process as
illustrated below.

46
 WHAT IS AI?

In the illustration, it looks like AI does all the work for you, and that is
almost right. The idea here is to take advantage of the fact that AI is good at
understanding patterns in a context. This fact can be used to make AI create
action suggestions for you.
What I am about to describe to you is called predictive analysis. Predictive
in this context means forecasting, and that’s exactly what AI does for you.
This is how it works. You start by setting your AI a goal. Then you ask your
AI to tell what the best way to achieve the goal would be. In this example, you
would ask your AI to advise you what the best marketing effort would be if you
were to sell ten extra pairs of shoes next week.
Your AI would then look at patterns in user behavior. Who bought from your
shop and who didn’t buy? If the group that bought should be increased by 10,
what would the cheapest way to do it be?
So, your AI gives you a plan on how to best achieve your goals. Your AI could
tell you that the way to reach your goal would be to post 100 extra ads on
Facebook, targeting women who are between 20 and 30 years old. Or what
else it could see would drive your sales most effectively. So your AI predicts
that if you do this, then it predicts that the consequence would be this.
In this example, you use this mechanism to set a goal:
If we want to achieve this, what should the effort be?
Your decision driving knowledge is now founded in your knowledge system,
not with your employee. You do not know - and maybe you don’t need to know
- what lies behind the proposal that your AI is making for you.
Using AI in this way does not require time or competent analytics trained
staff. You are now being presented with suggestions for actions. In this way,
you do not need a professional person to interpret and analyze data for you.

47
 AI FOR CEOS

This is why the flows in the AI drawing are so much simpler than the first
example.

Assistant to help you make new products

There is another domain in which Assistants can help you perform better too.
The principle that you have just learned where AI is used to give you a
solution to a problem (more sales) can be used in a very different field too:
The creation of new products.
A real-world example explains how.
A subcontractor for Airbus that also makes drones, used AI to specify how
their new drone should be built.18 .Their approach was exactly the same as in
the analysis example from before. Meaning, they defined the success criteria
or the goals of the drone building project. They wanted the AI algorithm to
design the frame for the drone. The frame should be made in fiberglass. In
this case, they were quite practical, so the requirements were something like
this:

Solution Requirements:
• The drone should be able to handle a weight of 1000 grams
• The drone should be able to fly for 20 minutes
• The drone should be able to operate in weather conditions from moderate
to strong breeze

Preconditions:
• The drone was to be built with fiberglass
• The solution should be able to carry four motors of a specific size and
weight
• The solution had to weigh a maximum of 250 grams

18
https://www.bbc.com/future/article/20181129-the-ai-transforming-the-way-aircraft-are-
built

48
 WHAT IS AI?

They then asked the AI algorithm to design a drone with the requirements
that could fly as quickly as possible. The result was a 3D drawing as seen here:

https://www.youtube.com/watch?v=odHC-gxJhG4

The process of this kind of product development has similarities to the Chess
example too.
It is very likely that the best new drones are not built by AI, or by humans. But
the combination of humans and AI can reach results at new levels of product
development.
The company also asked the AI algorithms to design a new supersonic
aircraft for Airbus. However, they have not built it. But they have fitted some
new suspension components that were used in the supersonic aircraft design
into the production of real Airbus aircrafts models. The parts were designed
by AI, and they are stiffer and lighter than what has previously been possible
by Airbus to manufacture.19
In this example, the AI algorithm built a physical product.
If you can specify what the important characteristics of your product should
be, and if you have data that allows for pattern analysis, then you will be able
to make AI create a suggestion for what it might look like, either as a complete
product (the drone) or a new component (the aircraft part).
If it is possible to design a supersonic aircraft with AI, then it is most likely
also possible to create any other kind of physical product too. We will get into
more details on how you do this soon.
Both the analysis example and the product development case are based on
the premise that AI is good at seeing patterns and connections in data. This is
what allows AI to predict the best actions to achieve a given condition.
The product development case has one more dimension built into it. In order
to make qualified suggestions for new product designs, it requires that the AI
has an understanding of the physical properties of the product material. So,

19
This page shows how the part was designed https://www.autodesk.com/customer-
stories/airbus

49
 AI FOR CEOS

the AI must have learned how fiberglass can be used in a product, how to cast
it, it must know the quality requirement for fiberglass so that it can come up
with suggestions that will work in real life.
So, the bar is higher to get started if you want to use AI to make new product
designs. But once you have an AI that understands how to work with fiberglass,
as in our example, then it is easy to get many more new product designs that
are based on that material. The initial cost of starting up is high, but it becomes
almost free to get as many new product designs once you have the system in
place.

Bots

We all know what bots are. They are either speech bots or chatbots. You know
chatbots from online services where they can provide more or less advanced
answers to customer questions and problems.
The best-known speech bots are Google Home and Apple Siri. What makes
them smart is that they open up a new form of communication with computers
that is much closer to our daily routines and how we live our lives.
You need to think of bots as a new way in which employees, customers,
and other stakeholders can access knowledge in your business in a way where
knowledge is accessible to them on their premises. Knowledge that they access
in a dialogue that is based on their needs and the situation they are in. So, they
get the information they need in a way that makes sense to them.
Thus, bots open up entirely new premises for how you can interact with all
of your stakeholders.
Bots, as we know them now, are just the first part of what will become a
paradigm shift in the way your company communicate. You will go from
digital mass communication to personal AI-driven dialogue. And that will
change everything in what you can do for the people that are most important
for your business.
Right now, your digital communication channels are just that—communi-
cation channels. That means communication from you to your stakeholders.
Communication is a one-way street. You say the same to all of them.

50
 WHAT IS AI?

Yes, you have an Investor section on your website that is written differently
from your online product catalog, because the two areas are designed for
different audiences. But the different parts of your site are not personal. You
write the same content to all your investors and all customers. But AI-driven
dialogue is changing that premise.
Twenty years ago, digitalization made communication with your customers
free. AI now makes the dialogue with your customers free. It changes the
premise of why and how you want to communicate with them.
It is clear that on the customer side, you will get a whole new range of
opportunities. You can bring customers closer to your business in all aspects of
their life cycle if you can have a qualified and free dialogue with your customers.
Potential customers, new customers, existing customers, customers with
an ongoing need for dialogue with your company, and passive customers, all
have different needs that can be solved with an individual conversation with
them.
When dialogue becomes free, then you will also be able to choose to
communicate with your surroundings and stakeholders in new ways.
Take your approach to your markets as an example. Barriers such as
language and time zones will be eliminated with AI-driven dialogue. Your bots
will be able to serve customers in Japan in Japanese around the clock. Or in
Portuguese in Brazil. So, many of the natural bars that have created national
or geographically dependent markets will disappear.
We return much more to how you work with bots, in the operational section
in this book.

Automa on

The last feature component in the knowledge system is automation. Automa-

tion could also be called the management layer because it controls automation
processes and business logic. Here is what this means to you.
Remember the curtain van example? Here, there was a business logic
that decided that an email should be sent to customers who have booked
an appointment. Such logics are activated via the automation layer.

51
 AI FOR CEOS

Either via an integration with their existing mail system, or via a mail system
that is built in as part of the design of a Knowledge System.
Some might say. So, what is new? That is what IT has been doing for us for
the last 20 years. Which is a perfectly valid point. The unique part is not what
AI can do for you, but how it is done.
AI can perform tasks that are much more sophisticated than IT can do for
you. So, you can let AI take over some of the jobs that you need a person to do
for you now because your IT is not capable of solving them for you.
An example will explain the differences between IT and AI in this respect,
and will let you in on your new options.
Let’s say you make physical products like lamps. Your business focus today
is on a consistent quality of the product, optimization of production flow, and
cost reduction.
What you would like to deliver is a uniformly excellent quality of your
products, so that you comply with the legislation and so that you have a good
reputation in the market and avoid time-consuming handling of returned
defect products, etc.
You also want to optimize the production flow so that you make the products
as quickly as possible and send them to the customers without too much
warehouse storage time. This ensures that you do not have too much capital
tied up in your inventory.
Finally, you want to make sure that you make the products with the least
effort, and therefore you automate the production flow as much as possible.
It makes production cheaper.
The result of this set-up is that you produce good quality lamps.
However, all the lamps are basically the same.
There are specific color differences, different models, etc. But you do not
have an interest in having too many different products in your product range.
Too many product variances will create conflicts with the elements that make
it possible to produce your products efficiently.
So consumers get the same lamp no matter who they are. Consumers are
willing to accept a lag of choice and influence on the product design because it
means that they will get an excellent cheap lamp.

52
 WHAT IS AI?

If you let AI take over (parts of the control) of your lamp production, then
you will be able to gain a competitive advantage in a way that gives consumers
more control and more choices without compromising your efficiency in the
way your products are produced.
For example, you will be able to let consumers design their own lamps. They
can choose the color, size of the foot, have personal inscriptions, etc. Anything
imaginable that could be of interest to the creative consumer. And you will
be able to produce such individual lamps cheaply, in a consistent quality and
without unnecessary inventory. So, no compromise concerning your liquidity
or efficiency requirements that was explained before.
All this is possible because you will let AI take over the control of (parts) of
your production flow.
Your IT systems, as they are now, are designed to manage efficient, con-
sistent production. This is achieved because your production software is
based on rules and processes. It is the rules of the system that enable you
to produce efficiently. But it is also the rules that prevent you from producing
an individually designed product. Role-based software cannot handle too
many exceptions and too much complexity, so it is not an option for you.
But AI, which is not based on rules, can handle individual requirements
differently.
AI is based on knowledge and experience. This means that if AI controlled
your production flow, it would be able to handle both efficient production and
that you made individual lamps. So you can serve 1,000 different customers’
1,000 different wishes to create their personal and unique lamps.
Such agility in productions opens up for individual mass-produced products.
And that will change your competitive situation. For that removes the premise
that as a consumer, you can only get good and cheap products if they are
simultaneously mass-produced.
Consumers may not be interested in individual lamps, but all people have
a degree of self-realization and would like to own unique and personalized
products.
The companies that manage to open this flank up to their customers will
be able to produce sneakers, cars, glasses, bikes, mobile phones that are

53
 AI FOR CEOS

personally developed to their individual customers’ unique specifications.

What you need to know about AI so far is that you have three different
components to work with when it comes to AI - you have assistants, bots, and
automation.
But there is a technique layer under each of them that allows them to perform
the tasks that we just reviewed. To really understand what you can use AI for,
you also need to know the technical components that make the three areas
possible. You will be able to do this after reading the next section of the book.

AI technologies

Can you remember the discussion on how to define intelligence, and what
intelligence means?
An important point was that it is difficult to define and measure human
intelligence because intelligence can unfold on many levels - social, math-
ematical, musical, etc. As an individual, our intelligence is the sum of our
various bits of intelligence.
It’s the same way with AI. Here, too, there is no one type of intelligence,
since AI has many different branches.
Some AI solutions could be designed to understand and decode the spoken
language. Another AI could be trained at being able to read written text. Others
again can be used to analyze large amounts of data and look for patterns.
This diversity in the type of intelligence that AI can possess is one of the
greatest strengths of technology because it means that AI will be able to get
closer to the kind of intelligence that we as humans have.
Just as it means that by combining different types of AI intelligence, we will
be able to build a unified system. A system that could perform much more
complex and demanding tasks compared to the IT systems that we know today.
This AI system is the Knowledge System.
A Knowledge System can have seven different types of AI included. You can
see all of them and how they fit into a Knowledge System in this illustration:

54
 WHAT IS AI?

You will learn what each of them can do for you later in this chapter. You will
understand this better if you know how AI works on a general level, so this is
where we start.
As with human intelligence, it is the combination of the many levels of
intelligence that makes the big difference. Getting a route from a GPS is smart.
GPS is already better than you to navigate. But if you combine the GPS route
with a chatbot, then you get an assistant that you can place an order to. Tell
me the fastest way to London. And on the way to London, you can ask your
route assistant about the expected arrival time, when the next gas station will
appear on your journey, etc.
The number of areas using AI is continually increasing. From a commercial
perspective, seven AI technologies stand out. They will now be introduced to
you. Knowing how they work is a stepping stone for you in your journey to
understand how to create value from AI in your business.

Image recogni on

55
 AI FOR CEOS

This technology allows you to recognize objects in images—typical faces or

numbers20 . Image recognition is most common in cases where you use AI to
make products or services better or develop new services.
Later in the book, you will be presented with a case where Stanford uses
image recognition to screen for cancer. Their AI can more effectively identify
cancer cases, among other things because the image recognition part of their
solution can see visual patterns in cells that, in a context with other indicators,
can help diagnose cancer at an early stage.

Text mining

Text mining means linguistic analysis of text21 . I admit that at first, it may
not seem that interesting, but this technology is a core element in much of
the value that you can create with AI.
Text mining is a cornerstone in the set-up that will help you communicate
with your customer in radical new ways because you will use the technology
to understand the meaning of text.
Look at it this way: Whenever someone is communicating a message to a
person, the message consists of the actual words that are spoken (or written),
and some kind of meaning.
Take this example “Let’s go to McDonalds”. This sentence does not mean
that the person literally wants to walk to McDonald’s. It means that the person
would like to visit a McDonald’s restaurant. As humans, we know this.
So, if you need a system to be able to interact with humans, then you would
need to make this system decode the actual meaning of a text in the same way
that a human would.
This is what you use text mining to do for you. So, it is used as an “infras-
tructure” component in set-ups where humans interact with a knowledge

20
If you are a tech expert and what to know how to work with Image Recognition, then I suggest
you read this book: Programming Computer Vision with Python by Jan Erik Solem
21
In my view, the best book on Text Mining is by Charu C. Aggarwal. It is written for developers
and is called “Machine Learning for Text.”

56
 WHAT IS AI?

system so that the system can detect the meaning of text.

At my company, Canecto, we work a lot with text mining to help make our
clients’ websites perform better. We use the technology to read the content of
our clients’ websites. Based on this, our system can make recommendations
content to change on a site if you want users to take specific actions (e.g. buy
a product).
Thus, text mining should be seen as a kind of central AI infrastructure
component that can help decode the spoken language and text. It then allows
other AI algorithms to work with the content and provides qualified feedback
on the intentions of the meaning of the texts being analyzed.
This technology is a key element of Chatbots, and as you saw in our example
from Canecto, it can also be used in products and services.

Natural Language Processing

Natural Language Processing also called NLP, is a branch of artificial intel-

ligence that relates to human and computer interaction, through the use of
natural language. The goal of NLP is to read, decipher and decode the meaning
of the human language in a way that makes sense22 .
So, compared to text mining where the goal was to decipher the meaning of
the text, we are now focusing on what the intentions are. This is, of course, a
much more complicated task.
NLP is a difficult technical discipline because the human language is
challenging to decode the meaning of a digital context. There are so many
nuances in the language. Just as language contains many levels of abstraction,
can be sarcastic, implicit, etc. All such matters can make the decoding of the
actual meanings and intent difficult.
You will never have to develop an NLP algorithm for your business yourself.
There are specialist companies that provide the service as a service that can

22
Natural Language Processing is a topic that you must have a technical background to
understand. I don’t know now any non-technical books on the matter. So, the best tech
book, in my opinion, is “Introduction to Natural Language Processing (Adaptive Computation
and Machine Learning series)” by the Georgia Tech professor Jacob Eisenstein.

57
 AI FOR CEOS

integrate with your other AI applications. It will always make sense to use
such a service as the development costs of making one yourself will be too
high.
In commercial contexts, NLP will most often act as an infrastructure
component. For example, in connection with Chatbot, as previously described.

Natural Language Genera ons

Natural Language Generations can be briefly described as the reverse of

Natural Language Processing. It is the AI technology that makes it possible to
translate text into naturally spoken language23 . In the curtain van example,
it would have been the part of the bot that makes it possible to reply to the
customers.
The range of perspectives with this technology is enormous, and many Use
Case scenarios will emerge because what you have here is a technology that is
capable of creating meaningful content.
Natural Language Generations enables you to create understandable content
based on various sources of data. This makes the technology suitable as a core
component in advanced digital agents. An example explains why.
Let’s say that a patient has been diagnosed with a specific disease. As
part of the treatment process, different roles need a different kind of patient
information to perform their jobs.
So instead of giving all roles the same information, then you can use NLP to
design the output that fits each individual need.
A specialist (doctor) will be able to get an answer that explains, at a high
technical level, the content of the patient’s journal. A nurse will need a
different type of information from the same journal to do their job. And the
patient themselves may be able to be notified of the results in a way that is
understandable (and ethically justifiable) to them.

23
Natural Language Processing is also a topic that is only approached in the literature from a
tech perspective. I find the best book to be “Handbook of Natural Language Processing” by
Nitin Indurkhya.

58
 WHAT IS AI?

In other words, Natural Language Generations will be able to create

audience-specific content. The perspectives of Natural Language Generations
are to go from mass communication (everyone gets the same information) to
an ultra-personal message that adapts the form, structure, and message to
the framework to which users at the individual level respond best.
This technology has the potential to change how you can run your business
on many levels, and I will get back to how to utilize this in more detail in the
book’s operations and strategy parts.

Recommenda on engines

Recommendation engines are a technology that we all encounter several times

in the course of a typical day. You meet the technology on e-shop where it
drives the feature that gives customer suggestions on a similar product to buy.
The technology is also used on Netflix to suggest which next movie to watch,
or which songs on Spotify that are most similar to the ones we have on our
playlist or to find similar books on Amazon24 .
Thus, it is primarily a commercial technology that relates to products and
services.
It is primarily among the seven on the list because of its high degree of use.
It serves a very specific purpose, namely, to find the products or services that
match a given user’s behavior.

Predic ve analy cs

Predictive analytics is primarily a marketing tool that allows you to define

audiences with a given behavior very accurately25 . Thus, the analysis can form

24
The book on Recommendation Engines that I can recommend the most is a part of The MIT
Press Essential Knowledge series. It is simply called “Recommendation Engines”, and is
written by Michael Schrage
25
Predictive Analytics is so widely used that there is now a book available in the “For Dummies
series”. It is called Predictive Analytics For Dummies”, and is a great place to start. It is written
by Anasse Bari.

59
 AI FOR CEOS

the basis for highly targeted campaign activity.

You already know it from the example at the beginning of this section of the
book where it was used to select the best marketing efforts.

Prescrip ve analy cs

Prescriptive analytics moves the bar from predictive analytics one level up.
Now, you not only see what effect is best but also the expected reactions among
those you want to target26 . If you have studied economics, you might recognize
this as the same dynamics found in Game Theory.
In the AI world, it is a great tool, especially in medical research. Here, you
can use prescriptive analytics as a way to screen the effects of medical trials
before they are carried out as lab tests.
So, you would simulate the experiment before starting the expensive
laboratory tests. It is especially smart in this field where only 1 in perhaps
1000 medical products reach the approval stage.
The more trials that can be simulated, the higher the accuracy there will be
in the end because you would have learned that some tests will fail before you
start them in the real world. This means that you can spend your valuable lab
time on research that has a much higher likelihood of succeeding.
This technology removes some of this risk from medical research, and can
potentially speed up the product development time.
It is the seven technologies that collectively perform the elements that define
Artificial Intelligence. These are your Lego blocks on which you can build your
company’s digital intelligence.
How to do it and what tasks you can solve with it will be addressed in the
remaining part of this book.

26
Prescriptive Analytics A Complete Guide is, in my view, the best book to read if you need an
introduction to the topic. It is written by Gerardus Blokdyk.

60
 WHAT IS AI?

How can Ar ficial Intelligence create value?

The next two sections of the book will help you understand how you can now
use the Knowledge Systems, the 7 AI technologies, and the other AI elements
that you have now learned of in a business context.
To set the scene for how you can create value with AI, we start with models.
But this time it is not AI models. It is real models, photographers, and a
photoshoot.
You are about to learn how all of them will soon be seeing fierce competition
from AI because AI can create exponential growth.
If you think it seems weird, then I can fully understand you. It is two very
different worlds: Photos, photographers, models, and AI. What´s the link?
However, I predict that a large portion of the world’s 20,000 professional
photo models soon will have to look for another job, because how they
create value is an obstacle for the realization of exponential productivity
improvements. Here is why.
The requirement for an excellent photo model is obvious. Being photogenic
is a part of the job. However, being photogenic alone does not make a great
model. He or she must also be creative so that they can sell the products they
market in the best way. They must be able to “work the camera”.
The models must also be flexible; they must be willing to travel a lot, to
work for many days, and have long working days. So it is a line of work that
requires a high degree of both social and cognitive skills, and thus something
most people think is very far from what AI can do for us.
So why would a skilled and professional photo model be made redundant by
AI? The reason is economics, time, and technology. But most of all, because of
exponential growth options.
Anyone who has experienced a photoshoot knows that it is expensive. Even
the simplest studio recordings require a relatively large apparatus to get the
images you need.
But this is, in fact, the least problem for the models, photographers, and the
owners of a photo studio.
The big problem for all of them is that they are a bottleneck. You use the

61
 AI FOR CEOS

photos they create to sell products or to brand a business. But, it takes time
and planning to work with a photo shoot.
Let’s say that you create women’s fashion and you have created a new red
dress. To sell it, you need photos. So a photoshoot has to be planned.
It requires you to book a model that matches the expression of your
marketing concept. Also, a studio or a location and a photographer must
be booked.
This whole process is a bottleneck to your marketing efforts. You cannot
make sales material, post on social media, and create online articles or print a
catalog before the pictures are ready.
In other words, there is a significant limitation in the period from when
the clothing concept is ready to when the promotion can start. And the photo
shoot is a critical component of this limitation.
This can be illustrated as a process flow diagram with three coherent streams:
Marketing, production and the photos, which is shown in this picture.

So there is a great business interest in not making photoshoots if you can

avoid it in any way.

62
 WHAT IS AI?

Because if you can avoid it, then you can market the products earlier. Maybe
you can test what works on customers before production starts. And you can
release new collections on an ongoing basis without relying on a planned
photoshoot.
All of these elements are components that will make you achieve exponential
efficiency.
All those elements support the idea of achieving exponential growth. Our
business will have better options if we can design it in a way where we don’t
have any dependencies on having to take actual pictures in a studio.
If we could achieve this, then it could mean that we could create more
collections, that we fast verify if they match our customer needs, and have
more efficient production management.

Here is where AI comes into play. To be more specific, a branch of AI that

provides the opportunity to work with artificial photo creations. It is called
CGI (Computer Generated Images).
As an example of the technology, you can watch this YouTube video. The
clip contains images of artificially created people. None of them exists in real
life:
https://www.youtube.com/watch?v=kSLJriaOumA#action=share
This technology is about to become available on a mass scale. When this
happens, it will mean that you can make all the photos you need with virtual
models.
As early as 2014, 75% of the IKEA catalog was made from computer-
generated images27 . So, it is not science fiction. It is a technology that follows
the normal tech adaption patterns. It starts at the giants such as IKEA and
then becomes more accessible to a broader line of business.
The cost of the technology will be so low that very few companies will ever
be able to justify using models for their product images.
The use of CGI does not mean that the modelling industry is becoming
extinct. But it will clearly mean that the number of global photo models is not

27
https://petapixel.com/2014/08/28/flip-ikea-catalog-75-photography-see-cgi/

63
 AI FOR CEOS

20,000 as it might be today. There will always be supermodels in the world,

but the majority of the models are at a high risk of losing their jobs.
And the point is that the Knowledge Systems do not do the jobs of the models
and the photographers better than they do. It all depends on your perspective.
If you look at the output as a product, then a photo will always be better if it
is made in a studio by a photographer. But if you are looking at the output as a
process, then the winner changes. A CGI photo will always win over a studio
photo because it has no dependencies to the product creation process.
Using Computer Generated Images means that we win calendar time, you
get agility in the way you can act in business. The new production flow is
shown in this illustration.

The companies that use CGI will get the pictures right away and do not have
to depend on photoshoot bookings and production plans. If they need a new
product photo, then they can have it 2 hours later. Not two weeks later, when
the right model, photographer, and studio are available. And in the end, it will
be such matters that determine evolution.
The economic factors and the new level of production flexibility will win
over the photoshoots. Despite its apparent defects, CGI will be the preferred

64
 WHAT IS AI?

solution for many companies.

As you can see, this goes beyond jobs and AI. It has to do with the core of
how you run your business. So, what you have just been introduced to, how
AI can influence jobs and processes is of the core part of the value creation
process.
The goal is not to use AI to replace a job, process, or IT system. The goal is
to use AI to create a different kind of value. A value that is better than the one
you can create with your current set up of jobs, in your processes, and with
your IT systems. So, there is a better way to create a new kind of value.
This case represents one of the core elements of working with AI. For now,
it is just a single case. In the next sections, you will get tools that let you know
how to work with AI to help you identify the best such examples for AI in your
business. So, we go from a single case to an applicable business framework
that also works for your business.

Next steps

You have now been introduced to the basic concepts of AI.

Hopefully, this has made you more curious and interested in how you can
actually work with this new technology. So, you can create new products,
become more productive and gain insights that you would otherwise never
have been able to get.
Knowing the technology is one thing. Knowing how to work with technology
is something else. Knowing where you create the value is what really makes a
difference. The rest of the book is dedicated to helping you understand how to
do this.
You will soon learn more precisely how to find and prioritize the most
relevant tasks that could be solved by AI in your business. Just as you will
learn how to use AI as a strategic driver that has the potential to change how
your business could be run.
You will also learn why there is a connection between the fact that you can
only choose five different colors for your Tesla and AI.

65
 AI FOR CEOS

Before we get to all of that, we have to start in another place, a place that
sums up all the reasons why you should work with AI in your business. AI
changes how you can create value in your business. So, that is where we begin.
By looking at how you create value in your business, so you can see where AI
can help you the most.

66
 3

Opera ons and AI

W
e have now come to the part where you will be able to see how AI
can be used to solve tasks and problems for you.
So, you can decide whether it makes sense to apply AI to your
business and how this could be done. Some of the things AI can do for you are:

• Make it cheaper for you to develop new product prototypes

• Give you an innovative advantage
• Replace your customer-oriented processes
• Automate your customer service
• Increase your production capacity
• Make your business more effective
• Help you make decisions based on complex information
• Make better digital products
• Sell your digital products online more efficiently

As you can see from the list, these are very different areas. Some of them
reduce your costs; others create new offers for your customers; others give
your customers a better experience or increase your productivity.
The common denominator is that they create value for your business in new

67
 AI FOR CEOS

ways. You will learn that some of the initiatives will be more relevant to your
business than others. This part of the book will teach you how to see where AI
provides the most effective value for the way your business is run.

Hence after reading this part of the book, you would be able to identify where
in your business you could use AI the best. You will be able to do this because
you would know in general what AI can do for your business, and how this can
be applied in specific Use Cases in your business.
You might remember from the introduction to the book that I mentioned
that AI can create value, and your business creates value too. And that you will
get an understanding of why the two things are interconnected and what it
means to you and your business. This is what we are going to do now.
The way you will learn this is by following a flow with four sequences that
leads toward the best match between how you function as a business and what
AI can do for you.
The first step in the sequence describes your business and how you create
value. The focus is then shifted to how AI can generate value for you.
One of the reasons for this approach is to enable you to spot the most obvious
ways for you to use AI in your business. Because it makes it easy for you to
compare what you do to what can be achieved with AI.
Another reason is that the framework can be used to identify where your
exponential growth scenarios could emerge and how AI potentially can
challenge the goals of your business. Such considerations and many more are

68
 OPERATIONS AND AI

on a strategic level, so you will have to wait for the strategy section to get that
learning.
Conceptually, you can see this approach as a funnel where you are trying to
mix what you do as your business with the capabilities of AI, as this illustration
shows.

The first step is to reflect on what it means to create value in your business.
I have mentioned that AI creates value as one of the great strengths of the
technology a few times now.
AI has the potential to change the dynamic of how you can create value. In
its most extreme case, this will redefine how your business could be run. You
could even become employeeless as we will discuss in the strategy section. So,
you need to understand how AI can function as a value creation force for your
business.
The dynamics behind AI in a corporate context are all centered on new ways
of creating value. To fully understand how you can work with AI and how it
fits in, you need to start by looking at how you create value in your business
today.
So, this is where we will start. First, we need to have the same view of
the value-creating process in your business. This serves as a foundation for

69
 AI FOR CEOS

identifying how AI can create value and solve tasks for you.
Some of you might want to do the actional analysis that we are about to
engage in on your own business while reading this book. If this is your plan,
then you should know that all the tools you need to do so are tools available
for download at cognifirm.com/tools.
But now, we get back to your business and how you create value.

How do you create value in your business?

Since the start of organizational theory, scientists have worked with different
ways of looking at a business.
You surely know Porter’s Five Forces28 which have a competitive angle on
what a company’s best market position would be. Another example is Blue
Ocean Strategy29 . That model offers an approach that helps you to identify a
unique value proposition for your business.
Both approaches can be used to describe your business. However, their
purpose and perspective are different. So, it makes no sense to discuss whether
one is more right or wrong than the other.
They work with different prerequisites and seek to describe different
conditions on the same subject, namely your business and how it is run.
Basically, they are just two different approaches you can choose to look at
something as complex as a business.
It is with such a mindset that you need to see the theory that will be presented
to you now. The focus now is not on your competitors or on your profits. This
theory will help you understand:
How do you create value in your business?
Because that’s basically what AI can do for you. The technology can change
the premise of how you create value. So, we need a way to describe the value

28
Competitive Strategy: Techniques for Analyzing Industries and Competitors by Michael E.
Porter
29
Blue Ocean Strategy, Expanded Edition: How to Create Uncontested Market Space and Make
the Competition Irrelevant by W. Chan Kim

70
 OPERATIONS AND AI

creation in a business to identify where it makes sense for you to use AI.
Let’s start by looking at what it means to create value. In your business,
you create something that others will pay you for. It may be because you are
making a product or providing a service.
In a manufacturing business, the product is physical, a communication
agency makes money by making its customers better at creating compelling
communication. Their value to others is a service.
The way you create value is based on what you do for others that they will
pay you for.
The way your business works is no accident. There is a reason why you hire
the people you have, that your location is where it is, and your products are
not the result of a random process either. You have made plenty of conscious
choices on what it takes to create the value that your business generates.
Your business is designed around a model for how you create value. You
probably also want it to get better too.

Goals

The way you manage the direction of your company is probably centered on a
section of goals.
The aim of your goal is to make all parts of your company work in the
same direction, so you become better at what you do. Maybe your goals are
formulated as a strategy; maybe they are more loosely expressed. It may
be that you have one overall goal for your company, such as “best for our
customers”.
The most normal organizational scenario is that your business is divided into
several sub-sections. Such as sales, marketing, production, and finance. Each
department is usually managed by a set of individual goals. Accounting, sales,
production, and marketing are guided by goals relevant to their respective
domains. And the sum of your goals ensures that you become better at creating
the value that you aim for.
Setting goals is a core management tool that ensures that you, as a company,
succeed in achieving your corporate ambitions.

71
 AI FOR CEOS

Output

Setting goals is essential as a management tool. However, a goal only has

value if it gets implemented. So, you need to make the goals more operational
and specific.
Let’s say the sales department aims to increase revenue by 20% in the next
financial year, and that is their core goal. They will not be able to achieve
such a goal without making changes, or perhaps having a new focus. In other
words, they can only achieve their goal if they take some new actions.
In this example, they need to decide what will drive the 20% revenue growth.
It could be achieved in many ways: They could plan to launch new products to
existing customers, or by entering new markets, or maybe by increasing the
prices of their products.
So they need to plan for how each areas goals shall be achieved. The plans
are a balance between their capabilities, the resources they have available, and
their managers’ goals. They will consider the three areas and decide the best
way to achieve their goals. That’s their output.
An output example could be:

72
 OPERATIONS AND AI

We decide to start export to Sweden because by selling to the Swedish market

we can achieve our goal of revenue growth of up to 20%.
So the output is the actions that must be addressed to achieve your goals,
what you need to do to reach your goals. The goal is what to do. The output is
how you plan to reach it.

Create

Creation is what you do to generate your output. If you run a factory, then
this is the part where you produce your goods, if you run a bank, then this is
where you sell new pension products to your customers. Creations are what
you produce and how you create it in your business.
Creation also covers tasks such as sending tax returns to the government and
doing employee satisfaction surveys. And cleaning your parking lot. Creation
covers all that is required to run your business, not just the tasks that can be
directly related to a business goal.
The keyword here is tasks. The tasks that you need to do to run your business.
In order to solve your tasks, you have employees, you have processes, and
you have (IT) systems. Your employees perform jobs, such as Key Account
Manager. You also have processes that handle unpaid customer bills, for

73
 AI FOR CEOS

example. And you have systems that ship your products at your warehouse or
send an e-mail with offers to your customers.
The relationship between jobs, processes, and IT systems can be illustrated
as follows:

This is the way you should look to understand the concept. Some of your
jobs do not require the support of any IT system, for example, when your Key
Account Manager travels to London and meets with your new customers.
Other jobs require that you have systems to help you, for example, when
you need to receive a foreign bank transfer from your customer in London.
You also have tasks that are automatically solved by your IT systems. It could
be an email newsletter that is sent automatically to everyone who has signed
up for it or your accounting system that automatically processes most of your
invoices.
Finally, you have processes. Your support department consistently answers
customer questions as a process. They do so because it is the most cost-
effective way to help your customers with answers. To help them do this
even better, they make use of various IT systems.
Your tasks are solved by an interaction between jobs, processes, and IT
systems. Sometimes the three components work together; other tasks are
solved using just one of the options.
Now, we are getting closer to the AI part because a big part of what AI can

74
 OPERATIONS AND AI

do for you is to challenge the best way you can create value. AI is a new leg in
this equation, and it will disrupt the premises on how you solve tasks in your
business. AI does this by challenging the premises on which your tasks can be
solved. You will learn much more about how this is the case later in this book.
Before we get to that part, we need to look at how your business handles
capacity. The sum of what your employees can do, and your production
capacity are the total value you can create in your business.
So, getting back to the relationship between goals, output, and capacity.
When you decide your ability to reach your goals, then you would base your
reflections on your current ability to handle the tasks that you perform today.
In other words, your capacity.
If you are a communications agency, then you cannot handle 10,000 hours
of customer consulting annually with five employees. Such a volume requires
a change to your agency’s capacity. You would either need more employees
or maybe a different use of technology so that the advice you sell becomes
digital.
The relationship between creating, output, and goals can be illustrated as it
is shown in this diagram.

75
 AI FOR CEOS

Complexity

The last element in the value creation flow is the complexity of your tasks.
This is how you should look at it: All tasks that you have to deal with contain a
level of complexity.
Some tasks are simple. Send an invoice statement to your subscription
customers every month. Such tasks will be managed by an IT system for you.
Other slightly more demanding tasks may be solved through a process. For
example, accounting. 90% of your bills are properly paid automatically in
your accounting system. However, some bills are so large that they have to be
processed manually. So, this case requires you to master a slightly higher level
of complexity management, compared to the subscription customer task.
You will also have tasks that are too complex to be solved through a process.
Entering the Swedish market requires an understanding of how the market is,
perhaps even a network in Sweden, etc. Therefore, complex tasks are solved
as part of a job.
Having to deal with complexity gives you a natural constraint. You cannot
enter both the Swedish and Portuguese markets simultaneously because few
in your company are likely to have the knowledge, network, and insight that
this requires to do both such tasks. The span of complexity is too large.
It is such complexity consideration that determines that Tesla cars only are
available in 5 different colors.
You can’t get a Tesla in green. It’s not because it’s more expensive for
Tesla to make a car green compared to painting it red. The limitation of color
choices has to do with Tesla´s production complexity. By keeping the number
of colors down, it is easier to optimize car production, thereby increasing the
productivity of the factories.
By reducing complexity, Tesla increases productivity, and thus they come
closer to their goal of creating mass-produced electric cars at an affordable
price.
There is a direct correlation between your overall ability to handle complex-
ity, your ability to handle performance, your output requirements, and your
business goals.

76
 OPERATIONS AND AI

The context can be described as a sequence of context elements that

collectively describe the flows you work with when you create value in your
organization—the relationship between the four areas are shown in this
illustration.

This way of looking at value creation can be used to give you an impression of
what it takes to solve tasks and thus create value in your business.
You need such a view in order for you to get an overview of where AI can
help perform tasks for you.
Some AI projects can help you deal with complexity. Others can help you
handle the way you solve tasks better. Other AI projects can redefine what is
possible for you in business, thus changing your goals and the output you can
deliver in your business.
There is a total of five different areas where AI can change the way you create
value. Within each of these areas, you can find AI project candidates who can
create value for you in a way that is possibly better than what you do today.
You must understand what kind of value AI can create for you. When you
know the dynamics behind AI and understand how you can use it to create
value, then you are capable of identifying the areas where AI might be able
to work for you. That is precisely the goal of the next part of the book. To

77
 AI FOR CEOS

understand how AI creates value.

How does AI create value for you?

Knowing where AI can create value in your organization is your first step in
identifying if AI is an applicable technology for you to use. This part of the
book helps you reach this insight.
This means you should be able to see where in your value creation process
it makes the most sense to use AI after reading this chapter of the book. In
other words, you will soon be able to see how AI can create value for you in
your business.
For you to be able to do that, you need to understand the five areas that AI
has the potential to change. You have to look at it this way. Overall, AI can
create value in your business in two different ways:

• Efficiency: AI makes you better at solving the tasks that you as a company
are aiming at solving
• Market Improvements: AI enables you to understand your customers
and gain access to them in ways that give you radically new market
opportunities

One factor relates to how your business operates and the other factor changes
your commercial possibilities.
In management theory, these two parameters normally are not combined.
Usually, one will look at efficiency as an independent domain, and look at
commercial areas as something very different. There are two very different
areas, without significant overlap.
However, AI affects them both by changing their premises at the same time.
So, your perspective has to weigh in both factors if you want to have a complete
picture of how AI can affect your business. The common denominator is
new opportunities to create value in a way that covers internal and external
relationships.
AI and efficiency means that AI could challenge your IT systems, your

78
 OPERATIONS AND AI

processes, and the jobs your employees perform. AI will be able to solve
many of the tasks currently handled by those three factors better and more
efficiently.
The curtain van is a tangible example of how efficiency can be increased
with AI.
In this example, the processes were changed after AI was able to conduct
conversations with new customers. Some job functions were also changed as
an employee no longer had to handle the bookings. And the way meetings are
booked meant that their IT system (calendar-booking) was used in a new way.
So effectively is a lot about how many similar examples will make sense to
work with in your business.
The more of this type of case you will find, the higher the potential AI will
have as a possible driver of change. You will be presented with a method for
that particular task very soon.
The second parameter relates to your products, your customer insight,
and the way you access your markets. The sum of which I call Market
Improvements.
It’s called Market Improvements because AI gives you new opportunities to
sell more by understanding how to adapt to your customers’ needs more
dynamically. So, using AI can facilitate changes in products and market
conditions for you.
If we start by looking at the efficiency part of the value creation process,
then you can achieve three effects. You can:

• Become more productive

• Increase your ability to handle complexity
• Increase your level of service

We now go through each of them so you understand how they work and so that
you can judge if one or more of them could be potential AI project candidates
in your business.

79
 AI FOR CEOS

You can become more produc ve

AI can be used to solve some types of tasks more efficiently compared to

how they are fixed today. The goal of using AI in this context is thus a more
productive operation of your organization.
AI can learn to understand how to solve a specific task, after which it will be
able to solve it instead of the person who would otherwise have done it. So, a
bit similar to Robotics, but at a more sophisticated level.
The short explanation of Robotics for those of you who do not know it, is
that it is a way of improving your processes by making two or more IT systems
handle a sequence of tasks automatically. You do this by installing a “robot”
that makes the underlying IT systems perform a specific task.
Such a task could be if a new customer signs up, then you might need to key
in his data in multiple systems in your business (CRM, billing, etc.). Instead of
having a person to do manually in your systems, you could have your Robotics
solution do it for you in all the systems you would like the customer to appear
in.
Robotics solutions work by following a defined set of rules for how to work
with your IT system.
AI can also manage processes that require working with multiple IT systems
but at a higher level of sophistication because AI can be taught how to solve
much more complex tasks, including jobs that cannot be described by rules.
My newest Start-up, Cognifirm have designed software that can do this for a
business.
Efficiency and productivity is core parts of how AI can create value, and it
will be in focus throughout the rest of the book. For now, you should just know
that when you read that up to 40% of all office jobs could be taken over by AI;
it is very likely this type of AI application that the authors of the articles you
have read have in mind.
The design of AI productivity solutions requires the use of both Assistant,
Automation, and Bots. How you work with it will be described in an upcoming
section.

80
 OPERATIONS AND AI

Increase your ability to handle complexity

I admit that being able to handle complexity could seem like a strange need to
have. The way you should look at it is this.
The human intellect can manage mathematical complexity to a certain level.
Our limitation is rooted in the capacity of our brains. We also have a limit in
our ability to define solutions to the problems that we encounter. So, we don’t
see everything. Especially if the amount of data we work with is very complex.
But AI sees data (and thus the world) differently, and this creates two new
possibilities:
• Better analysis and insight
• The opportunity to create new products and services
Analysis tasks primarily mean that AI can find meanings and correlations
in data sets that we probably would not have discovered. So you will be able to
make decisions on a more informed starting point.
AI can also come up with product designs based on your requirements. It can
open up new opportunities to make the products different, or design product
with features we might never have imagined.

Increase your level of service

The service section has to do with how you interact with your organization’s
employees, customers, and other such vital stakeholders.
AI makes it possible to go from (digital) communication to (digital) dialogue.
The fact that your communication is digital means that it scales to many
virtually free through online services such as the Internet, emails, and mobile
apps.
But communication is one-way and goes from you to the recipients. You
tell the same message to everyone (I know there are nuances, but basically
that’s true).
With AI you can now have a digital dialogue with your stakeholders. So,
you can talk, advise, influence, and interact with them. With the same cost
structure as with digital communication, because the dialogue is handled by

81
 AI FOR CEOS

bots.
We go from electronic communication to automatic individual dialogue. I
will return later in this section with more on how the opportunities to get
closer to your stakeholders could work for you.
If you look at what AI can do to make you more efficient, then there are great
but also very different potentials. The common denominator is efficiency,
but there is a big difference between replacing processes and jobs with AI
and using AI for analytics. Therefore, the three areas will be evaluated as
independent sections in this book.

Market Improvements

The second of the two areas where AI could help improve your business is called
“Market Improvements”. Here, AI helps you sell more to your customers; it
can assist you in creating better products and by supporting you to get closer
to your customers.
There are two elements in the Market Improvement part that you can
improve with your AI. These are:

• Product development - You will be able to create new prototypes quickly

and be challenged on how it is possible to design your products
• Market agility - You will be able to search, cultivate and exploit new
markets in a structured and automatically way

This part of the value creation aims at creating more qualified sales process,
creating new marketing opportunities as well as supporting new possibilities
to develop your products.

Product development

AI can be used for a new form of product development. The way you should
look at it is this: Any professional like engineers, developers, or designers are
trained to solve problems based on a given methodology.

82
 OPERATIONS AND AI

If an engineer is asked to design a new bridge over a river, then she will
come up with a proposal for a design that is based on her ability to understand
the problem, her educational training, experience, and knowledge of bridge-
building.
The engineer makes a great proposal for a new bridge, based on her way
of seeing the world. The same can be said about the designer who makes
new clothes, or the developer who makes new software. It is our professional
background and our experience that makes us competent to solve problems.
But it is also our professional background and our experience that prevent us
from seeing solutions to problems in new ways. Who says the bridge couldn’t
be built by using recycled plastic? That - however - is not a real option to us
because our bridge-building engineer only knows how to build bridges out of
concrete and iron. This is where AI comes into the picture.
You can use AI to come up with new ways to solve your design problems,
which is created without being hindered by the way you have learned to see
the world. But, still in a way actually solves the problem. AI will understand
your problem differently, so it will deliver different answers as solutions to
your problems.
Maybe AI’s answer is better than yours. You only know that when you try.
But AIs answers are in any case different and will challenge your current view
of the world. How you do so is a topic in a later part of the book, including the
strategic perspectives you can gain by using the technology is this way.

Get closer to your customers

Another way AI can help you commercially is by helping you to get closer to
your customers and your potential new customers.
AI can help you ensure that you offer the right products, to the right
customers, in the right markets. From a commercial perspective, this is not a
new goal. Since the dawn of trade 5000 years ago, this has been the aim for
anyone involved in commerce.
But, the way AI can do it for you is new. The way to look at it is to view trade
as a process that you aim to optimize. The components that are in play are the

83
 AI FOR CEOS

elements that make up your product, your market, and your customer. You
are aiming at reaching the joint center of the circles, as illustrated below.

Again, we draw on AI’s ability to see patterns. Initially, you can use AI
to understand what it is that defines your customers in the center of the
illustration.
You probably only sell to 5% of your potential customer base today. AI can
help you see the patterns that set your customers apart from those you don’t
yet sell to. And you can get qualified suggestions on whether it is your products,
your market or the customers you sell to, that need to be adapted or changed
if you want to sell more.
The way AI does it for you is through a series of sequential learning loops.
Exactly how to do that is described in more detail in the book’s strategy section.
If you compare the five areas where AI can create value with the way you
create value in your business, you get this diagram:

84
 OPERATIONS AND AI

This means that AI can challenge you in three different ways:

• In what your business goals might look like

• AI can make that way you manage your tasks more productive
• AI can redefine the amount of complexity you can handle

The next section of the book covers how you find out in more detail if any of
the five areas can create value for you and your business.

Iden fica on of your AI project candidates

Now you have learned how you create value in your organization, and you
have been introduced to the five areas where AI can create value. You now also
know how to merge the two perspectives.
The time has now come to get closer to your business. We now shift focus
and digs deeper into the individual areas. The aim is now to make it possible for
you to evaluate if one or more of the five areas can be used for actual projects
in your business.

85
 AI FOR CEOS

To help you decide if AI makes sense for you now, means that we go into
more detail with each of the areas. You will now learn how each of them creates
value and how you can activate them in a much more operational perspective.
We start with product development.

Product development

The technique used when working with product development and AI is called
“Generative design”30 . It’s a way that potentially can reverse the design
process, and creates new techniques for you to develop your products.
Let’s say you are a furniture company that needs to design a new chair.
Usually, you will start the process with a designer or design team that
brainstorms new ideas.
You will then have a process where the best ideas are chosen. The ideas are
then verified in a computer design program. You build different 3D models
of the designer’s sketches in the software to get a sense of what the product
would look like. The next step is then to make physical prototypes of the most
promising designs.
With AI, you can get new aspects of the design process first phase, namely
where you brainstorm the design. This is possible because you can create
AI-based design suggestions for the chair that you are designing.
The reason why AI can design products for you in this way is because you
would have defined the conditions under which your chair must be made.
Such conditions could be physical features, like height or choice of material.
For example, the seat must be 50 cm over the floor, the chair must be made of
oak, and it must be foldable for easy storage.
You could also have requests that define the visual expression of the chair.
Like, for example, that it should have an organic appearance with no sharp

30
“Generative Design (Form + Technique)” by Asterios Agkathidis shows 11 different projects
that have been made with Generative Design software. The book is a great place to start if you
want to understand the topic in more detail.
https://adsknews.autodesk.com/stories/future-things-get-made-harnessing-power-
artificial-intelligence-next-frontier-design

86
 OPERATIONS AND AI

corners, or that it must appear visually thin, and therefore must not be more
than 2 cm thick in any place.
Based on your requirements for the chair’s features, your AI can then design
solutions that meet your needs.
Once you have defined your requirements for the chair and described the
prerequisites, then you would be able to get as many design suggestions as
you wish. You can make 50 or 100 different designs of your chair; it is entirely
your choice how many you need.
Thus, AI helps you create new products by:

• Potentially shortening designs the process,

• Increasing the number of solutions you can work with
• Challenging you on your design concept, and how the perception of your
product’s appearance could be.

This link contains an example of products designed by AI:31

https://youtu.be/CtYRfMzmWFU

The value of Genera ve design

The chair example explains the potential of the technology. In addition to

getting alternatives to your design process, you can also use the technology to
create products that would otherwise not be technically possible to make. Or
to develop new products with less effort than if you don’t use AI.
You can train AI to come up with design answers to problems like:
• Make a design proposal for a new bridge
• Make a design proposal for a party dress
• Make a design proposal for a new piece of software
The design may be different; you may be able to use new materials; you can
quickly design prototypes that can be tested by real customers. In other words.

31
https://adsknews.autodesk.com/stories/future-things-get-made-harnessing-power-
artificial-intelligence-next-frontier-design

87
 AI FOR CEOS

There are a lot of elements in the creative part of producing products that will
have other limitations than they have today.
AI has no professional background or experience. But it can be trained
to know how components of a domain behave. This can be used in product
development to get new suggestions for how we, as an example, build a new
bridge in a different way that an engineer might have suggested.
Whatever kind of task you want to help you design, the approach is the same.

1. You define your requirements. For example, the design of a new folding
chair
2. You train your AI to understand the materials that the product should be
made of. For example, the physical characteristics of plastics.
3. Define the terms the case must be produced under. For example, the
chair must be made of oak; the seat must be 50 cm over the ground. The
shape of the seat must be within some limits.
4. Ask your AI to come up with design suggestions for your oak chair, bridge,
dress or new software.

Creating such product designs requires a whole new way of working where
you work with your AI. It means that your AI does not work for you. It works
with you. So the AI is much more like a virtual co-worker than software.
The consequences of this kind of working environment will be a topic that
you get more specific insight into in the leadership section of the book. So, for
now, you should not have as much focus on that part.

Does Product Development have business poten al for you?

There are three primary situations where AI in product development has an

advantage.

• If it is expensive for you to do product development. The cost structure

is different for AI product development than for traditional product
development. With AI, you “pay” pretty much the full price by getting

88
 OPERATIONS AND AI

your AI ready to compose new product designs. After that, new products
can be designed almost for free. Thus, a high initial cost and low operating
costs.
• If you develop many new products. If you make many products and product
development costs are a big part of your total expenses, then working with
AI could make sense to you.
• If product innovation can give you a competitive advantage. AI has the
potential to come up with groundbreaking new designs. Such designs
could provide you with production or competitive advantages.

You must weigh in the benefits against the price you pay and the risk you
take. It will probably take a long time to get your AI ready to make qualified
suggestions for new products, and there is a real risk that you will not reach a
viable result.
The barrier to getting started with this type of AI is probably relatively high
for you. In order for you to use AI to design products, it requires that your AI
understand the physical properties of your products, quite accurately.
For example, your AI needs to know the limits to which a given material
can be stressed, to what extent it is physically possible to reshape it, what
physical attributes it has, etc. So it can be a big job to make your AI ready to
make qualified prototypes for you.
When you start building AI, there is no guarantee of the quality of the output
that you end up getting. You do not know in advance whether the product
proposals that your AI comes up with are revolutionary or whether they are
useless. So, you run the risk of wasting your time and energy.

Remember that I told you that you can download free tools on how to evaluate
your business and AI. You can find a template that helps you evaluate if product
design could be an area that you should look into. The URL where you can
download it is cognifirm.com/tools.

89
 AI FOR CEOS

Service

We start our explanation of how you can use AI to create better service starts
with an example.
Let’s say that you are considering switching your mobile subscription plan
because you have the feeling that you are probably paying too much for your
current one. You also experience that it does not cover international roaming
as well as you would like.
One way you could start a search process for a new subscription plan could
be by checking online price portals to see which operator appears to be the
cheapest.
Then you could click into the cheapest telecommunications company´s
websites to check prices and the product features. You travel a lot in the US, so
international roaming is important to you.
But you weren’t quite sure if there were any special conditions for the US
roaming in the product deal you saw on their website. Another critical element
need for you is how well the data coverage works in a particular part of your
country — Namely where your holiday cottage is located (in a forest).
If you can be sure of the network quality is ok there, then you can stream
Netflix over the phone and thus avoid having to pay to install fiber network at
your cottage.
There was some information about the geographical coverage on telecom-
munications company’s websites, but it required a few clicks to find the
information you were looking for. In particular, the issue of coverage in the
cottage area was complicated for you to obtain information about.
Nevertheless, you end up ordering the subscription plan you thought was
best for you. You received a letter in the mail with your new SIM card. When
your old subscription expired, you made a change to your new mobile carrier.
Looking back at the process, it was reasonably straightforward. So how can
AI make such a customer experience better?
AI can do this by letting interaction with bots handle the dialogue with you.
So, you don’t have to find the information you need on a website. You get the
information you need through talks and conversations with AI bots.

90
 OPERATIONS AND AI

You could do that by calling the telecommunications company. There was

no queue. The bots were ready right away.
You explained to the Bots what you were in the market for, a cheaper
subscription that supported relatively much data usage in the United States.
And that you need certainty about the coverage precisely where your cottage
is located.
The conversation you had on the phone was with the telecom company’s
bot. It suggested the subscription deal that exactly matched your needs. It
was a nice-product with extra US roaming data that you did not see on their
website. So, it was cheaper than what you otherwise would have chosen.
You also asked the bot about when your new SIM card will arrive, and what
to do in the transition period between the transfer of two subscriptions plans.
You got qualified answers to such questions too.
When the conversation was over, you received an email from your new
telecommunications provider. The mail contained personalized and precise
instructions on what would going to happen, including when you had to take
actions in the process.
The instructions also included a precise indication of the coverage at the
address of your cottage. The coverage information even contained benchmarks
against the other operators’ networks, so you were assured that you had
chosen the right solution with the best coverage.
The process could best be compared to the level of service you would have
received in a 1950s department store. A store where a competent assistant
would have approached you. An assistant who knows you as a customer, and
the products you could have an interest in.
What you experienced, in this case, was AI-based service.

The value of AI-based service

As you have just seen from the example, service has to with the way your
organization’s employees, customers, and stakeholders are treated. AI opens
up for a more personal and direct service based on the needs of the person you
aim to serve.

91
 AI FOR CEOS

That is, where we go from automatic communication to automatic individual

dialogue. The goal now is to meet the needs of the people we aim to serve. We
do this by:

• providing the right insight to them

• in areas that are relevant to them
• in the situation, they are in right now.
• presented in the way that they understand the best

AI-driven service has two braces, an internal that is designed to serve the
employees’ need for knowledge, training and education, and an external one
that is aimed at the external stakeholders’ needs for expertise from us.
It is a paradigm shift to go from saying the same thing to everyone (like
we do on the Internet) to having a relevant personal dialogue based on the
specific needs of each user.
Communication is a one-way street and targets many people with the same
message (perhaps divided into segments). Dialogue is individual and based
on an understanding of an individual’s specific needs. This is equivalent to
ether selling spectacles by advertising on a billboard at City Hall Square, or
through the advice of a qualified optician.
The key to AI is that, like Internet communication, AI is free once you have
developed it (roughly speaking). Because AI bots, like the Internet, are digital,
and thus (basically) it doesn’t cost too much to go from a few to many users
of a solution.
It goes without saying that companies that can exploit the opportunities
for close customer contact as one-on-on digital dialogue will be able to offer
excellent customer benefits in many areas.
They can bring customers closer to their brand; they can provide a service in
every part of the operation where there are customer interactions, at quality
levels that are not possible today.
The service part of AI is particularly interesting in the industries where the
products are relatively similar, such as internet providers, telephony, and not
least, e-commerce.

92
 OPERATIONS AND AI

It would also be reasonable to assume that AI-based service has the potential
to influence companies’ brand, sales potential, and quality of support. Thus,
it will be able to move sales from competitors that do not match such AI
initiatives.

How does it work?

The way customers will be serviced will be through bots. It could be through
dialogue as in the phone case, or it could be a chatbot on a website.
Bots require a completely different kind of system design and thinking than
what is known and done today. A bot is a much more intuitive user interface
than, for example, a website.
A Bot can be used to give users access to information that would typically
have been displayed on a website today. But in a way that is in dialogue form,
thus supporting the individual user’s unique need for knowledge.
Bots have three key elements: First of all, it must understand you as a user.
So, there is functionality to make sure you can talk to it. That is, it must be
able to transcode language into meaning and intent. The bot must then return
a qualified response to the user. It must be substantiated by business logic
that finds the answers the user seeks. The last thing it needs to be able to do
is to speak to you so that the bots responses will be presented in a form you
understand.

Is AI-driven Service business poten al for you?

There are a lot of Use Cases where bots have their ability to provide service to
you. Bots provide value in two different dimensions:

• Effectiveness: Jobs and processes can be taken over by Bots. They provide
a more efficient organization.
• Customer Experiences: Bots have the potential to enhance the perceived
quality of users’ relationship with your business.

93
 AI FOR CEOS

What you want to use bots for, in this case, is to take over tasks that require
lots of resources today. That would normally be various kinds of services to
your customers or your employees. What you are looking to replace with AI is
areas where you have many resource-demanding processes.
However, a small warning. Not all processes are equally well suited to being
taken over by a bot. In the next chapter, you will find a method to validate
the real potential of an AI project, and there you will also be able to see if the
processes you identify here actually is a strong potential for being taken over
by AI or not.
Personal automated dialogue through bots has excellent potential to bring
your customers closer to your business. But it also places high demands on
your technology.
The bot used in the phone example is probably a 2nd or 3rd generation bot,
which was developed on top of a simpler bot. So, you have to look at this as
an evolvement process. You will grow into being able to provide even better
customer service when the maturity level of your customer bots is continually
evolving.
As with the product prototypes, you must relate to the relevance of the
area and find the most critical projects. Are there service areas where you
can find candidates for the use of a bot in your business today? Remember,
cognifirm.com/tools can be used to do this online.

Produc vity

AI can also be used as a way of solving your tasks, making you more productive.
The reason for this is that some of the functions you are currently addressing
with either your IT systems, processes, or jobs could be solved more efficiently
with AI.
Let’s continue with the telecom example from before, but now we do not
look at the case from the customer’s side, but from a management aspect
where the focus is cost and efficiency improvements.

94
 OPERATIONS AND AI

A short recap of the case: Customers could now talk to a bot which helps
answer products and services enquires as a substitute for either finding the
information on their website or talking to a customer service employee.
Some customers call with questions about their bills, others need help
setting up their new phone, others are in a relocating process, and have
questions about how their new situation will affect their phone services.
It will always be a Call Center that helps customers with such questions.
There is a fairly high degree of predictability in the types of questions your
customers need to get answered. That’s why you have created processes that
help you deal with them.
The processes ensure consistent and correct responses to your customers.
So, when a customer calls in with questions on what to do in a relocation
situation, then a customer would get the same answer no matter which
employee the customer would have been talking to.
You ensure this by letting the call center employee practices the same
technique to clarify the customer’s needs, and by using a template for which
answers to present in a given situation.
Processes also make you more productive because you can organize your
business in a way that ensures that you can solve the customer tasks as
efficiently as possible. That’s why you use IT to support as many functions in
your workflows as you can.
One of the core reasons why you have a heavy focus on IT is because the
major cost of running a Call Center is salaries to your employees.
From a management cost perspective, the more tasks that can be solved
without employee involvement, or with as little employee involvement as
possible, the better.
The balance between cost and customer service are the two parameters that
every Call Center must always find the right level for.
This is where AI comes into the picture. AI can solve a number of the tasks
in a Call Center that employees today perform. Tasks that an IT system can’t
take over.
Let’s get back to the onboarding case. The chatbot guided the customer in a
competent way to get answers to the questions he had.

95
 AI FOR CEOS

The outcome of the case showed that the service level was perceived as high
by the customer and the cost level was low because there were no employees
involved in the process.
Thus, this kind of customer service task falls into the category of jobs that AI
will be good at solving. One of the reasons is because this task has a predictable
outcome. You know what the right answer is to give the customers in a specific
situation. In such cases, AI can be designed to solve the task for you.
So, you can use AI in your customer service to design a set-up that can
answer customer’s questions that your customer service employees manage
today.

The value of AI-based service

The reason why AI can drive your efficiency forward is because it reduces your
capacity constraints.
The tasks you solve in your business today can roughly be divided into two
categories, processes and jobs. Processes are the ones we just went through.
These are frequently occurring tasks that would be solved the same way
regardless of which employee solves them. You have organized them as a
process. Your Call Center tasks are probably almost exclusively solved by
process tasks because it is the way to ensure efficiency.
But you can’t do that with all types of tasks. Your export salesmen’s job
cannot be defined as a process. It makes no sense to describe how to fly to
Milan and negotiate with a possible new customer as a process. So that’s why
you have jobs that contain “not process” tasks.
You also have IT systems that help you become effective. Some are designed
to be process supportive. That is the way systems, such as Salesforce and SAP
work for you. Other IT systems help individuals do their jobs. They use Word,
email, and Excel which are all designed so that a person can solve specific
tasks.
Common to both jobs and processes is that there is a capacity limitation.
Because they both require people to perform tasks. In the process, the
customer service employee must talk to the customer, and in the case of sales

96
 OPERATIONS AND AI

your employee must fly to Milan. So, the number of employees sets the limits
for how many tasks you can solve.
However, tasks solved with AI have no capacity limitations. When customers
who previously called in with questions about subscription, it is always
possible to serve them. Whether you have to deal with 10 or 10,000 at once
makes no difference.
This gives you the opportunity for the potential scalability of your services,
because you no longer have the capacity as a factor to take into account when
solving the task.
It can have potentially great strategic implications for you to start thinking
that way, and this topic will, therefore, be addressed on an ongoing basis in
the book’s strategic section. There you get specific tools to identify where you
can work with AI so it boosts your productivity.

How does it work?

The way a productivity system will work is by learning a user’s behavior. It

could be a customer service employee handling a support case.
Let’s say it is an event that happens quite often and takes time for the
employees because they have to work in multiple systems in order to solve the
case. Each task in itself is not difficult; it just takes time to do them all.
But with AI, you could train a Knowledge System to do as your support staff
does when using the systems to solve their tasks. The Knowledge System
would receive data on how your employees use your systems, and based on
that, it will look for patterns that determine the choices your employees make
when they interact with the various system they use.
The system is now learning. At some point it will now be able to suggest
how the tasks that your employees solve can be done. The system will suggest
what to do when creating a specific new support case.
You will then tell systems whether its suggestions are right or wrong. Based
on your feedback, the system will build a knowledge of the patterns that
separate the right outcome over the rest. The system thus becomes “smarter”.
At one point in this process, the system has become so great that it can solve

97
 AI FOR CEOS

the task in exactly the same way your employee would have done it.
This concept of training your AI is a core element when you develop new AI
solutions, and there is a chapter in the Leadership section of the book called
“The art of training your AI” that is dedicated to explaining how this is done.

Does AI bases process development have business poten al for

You?

On a high level, you should see productivity improvements in the same way
you look at your Digital Transformation processes.
The goal of a Digital Transformation process is to identify areas in your
organization where tasks can be digitized. The consequence of digitization is
new processes, and perhaps new ways of organizing the way you work. The
change driver is making manual tasks digital, and the result is a more efficient
organization.
AI-driven productivity improvements also make your organization more
efficient. But it does so by making manual decisions digital.
So, the tools and methods that you use when looking for manual process
candidates in your Digital Transformation process can also be used when
looking for productivity improvements with AI. Now you just don’t look for
manual processes; you look for manual decision making.
I know it sounds strange to call a decision manual and digital, respectively.
But that is a consequence of the premise that your business is now getting a
digital intelligence. You enable your company’s digital intelligence by letting
it make decisions for you.
You let your AI decide whether customers should have a newsletter, whether
they should have credit, or whether they should have a discount for keeping
them longer as customers. These are all examples of decisions that your AI
will take on behalf of your company.
It is a real and relevant consideration to what extent you want your AI to
make decisions on behalf of your business. Should it be allowed to terminate
a deal with one of your suppliers? Would you let it give discounts to some of
your customers, or would you let AI decide if a customer is applicable for your

98
 OPERATIONS AND AI

products and services? A core element of the strategy section of the book is
dedicated to helping you get a grasp on such considerations.
In practice, there are two dimensions of AI productivity improvements.
Either you can become more efficient or you can increase your capacity. The
efficiency part is primarily about replacing jobs, processes, or IT with AI.
Capacity enhancement aims to create more with your current resources.

Complexity

The way you look should look at AI and complexity is that AI enables you to
solve more difficult tasks than you can do today, and it may also help you to
automate more functions than you cannot do today. So, a better handling of
complexity in your business.
On a more practical level, this means that you might be able to redefine
what you can do as a business for your customers and also what products and
services you can produce. Here is why.
To help you understand what managing complexity means, we continue
with our telecommunications case. But now we are focusing on another big
field, namely what is known as Churn. Churn is when a customer leaves a
subscription product such as a mobile subscription.
The business model of the telecommunications industry is that they invest
in very expensive infrastructure. They must then recruit as many customers
as possible, to get a return on their investments.
This is why one of the key drivers of this industry is marketing. They invest
heavily in marketing to acquire new customers. When they get a new customer,
they must ensure that the customer stays with them for as many months as
possible. To ensure that they get their marketing costs covered and to ensure
that the customer also contributes to cover a fraction of their infrastructure
costs.
These business dynamics means that the telecommunications industry is
very interested in understanding what drives their Churn. So, they can keep
their customers for as long as possible.
They are interested in understanding what makes customers leave their

99
 AI FOR CEOS

business. They also want to be able to do something proactive before

customers leave them. So, they can contact customers who are likely to exit
before it happens.
Their big challenge is that once a customer cancels a subscription, they have
already subscribed to a competitor´s product. They cannot be recovered at
this point. They only win if they can retain those who are on their way to leave
before they take action.
One of the ways telecom companies are addressing their Churn problem is
by making computer models that can explain why customers leave them by
looking at customer data.
If you can use data to describe the difference between the customers who
stay and those who leave, then you can come closer to an understanding of
why they go. And you can identify those of your customers who are at risk,
and thus you can act on your knowledge. Such actions could be; contacting
customers, giving them price discounts, more services, or whatever you can
see you need to do to keep customers who are at risk of leaving.
You may remember earlier that one of the reasons why our customer left
his previous telecommunications company was that there was poor coverage
in his cottage. In his case, it was a contributing factor for his exit.
AI could have helped the telecommunications company to focus on such
areas.
If it turns out that there is a segment of the customers who are leaving
them who have a common pattern in their behavior. Such a pattern could be
that they own a cottage in an area where the telecommunications company’s
coverage is inadequate. In such a case, we would have a factor that may or
may not explain part of their Churn.
What AI does for the telecommunications company is that it finds the factors
that best describe why customers leave them. It may well be that customers
who experience good coverage in their cottage area also leave them. So, the
cottage coverage does not explain why customers churn because there is no
difference between the good customers who stay and those who leave them.
This is where complexity management comes into play.
Because there can be hundreds of different reasons why customers leave

100
 OPERATIONS AND AI

a business. We can have theories and hypotheses as to why they do. But AI
helps us identify the areas that can best explain a given behavior. This is called
predictive analytics and is an area that we will touch on more regularly in the
future sections of this book.
Many companies first AI experiences start with analytics and this branch
of the technology. This is because there is tangible business value in the
knowledge that you gain. And because the data infrastructure required by the
analytics tools is often relatively accessible in most companies.

The value of AI and complexity

Handling Churn represents one part of being able to handle complexity,

namely in the form of advanced analytics. The second part allows you to
create new products and services.
The overall reason why AI is good at helping you with complexity is that AI
is good at seeing patterns in complex circumstances. The human intellect can
handle mathematical complexity to a certain level. For example, we can see
a correlation between data in a spreadsheet. If we want to go further in our
explorations, then we use advanced formulas based on mathematics.
However, both approaches have a limitation in what kind of patterns we will
end up identifying. The constraint is rooted in both our is our intellect and
our imagination in defining the patterns we are looking for. So we don’t see
everything.
There will be patterns in the areas that we are looking into that we cannot
bring ourselves in a situation where we can detect them. This case is especially
true if the amount of data is very complex.
AI can be designed to help us find the patterns we don’t discover ourselves.
That’s why AI is good as an analytics tool, as was just proved in the telco case.
AI can also be used to create new insights that may turn into new products. So,
another way of using AI to handle complexity.
AI is the foundation of the part of the analytics disciplines called predictive
analytics. We already covered this a couple of times now, so we will now focus
on the other part of managing complexity how AI can manage complexity and

101
 AI FOR CEOS

create new products to you.

An example from Stanford University will help us explain how.
At Stanford, an AI model had been trained to identify if dangerous birth-
marks could develop into skin cancer32 .
The model’s results were tested against the best dermatologists’ diagnoses.
Not only did the AI model have greater accuracy in its assessment of whether
the birthmark was a skin cancer precursor or not.
The AI model also ended up having 12 different criteria that it used in the skin
cancer assessments. The dermatologists use 9. The machine has thus found
three new factors which were an indication of skin cancer, which importance
the best dermatologists were not aware of.
The models they used at Stanford were not “hindered” by our human
professionalism or our scientific view of how a given subject - in this case,
dermatologists - perceives reality.
The AI model approach was objective and rational in their understanding of
the given subject (Skin cancer) that they were trained to understand.
The models related to the patterns they saw in the data that were available
to them and, based on this, form their perceptions of how skin cancer versus
non-skin cancer could be described. That is why they came up with 12 areas
that could define skin cancer and not the nine that we, as human beings, have
chosen to use.
If Stanford were a traditional commercial enterprise, then they would be
able to exploit this handling of complexity to gain a tremendous competitive
advantage.
They would be able to make a better product and perhaps faster than
competitors to find the potential cancer patients.

32
https://cs.stanford.edu/people/esteva/nature/

102
 OPERATIONS AND AI

There is a third complexity leg which is probably where you can create the
most value with AI when it comes to managing complexity. It has to do with
your business and complexity.
If you remember the value creation flows, going from goals, output, creating
complexity, all the tasks you must be able to handle in such a flow have a certain
level of complexity build in.
It is complex to make an annual report, but it is a necessary task that you
have to make every year.
However, in some cases, it will be possible for you to select tasks that have a
different complexity profile than the job you handle today has. Or rather, you
can choose to handle a type of complexity that can be managed well by AI.
Let’s say your company is the telecommunications company from our
previous case.
In this case, the customer called the chatbot and got advice from it. What
also happened was that you handled a new level of complexity, that you did
not offer before. It was the part about products and prices that the Bot gave
the customer.
By using the Bot, you were able to tell the customer about specific coverage
where his cottage was located and to send him a personally designed feedback
on his inquiry. Thus, you also chose to deal with a different kind of complexity

103
 AI FOR CEOS

than you do today with your customers.

Such elements can potentially have a significant impact on the way your
business can be run. They can do so because you are eliminating your current
capacity constraints (call center employees). On top of that, you were now
able to solve tasks automatically that you choose not to do today because they
are too expensive for you to handle.
Those tasks are the personal letter and the information on local roaming.
You currently have all the data you need to send it to your customers, but you
choose not to because it will be too time-consuming for you to prepare for
each customer.
This kind of AI use, however, makes it possible for you to provide better
service, cheaper. That is why the area can potentially create significant value
for you. How you do so, that is explained in the Strategy section of this book.

Does complexity management have business poten al for you?

Analy cs

If you make critical decisions based on complex amounts of information, then

the predictive analysis is an area you should be interested in. Especially if the
action you take is important to run your business. Typically, those activities
have to do with customer relationships in finance and production management.
What is common to all areas is that making the wrong decisions could have
significant consequences for your business.

Products or services with a specific goal based on data

Products or services that are based on data have the potential to be improved
with AI. The quality can be increased, as we saw in the skin cancer example
just as it may be possible to add better service as the telco case also showed.

104
 OPERATIONS AND AI

Market Agility

The goal of market agility is that you sell the right products, to the right
customers, in the right markets, in a replicable way. When something is
replicable, then, it can be reproduced, and if it can be reproduced, then it can
be scalable.
So, in this case, AI helps you to understand why you have sales success and
what you need to do to be able to repeat your success in a new context, over
and over again.
This type of market approach is especially relevant if you sell digital products
online. The best cases are the ones where you also create your own products.
Because then you have control over the product attributes and features too. So,
it is applicable for industries such as telco, finance, and general e-commerce.
The underlying dynamics of AI in this context is a variation of what has just
been discussed in the complexity section. Namely, that you can describe the
difference between success and no-success on patterns in the data you are
looking at. And you act on that knowledge by doing more of what you do well.
So, AI here helps you stay focused on what actually drives your sales.
The first step is to understand what factors can explain your sales the best
in the same way that telecom companies tried to understand their Chrun.
This time, your focus is on the elements that drive your product, customer
and market fit. First, you aim to understand what the differences are between
the customers you sell to, and those you don´t. So, you have information that
can explain why you sell.
The thesis behind the approach is the explanation of sales success not only
lies in the characteristics of your product. Sales cannot be explained solely by
the behavior of your customers or by the markets in which you operate.
The explanation of why you sell is a sum of factors. There is always a split
of their importance. A pattern that best explains why you are successful. This
pattern can be found in elements of both product, market, and customer
relationships.

105
 AI FOR CEOS

So, this time we are also trying to understand patterns in complexity and to
discover insights that we otherwise would never know of. That’s why you need
to look at both products, customers, and markets at the same time. They are
linked together when trying to explain what drives your sales.
When you understand what factor combinations that truly drive your sales,
then you can copy them and repeat them under new circumstances. If you
can achieve this, then you would have created an understanding that can fuel
exponentially sales growth. So, this is important, and here is how it is done.
If the majority of sales are driven because you reach a unique type of people,
then you need to focus on this element. If particular product features drive
your sales, then you must include such features into other products. Or if it is
your way of approaching the market that explains your success, then it is the
element that needs to be copied and repeated. It is normal, however, that a
combination of sub-elements from all three areas best explains your sales.
Your goal is to find a Micro market that you do really well in based on
analytics of the three elements. What you are looking for is a good product-
customer-market mix. When you succeeded in identifying such a product-
customer-market mix, then you copy it. You aim to copy what defined the
success into a new sister market, and then copy it once more to another market,
and so on.
This process creates several small Micro markets, where you have products

106
 OPERATIONS AND AI

that are exactly adapted to the needs of this (small) market.33

The approach can thus be seen as a learning process where you learn what
works and then do more of it. Again, the dynamics that ensure this can take
place are based on pattern recognition.
For example, it could be a wish for you to sell a new indoor sports shoe. The
market could be school children, basketball players, or fans of e-sports. There
are many parameters on which a market can be defined. Similarly, a product
can have many characteristics, price, design, or materials. And finally, some
characteristic defines your customers. Their needs, specific needs for sports
shoes, their need for social status, etc.
The idea of a Micro market is to find strong and significant patterns among
all these parameters together. It could be that the market could primarily be
defined as girls who play basketball with a preference for pink and who watch
Keep up with the Kardashians on TV.
AI helps us find distinctive patterns that can explain the reasons that drive
our sales. And AI can help us identify what is the best way to go for the next
Micro market. Should it be based on customers’ interest in The Kardashians,
pink and TikTok, or on basketball, New York, and status? There are too many
parameters in such equations for any human to know, but AI would be able to
find them for you.

Does market agility have business poten al for you?

The Market Agility approach is of most interest for companies with significant
online sales, or considerable online sales potential. It is of extra relevance if
you create your products - that is, not just a webshop. The reason for this is
that you then also have an influence on the product design.
This makes this approach especially interesting for companies that offer
digital online services. Such companies have the opportunity to adjust product
characteristics (fast as they are digital), change their market focus, and the
customer group for which your service is to be designed.

33
You can learn more on this concept in the book “The Lean Start-up” by Eric Rice.

107
 AI FOR CEOS

Overview of your poten als

We have now gone through the five areas that define your AI potential on an
operational level.
However, we are not done jet. You need to factor in one more element. That is
the fact that there are some conditions where the AI is more or less applicable.
Those limitations have nothing to do with the value approach that we have
just gone through. They are more general conditions that could determine
whether or not AI is suitable for a given task.
So, you need to know of them because they could help you clarify if the
candidates that you have identifies actually will work in real life in your
business.

Valida on of the project candidates

Although you may now have found several problems that could be interesting
for you to solve with AI, it is not necessarily the best tool for you to use.
The reason is that AI as a technology has a profile where it is suitable for
some types of tasks and less good for others.

108
 OPERATIONS AND AI

Specifically, you have to deal with six different conditions which, overall, tell
you to what extent AI would be able to solve your tasks well or not:

• Social: Is it a task that requires interaction with people? Examples of such

tasks are teaching, care for the elderly, or childcare.
• Cognitive: If the nature of the task is intellectual and therefore requires
knowledge, reflection, and other thought activities, then it is a cognitive
task. Such tasks could be calculations, analytics, and legal advice.
• Physical: Does the task require you to do physical work to solve the
task? Tree cutting, sorting of parcels, or construction of buildings are all
physical tasks.
• Complexity: Is it a task with many elements in play that can make the
solution of the problem difficult to identify? Sending a rocket into space,
doing a weather forecast, or doing medical research are three examples of
tasks that require mastery of a lot of complexity to solve the task.
• Frequency: How often should the task be performed? It matters whether
the task is carried out often and in large volumes, for example, stock
trading, or whether there is a long time between tasks, like for example

109
 AI FOR CEOS

when you are building a new ship.

• Predictable: Is the outcome of the assignment likely, or is it unclear how
the solution might end up looking? In a lottery game, we know that ball
number 39 will not come out if the game has only 38 balls. So, feasible
outcome scenarios are known and predictable. A coaching process is
different. Here the result cannot be predicted because it depends on the
dialogue between the coach and the client.

As you can see, the parameters point in different directions and have different
focuses. The sum of how your potential AI tasks scores on each of them
influences whether or not AI will be the right approach for solving the task.
The first three social, physical, and cognitive have to do with different work
situations, and we start by looking at them.

Work Situa ons

The starting point for looking at work situations is an analysis made by

McKinsey34 . Their focus is on three different capacities it takes to solve a
task. A work assignment can be categorized on a scale in one or more of these
three categories:

• Social
• Cognitive
• Physical

The social tasks

Social tasks that deal with human interactions are generally challenging to
replace with AI.
The reason is that the current level of AI sophistication is relatively low.

34
https://www.mckinsey.com/business-functions/mckinsey-digital/our-insights/where-
machines-could-replace-humans-and-where-they-cant-yet

110
 OPERATIONS AND AI

Remember the talk of developing Super Artificial Intelligence. The prerequisite

for reaching that stage is that AI matches all the levels of intelligence that
humans possess. And AI, as we know it now, is not there yet. In particular, it
lags on the parts that deal with communication and interpretation of human
non-verbal communication.
So, technology is limited in the social tasks at this point in time.
However, that does not mean that there are no social tasks that cannot
be solved by AI. They are just not part of the core of the social tasks that we
perform daily. Robots already exist in hotels that can help guests check-in
without human involvement. This YouTube clip shows how it can be done.
And yes, it is in Japan:
https://www.youtube.com/watch?v=mpzIQt6l4xY
So, one should not dismiss all social tasks as impossible to be solved by AI.
There will be niches like this, and in the long term, AI will be able to make its
entry here as well. But, as a starting point, very demanding social tasks would
be a bad idea to try to make AI solve for you.

The cogni ve tasks

Cognitive tasks require clarification of knowledge. A lecturer at a university

distributes knowledge to her students, a doctor diagnoses a patient, a devel-
oper compiles code based on his experiences and knowledge of IT languages.
Cognitive tasks can be categorized according to their predictability. The
phoner who follows a fixed conversation flow is an example of a very pre-
dictable cognitive task. At the other end of the scale are very creative tasks such
as artist, composer, and other artistic areas, all of which are very unpredictable
in the way tasks are solved.
AI is a good candidate for solving tasks with a medium cognitive level. So, not
too demanding or too simple. There are different reasons why the extremes
are not so well suited for AI.
Simple cognitive tasks are calculations and monitoring. They are mechanical
actions that can be described with a logic called “if-then-else”. If this happens,
then do this.

111
 AI FOR CEOS

Thus, they fall into a category where conventional IT has its strengths, and
one would usually choose to let IT solve such tasks. So, it’s not because they
can’t be solved with AI. IT is just better here, and you probably already use IT
for those kinds of tasks.
At the other end of the scale, are complex and reflective knowledge tasks, as
well as tasks that require imagination and creativity.
AI cannot research, create art, or challenge our worldview with provocative
editorial articles in a newspaper. So, the kind of tasks that require cognitive
reflections is inadequately solved with AI.
The two examples are the extremes ends of a scale. We can use the
Knowledge Pyramid as a reference once more. Remember that data processing
is at the bottom of the pyramid, wisdom and reflection are at the top. And two
examples of extremes I have shared with you are at each end of the pyramid:
the data layer and the wisdom layer. AI is not great in those two but can have
high potential in the pyramids middle layers.

The physical tasks

Many work situations have a physical dimension too. It can be anything from
moving goods from A to B, to producing a physical product or tree cutting.
The process of creating a Mars bar at a factory is a series of production
processes. Here the raw materials are processed in a way so that the end
product is a series of chocolate products. They all have precise and consistent
quality. So, the production process is a fixed sequence of physical tasks.
At the other end of the scale of physical tasks could be the renovation of an old
house. It typically involves many different problems solving tasks with mutual
dependencies. It means that the renovation process creates unpredictable
situations along the way that indicate changes to plans, etc.
Physical tasks could have a simple level of complexity. For example, putting
a label on a bottle on an assembly line. In many such cases, an IT system will
be the best solution to manage the problem.
However, if the task requires a higher level of reflection, adjustments, and
flexibility, then the IT solutions often fall short.

112
 OPERATIONS AND AI

For example, vacuuming a floor in a warehouse. Such a type of task requires

that you monitor the surroundings, that you know where you have already
cleaned, and that you know where it remains to be cleaned. Such tasks are
more suitable for AI-controlled robots than IT.
It is because AI robots can navigate in such levels of complexity that the
demand for AI-based industrial robots is booming now.
If you can automate physical tasks that were not possible to automate before
then, you would have created efficiency. A potential even bigger benefit could
be that you now also can produce more flexible, because AI robots do not
depend on having to produce the same product the same way over and over
again.
Remember the Tesla example. We assumed that the reason to keep the
number of colors a car to a minimum was to streamline the production process
by reducing complexity.
With, AI complexity and efficiency do not longer have to be two opposing
forces.
An AI-controlled production assembly line has the potential to be able to
produce many different product variants without compromising production
efficiency. This can be achieved because production with AI is still automatic,
but it offers the possibility of supporting individual specified product produc-
tion in a mass production line set-up.
This gives customers more relevant products that can be produced without
cost increases.
So that means that Tesla might start producing green cars if they can find a
way to do it with AI, which does not conflict with their efficiency concerns.

The complexity and predictability of the tasks

The picture beneath shows how a Knowledge System can take over tasks, and
we have now added a new element. Namely, the level of complexity that a task
can have. It can be either predictable or unpredictable.

113
 AI FOR CEOS

Usually, when one would have to describe complexity on a scale, most people
would choose a range that goes from very complex to very simple. Because we
as humans understand complexity to a certain level. So, looking at complexity
in that way reflects the way we can process knowledge.
But it is not relevant to see the world that way when the focus is on AI.
Here, it is not the complexity level that determines whether AI can solve a
task well or not. AI is super good at dealing with very high levels of complexity.
That is not a problem.
But there must be a learning base for our AI. That is why, on an AI scale,
we deal with predictable and unpredictable complexity, and not as we do as
humans to see the range from complex to simple:

• Predictable complexity: The goal and outcome areas can be easily defined.
The correct behavior can be learned quickly from AI (and perhaps with
limited data available)
• Unpredictable complexity: Objectives and outcomes are unclear. Here, it
is difficult for AI to learn what the correct behavior is

114
 OPERATIONS AND AI

Therefore, a fundamental premise for the introduc on of AI is the

scale from predictable to unpredictable complexity. The easier
the problem output can be predicted, the be er it can be to solve
with AI.

Predictable complexity

Predictable complexity is if the actors in the area you look into have predictable
behavior.
This point can be illustrated with a bank example.
Banking customers who need phone service from their bank typically need
information or advice on products or have specific questions about their
account transactions. If they have questions about products, the bank knows
from experience what customers would ask for, and thus may have prepared
answers to many such issues in advance.
In other words, the complexity that needs to be addressed is relatively
predictable. Experience has shown what the customer needs are and how
to best give them the information they need. And - this is important - it is
possible to define precise success criteria for when the tasks have been solved
with success.
Knowing the span of possible outcomes and when it is solved with success
is central to AI.
It means that with enough data, AI can learn what the right behavior is when
a customer wants different types of advice. Once the possible outcomes are
known, you also know how to train AI to solve them for you.

115
 AI FOR CEOS

Unpredictable complexity

An example of unpredictable behavior is when the same bank plan for a

meeting with an entrepreneur. In such a case, the bank may not necessarily
foresee what new projects she wants to talk to them about when she shows up
for the meeting in the bank.
The entrepreneur may have come up with an idea for a project that requires
her to buy a building, or she may have made a deal with a Canadian partner,
so now she has a currency exposure in Canadian Dollars that she would like to
have covered.
In other words, her behavior is unpredictable, and the complexity (tasks)
that the bank employee must be able to handle is also more difficult to predict.
And the success criterion for the outcome of the tasks that need to handle is
challenging to make specific. Such conditions make the task difficult to solve
with AI.

Frequency of the task

There is another factor to consider, and that is the frequency at which the
incident occurs. Maybe you know that 15,000 customers always call in the
first week of every month with questions about their bank commitment.
So, a high-frequency task that will be valuable to replace with AI. A less
frequent job like meetings with business customers that occur on an ad-hoc
base between 25 and 100 times a year is probably not a cost-efficient case to
replace with AI.
So the volume also has an importance on the relevance of the case.

How far can you come with AI and solve the problem?

The sum of the reflections we have just reviewed can be illustrated in this
overview. The aim of it is to help you get an overview of where AI has strengths
relative to the sum of the attributes of a task.

116
 OPERATIONS AND AI

We know that a task can be either solved as a job, a process, or with AI. The
six validations areas can be seen as a requirement for any task that you would
like to address. This approach leaves us with a matrix of 24 different job
requirements. The span of outcomes is shown in the above table.
It is then possible to identify the best candidate for the job if you have to
solve the task by either a job, a process, or with AI for each cell in the matrix.
As an example, let’s say that the task was to host a reception.
The overall job requirement is probably relatively predictable. You know
who is invited and when the reception is to be held. So, this task falls into the
“predictable” category.
Arranging and conducting a reception, however is a relatively complex task.
Many people need to get together and interact in a social context. So, it is a
high complexity job in our score.
The next category is frequency. It is probably relatively rare for a reception
to be held. So here we end up in the “low” group.
The last element will evaluate if the job is mostly a social, physical, or
cognitive in nature The idea of a reception is for people to mingle and interact,
so this places the task in the social category.
Following this set of steps places the best way to manage a reception as a
job.
To most people, this would also be the most logical conclusion. Of course,
the best way to manage a reception is to make a human do it as a job.
The thing is this; not all the cells in the matrix have “jobs” or “process”

117
 AI FOR CEOS

as the best output. Your focus should be on where the conclusion is AI. I will
elaborate on why now.

The first impression when looking at the matrix is that AI does not seem
especially dominant. It seems like AI does not pose a particularly significant
threat to how we are solving tasks today. There are 24 fields, and AI is only
the preferred choice in the 7 of them.
The problem is that the volume of the tasks is even not distributed across
the 24 cells. The Mckinsey analysis looked at 2000 different job types in the
United States. The proportion of predictable physical and cognitive jobs was
65% of the total working time.35
So even though AI covers only 7 of the 24 cells, the volume of jobs in the
seven fields represent a significant proportion of the total amount of jobs.
That’s why McKinsey concluded that 40% of all US jobs are at risk of being
replaced by AI.
So, from a volume perspective, many tasks are candidates to be able to be
taken over by AI.

35
You can download the executive summary of the McKinsey report “JOBS LOST, JOBS GAINED:
WORKFORCE TRANSITIONS IN A TIME OF AUTOMATION”s a PDF on this link:
https://www.mckinsey.com/~/media/McKinsey/Featured%20Insights/Fu-
ture%20of%20Organizations/What%20the%20future%20of%20work%20will%20mean%20for%20jobs%
20skills%20and%20wages/MGI-Jobs-Lost-Jobs-Gained-Executive-summary-December-6-
2017.ashx

118
 OPERATIONS AND AI

The consequence of AI taking over jobs is one of the most vital areas of AI,
and you could write multiple books on this topic alone. I will address it once
more in the strategy part of the book.
For now, I will not go further with reflections on the topic. If you are curious
about your job and AI, then we have a tool on congifirm.com that you can use.
We have made an algorithm that calculates the likelihood of any job been taken
over by AI. It is based on six questions that you answer, and you can find it on
this address:
https://cognifirm.com/will-ai-take-over-your-job/
At cognifirm.com/tools you can also create a profile of all the projects that
you have identified from the elements that have just been reviewed.
This means that you can enter the projects, and by scoring them on the 6
different parameters, you get an assessment of the project’s AI maturity on a
scale of 0 to 100.
Whether AI is the right path for you or not depends on the degree of
correlation between your project’s score of the 6 parameters and the ideal AI
profile for each of the six areas. You can see that as an overview below.

As mentioned, the online tool will give your projects a score of 0 to 100, and at
the same time give you a red, yellow or green assessment of the potential of
your AI projects.
Green projects are good candidates for being taken over by AI, yellow ones

119
 AI FOR CEOS

can be implemented but with reservations. And finally, red candidates are not
tasks you should be solving with AI.
If you would like to work out an overview of your possible AI projects in your
company, then I recommend you use this feature. It makes it easy for you to
get an indication of whether the candidates you found could also be practically
solved with AI.

Next steps

We are now getting to the end of this section of the book. I hope that I have
shown you that AI has real value in many aspects of your daily operations and
how you run your business.
The important point, for now, is that you hopefully can see how AI can be a
way to achieve your goals. Because AI can help you find new ways to create
value.
But there is one more dimension to AI that has not yet been addressed by
me yet.
The focus we have had so far has been on how AI can help you get better at
what you do. It’s just not all AI can do for you. So, assessing a technology on
this perspective alone is a much to narrow approach.
In fact, it is fair to say that you have not even been introduced to how the
technology can create the greatest value to you yet, because:
AI can change the premises for how you can run your business.
The technology can challenge the premise of your business and what you can
do with it. In other words. There is a strategic dimension to the technology.
How to deal with AI from a strategic perspective is a big area that may prove
to be far more critical than how technology can help you with the tasks you
have right now.
Therefore, the next section is about how to use AI strategically.

120
 4

Strategy and AI

Should you become employeeless?

T
he previous section was about what AI can be used for and how the
technology can solve specific tasks for you in better ways. In other
words, an operational perspective.
Now the focus is on how AI has the potential to drive change that can
transform your business even more fundamentally. The driver is AI’s ability
to create strategic change in your industry.
AI is a strategic driver of change because, as a technology, it has the potential
to change the way you run your business and the way you compete.
What this means is that you can create business scenarios where AI can drive
exponential growth.
One of the reasons why AI can do this for you is because it makes it possible
to process knowledge and solve tasks without the capacity limitations that you
are experiencing today. This chapter will teach you how to look for areas in
your business where such kind of exponential growth cases would be possible
for you to achieve.
There is another strategic dimension that we will cover too. Using AI is
another way of creating value that can both complement and compete with
your employees and the way they perform their jobs. One of the dilemmas you

121
 AI FOR CEOS

now have to face is the extent you want to use the technology.
As you will learn in this section, some cases will accrue where your business
could be run without employees. You could become employeeless. So, you, as
a manager, have to decide if this is a way you want to pursue. And you will
learn that the answer is not as black and white as it might seem now.
So, there is a strategic dimension to how your tasks should be solved in your
business. As a manager, you need to decide the level of AI use in your company,
and to understand the consequences of using AI at different levels in your
business.
Until now, we have had a fairly internal look at how AI could change your
business. Meaning that we have not weighed any AI-driven actions that your
competitors might take. However, everything you can do with AI, so can your
competitors.
The last part of the strategy section helps you understand if AI can become
a competitive parameter in your business.
This section will help you identify where in your operations AI could create
competitive strategic advantages; it will help you see how to focus your AI
initiatives from a competitive perspective.
Finally, you will learn to know the consequence of AI-driven distortion in
your industry and evaluate the likely risk of this for your business.
Thus, there are three areas of strategic importance that are discussed in this
section:
• Exponential growth and value creation
• Scaling your business on feature, function, or corporate level?
• AI and competition
In Greenland, they have a saying. You will not be eaten by the Polar bear
that you are watching out for. You will be eating by the Polar bear that you are
not paying attention to.
This chapter will help you not to get eaten by the AI polar bear that you do
are not paying attention to, so to speak.
The aim is to help you get the right strategic perspectives on how you could
and should progress with AI with the right level of management attention.
So, once you have read this section of the book, you will have a basic

122
 STRATEGY AND AI

understanding of the commercial dynamics that can be driven by AI, so that

you can make a more qualified decision on the extent to which the technology
should have focus in your organization.
When reading the section of the book, you will realize that the three strategic
parts that you are about to be introduced to will not be of equal importance to
you, and they should, therefore, be addressed by you with different levels of
attention.
To help you prioritize your approach to how you should treat each of the
three areas, you will be introduced to a simple framework for evaluating their
importance to your business. This is where we start our journey on your
strategic AI focus.

How do you know how important AI is to you?

In order for you to place the importance of AI in a strategic context, it makes

sense to use a scale with levels of strategic importance. The higher the impact
level of AI will have your business, the more seriously you will need to address
this area.
To help you do this, you will now be presented for a simple four-level
strategic framework. You can though use this to judge your business on each
of the three areas where AI has a strategic consequence. The framework looks
like this:

123
 AI FOR CEOS

As you can see, your strategies can be viewed as a hierarchy of strategies with
increasingly importance to your business. At the bottom are the operational
strategies, and at the top are strategic initiatives that have a core consequence
for your entire business.
The strategy levels are defined according to the depth of change they make
on your business, measured on four levels:

• Corporate Strategies: The unifying strategy for your business that sets
the direction you want to go. The corporate strategies are the sum of your
business activities, and therefore, there is a close link to your business
strategies.
• Business Strategies: The level here is your business areas. That is, the
main areas your business is organized into. The typical cases would be:
Sales, marketing, production, or IT. So, your business strategies are closely
linked to the way your business generally is organized.
• Function Strategies: These are specific strategies within a business area.
For example, your IT strategy that could include decisions on, for example,
increased outsourcing or the use of Cloud solutions.
• Operations Strategies: Within your business areas, there could be strate-
gies on how to solve specific issues. For example, you could have an IT
security strategy or a strategy for how to launch a new product.

124
 STRATEGY AND AI

You can thus use the four levels to rank the strategic importance of your AI
initiatives. How to do this will be an integrated part of the three strategic areas
of this section.
Getting back to the Polar bear example, this means that such an approach
helps you keep an eye on all the polar bears, including the ones that, otherwise
might have ended up eating you.
So with that in mind, we start with the first point, which is your potential
for creating exponential growth with AI.

How to create exponen al growth

The operational elements discussed in the last section showed that AI can
make your organization more productive, and can create new commercial
opportunities that can drive your sales. So a cheaper organization to operate
that sells more.
There is one more dimension that the operational elements of AI can develop
into. That is to enable you to create exponential value in different sections of
your business.
You already saw the mechanism behind the concept in the photoshoot
example. By removing the elements that create capacity constraints, you
will be able to develop much value at no extra cost.
If the right conditions are present, value growth can become exponential.
Thus, a situation where, for example, an effort of 10 gives 100 more in value,
and an effort of 20 gives 1000 more in value.

125
 AI FOR CEOS

You are about to learn the dynamics of how this can unfold using AI, and how
to know if you can apply the dynamics into your business.
You already know the dynamics that drive such growth rates from digital
solutions. The most classic example is digital communication.
The Internet has made communication digital, and thus almost free. Once
you have created a website, for example, it costs - almost - the same for you
to run it if ten people a day read your web content as if 10,000 people visit your
site.
Of course, that is not entirely true, as there are definitely additional
operating costs for large volume sites. But, compared to the “non-digital
alternatives” which could be to call the same 10,000 people, send them a
letter, or talk to them at a trade show. Then, digital communication is free to
scale up.
It is the same dynamic we are now seeking with AI. However, with the
difference that it is not the digital (information) we now are getting for free,
it is knowledge and automated decision making.
The difference between the potential business impact between AI-driven
growth and digital-driven growth is almost beyond comparison.
To get the right perspective on why this is so, we return briefly to the

126
 STRATEGY AND AI

Knowledge Pyramid:

Free data and information have catalyzed the ability to create exponentially
increasing value for Internet solutions. The same mechanisms will also work
when we look at what happens when knowledge and decision making becomes
free.
You can now have the potential similar growth dynamics in areas where you
work with knowledge and decision-making.
Your new mindset in this context should be to think of tasks as something
that potentially could be solved free of charge for you, even in a way without
any capacity constraints. So, the focus goes from communication to solving
tasks.
In popular terms, you could say that the internet made it possible for you to
reach 1 billion people for free. With AI, you can solve 1 billion tasks for free.
Once you have set up your Knowledge System, it does not cost extra to
make the system solve one more task. So, the same dynamics as with digital
solutions. It does require an effort to get started, but solving 10 or 10.000 more
tasks is pretty much free.
The gap between the internet and AI potential can be illustrated, if we return
to the knowledge pyramid.
If you look at the knowledge pyramid, then digital tools mean solving
tasks with a relatively low level of complexity (e.g. displaying content in

127
 AI FOR CEOS

a web browser). Because digital is IT and that implies working with data and
information in the two-bottom layer of the pyramid.
AI will be able to conduct a telephone conversation, book meetings, and
other similar tasks that are more complex and demanding. Such tasks are on
the two top layers of the pyramid because they involve knowledge and to a
certain extent, wisdom.
The same dynamics apply to the market side. Once you have created a
marketing bot that has learned to book successful web banner ads that generate
sales in Denmark, then the step to do the same in Sweden, Germany, and
England is not a very big task.
The prerequisites for success are obviously different if you compare an AI-
solved task to a digital communication task. Also, it is not all places where you
work with the knowledge that you would be able to work with AI in this way.
In other words, you need to have a more detailed view of the dynamics to be
able to apply them to your business. You can reach exponential growth with
AI in two phases. This is how you should look at it.
The value that AI creates as an exponential growth curve can be shown as
an increasing curve, as illustrated below. The cure shows a case where the last
effort always produces the highest added value.

As you can see from the figure the value-added components are driven by two

128
 STRATEGY AND AI

underlying phases with two very different purposes:

• Phase 1: AI as a solution to a problem

• •Phase 2: Scaling the value creation

Their character and the dynamics behind the two are very different.
You always start by using AI to solve a specific problem that you have. The
previous section of the book was about how you could use AI to solve different
types of operational problems for you. It is the first of the two stages.
The next phase is initiated if you then start focusing on your capacity
bottlenecks. The aim is now to design your organization in a way where
you solve your problems without capacity problems. If this is possible, then
the preconditions for the next part are present. In other words, where the
exponential value creation takes place.
Remember the red dress case? Removing the need for a photoshoot was your
solution to a problem. Thus, Phase 1. The opportunities it created for agility
in your production flow afterward are what drive the exponentially increasing
value. Your Phase 2. This concept is shown in the picture below.

129
 AI FOR CEOS

Phase 1: AI as a solu on to a problem

The goal in the first phase is to create value with AI in a way where you use the
technology to solve a specific problem for you. In the operational section of
the book, you were introduced to 5 such areas:

• Efficiency: Help you do more with fewer resources

• Managing complexity: Helping you solve tasks that are too difficult or
impossible for you today to solve.
• Service: Providing a better and more personalized service
• Product development: Create products with features that you would not
be able to make today
• Market Agility: Help you sell more to more

Thus, an operational focus on specific actions that create tangible value.

130
 STRATEGY AND AI

Phase 2: Scaling the value crea on

Let’s return to the photoshoot example. Let say that you have just installed
the Computer-Generated Image software set up at your business.
You would then start to use it at a smaller scale, to learn the technology, how
to use it, and to find your most apparent Use Cases. As you gain experience
with how to work with the technology, an ever-increasing proportion of your
photos will be made as Computer Generated Images.
At this point, the crucial pictures are probably still made, live in a studio,
with a real model and a photographer.
However, the desire to be able to have even more photos accessible with no
production time or cost will start to grow stronger on you. Because relying on
a photoshoot creates limitations in the different branches of your production
set up. Mostly in your production, sales, and the ongoing customization of
the sales material.
So, in Phase 2, the goal is to remove the constraints that inhibit exponential
growth. That is why it is called “Scaling the value creation”. In our case, it
means unlimited access to new product photos. When you reach that point,
you have removed a capacity restriction, and it changes the preconditions for

131
 AI FOR CEOS

how you can run this part of your business.

Capacity management gives you exponen al growth

When you implement your first AI solutions, you will probably experience a
process with parallels to the photo example.
Namely, it is not the problem that the AI solution was designed to solve that
is the limiting factor for your ability to scale your value exponentially. The
limitation probably lies in other parts of your process flows and tasks that you
perform.
To know how to spot where you can scale exponentially, you must under-
stand this in more detail. So, for you to be able to do so, we start with a short
recap on how you create value in your business.

The way you create value

The reason your business exists is because you can create value for others
through the management of complexity. This concept is shown in the
illustration below.
The illustration is a way of showing a value-creating flow in your business.
The idea is to show how you create your products and service, guided by goals
set by management.
Your goals are then converted into specific output requirements that must
be created to reach your goals. To achieve your output, you need to solve tasks.
Tasks can be solved either as a job, a process, or with an IT system.

132
 STRATEGY AND AI

And now we’re back to capacity constraints and the potential for exponential
growth. Because, as you may have noticed in the picture, processes and jobs
are marked with a darker background. The reason for this is that they have a
built-in capacity constraint.
There is always a built-in capacity constraint in tasks that are handled as a
job or process.
There will always be a limit to how many lawyers and customer service
employees you will be willing to hire, and all people have a maximum limit
on how much they can deliver in a working day. Therefore, there is a built-in
capacity limitation in all organizations.
But it doesn’t have to be like that now.
Knowledge Systems have the potential to be able to solve tasks that previ-
ously could only be handled by jobs, processes, and IT solutions. And with
this, opportunities arise to remove some of the capacity constraints that you
are experiencing today.

133
 AI FOR CEOS

AI as an alterna ve

A Knowledge System is always designed to solve tasks. Either an analysis

task, a management task, or a communication task. The curtain van example
contained both a communication task (book an appointment with a new
customer) and management tasks (send mail out, etc.).
A Knowledge System thus also creates business value by dealing with
complexity just as you do today in a company when an employee is performing
a job.
So it makes sense to place a Knowledge System in the creation row in the
illustration because AI is creating value for the same reasons that you deploy
people, processes or IT in your business.
However, there is a significant difference between solving a task as a job,
a process, or with AI. The difference is scalability. Unlike the usual way of
dealing with complexity, there are no capacity constraints in a Knowledge
System.
As you probably remember from the beginning of this section, I told you that
AI has the potential to scale value creation exponentially by going through the
two stages:

134
 STRATEGY AND AI

• Phase 1: AI as a solution to a problem

• Phase 2: Scaling the value creation

We return to the redress example to understand how AI can be used to scale

your value creation processes.
The first phase will focus on solving a specific task that creates value for
your business.
In this example, it is the use of Computer-Generated Images (CGI) to create
marketing material. It solves a problem here and now, as it is cheaper and
more flexible than the alternative, which is an expensive photoshoot.

A consequence of our phase one implementation would be that jobs and

processes would change. After all, you no longer had to work with (so many)
photo models, and the job of having a photoshoot will gradually disappear.
So the fact that we now use Computer generated images affects the condi-
tions on with our promotion pictures, and our promotion materials can be
created.
As you remember from the case, you still take some photos in a studio.

135
 AI FOR CEOS

Thus, some of the old conditions for how you create value is maintained. As
a consequence, you will still have scalability constraints in your processes
and jobs. Therefore, you do not reach the realization of the full potential of
technology at this stage.
In Phase 2, the goal is to redefine the premise of how your value is created,
thus creating the conditions for exponential growth and value creation.
As you may remember from the photoshoot example, there is great value in
being able to reach a state where you do not have to take any pictures at all.
By not having to take the pictures physically, it would mean that you could
have marketing material ready as soon as your concept was complete. You
might be able to have your production batches at shorter intervals, and maybe
you could also handle more production batches in smaller sizes because your
marketing cost will be reduced.
The achievement of the gains that I just mentioned can all be traced back
to the capacity problems of your jobs and processes. You want to remove
them, because the result of this will be a higher degree of flexibility in your
production.
The goal is now another one than in the first phase. In phase one, you made a
solution to solve a specific problem. Now you are focusing on creating a set up
where you utilize what you have created in phase one to do things differently.
So, you challenge the conditions for how you create value.
Succeeding in Phase Two means creating the foundation for exceptional
growth in your business. So, a much more attractive value proposition that
what you could achieve in the first phase.
The goal now is to have some of the tasks that are solved by processes and
jobs today removed. The consequence of achieving this will be an organization
that has unlimited capacity because all the components of the tasks now can
be solved by a Knowledge System.
If you succeed in bringing the organization to such a state, then you will
have the unlimited capacity in the number of tasks you can handle.
The idea is illustrated in this illustration:

136
 STRATEGY AND AI

However, there is much more to it than just removing tasks from your
employees. What you are doing in phase two could have a much more
significant impact on how you run your business. You might be able to re-
calibrate how you create value and how your business could be run. Here is
why.

Calibra ng your value crea on

You should look at your options in Phase two as a calibration process. You now
have the option of re-calibration the ways on how you create value.
The illustration that you have just seen shows that AI will initially replace
IT, jobs or processes. This is what happens in Phase One, and in the five
operational areas that are described in the previous section of this book.
However, AI handles and manages tasks differently compared to how you
would have done using either the jobs, processes or IT to create value.
Let´s go back to the photoshoot example. Computer Generated Pictures
is not a better way to take your pictures in a studio. It is a better - but also
different - way of producing your desired output. It creates the same result,
but in a different way.
Your ability to create the right conditions to reach the exponential value

137
 AI FOR CEOS

stage depends on the extent to which you succeeded in rethinking the

premises for how you create value. And you do this by looking at the
complete perspective from goal to complexity. That is why this illustration is
reintroduced to you.

Your goals

If we step back and reflect on goals, your overall business goals remain the
same, whether you try to realize them with or without AI.
Your goals reflect the specific expectations that your business management
has for a given area.
As you remember from earlier, the feasibility of your goals is determined by
a combination of two opposite factors. On the one hand, your management
would like to move your business forward as fast as possible. On the other
hand, it must consider their organization’s ability and capability to deliver on
the goals.
In the curtain van example, it makes no sense to have a goal of getting
100,000 new customers in 12 months because it would be unrealistic for the
organization to handle such an expansion.
At all levels, “goals” are a balance between an organization’s expected ability
to deliver, set against a desire for business progress.
This balance is shown in the figure below.

138
 STRATEGY AND AI

As you can see, your organization’s ability to deliver ultimately is a conse-

quence of your ability to handle complexity. Thus, a weighting between the
two boxes, as shown below.

The two boxes must be in balance. On the one hand, it is your organization’s
ability to deliver. It will be determined by how much complexity you face, your
ability to handle it, and the tools you have to do so.
If you are a Call Center, then your complexity is the conversations you need

139
 AI FOR CEOS

to have with customers. Your ability to handle complexity is your employees,

the processes they follow, and the IT systems that support your processes and
jobs.
You have a limited number of employees, which means that there are
capacity constraints on how many customers you can serve. So, there is a
natural built-in limitation on how far you can move your business forward.
However, you will be able to improve your processes, reduce sick leave, etc.
So, on a smaller scale, you can make several incremental improvements, the
sum of which can make you perform better. Your current goals are probably
defined around the achievement of improvement in such areas. And your
business goals will thus be tied to an assessment of capacity and progress.
And now we get to the AI part. The introduction of AI in the form of bots
in your call center means that it can replace tasks that would otherwise be
handled by jobs, IT systems, or processes.
This, however, it just one of the consequences of using AI in your Call Center.
You gain much more because a Knowledge System can handle a new kind of
complexity.
As an example, now you can offer individual and personalized counseling to
your customers. That kind of service might otherwise have been too expensive
for you to offer until now. The reasons could be that it might require staff with
a higher professional education who would also need to spend extra time on
each client.
So, cost and access to capabilities stops you today from offering such a
service. The cost-benefit case cannot justify it if you must use employees time
as part of the service.
AI can affect your ability to handle two levels of capabilities. There is
potential to both increase your capacity and increase the quality of your
deliveries. This principle is illustrated in the figure below:

140
 STRATEGY AND AI

The point of this figure is to show how AI could be included as a strategic

component.
AI can change the premise of what goals you can set for your business
because your organizational delivery scope is changing. This is shown in
the figure below.

If your goal for the organization is defined without consideration to the

opportunities your Knowledge Systems give you, then your focus will be too

141
 AI FOR CEOS

narrow.
For you can do more and different things now, and that new reality must be
reflected in the (new) goals you have to set for your organization.
It is the ability to be able to do more and to do something different with AI
that drives your opportunities to achieve exponential growth. Because you
can increase your capacity - perhaps with a few restrictions. And you can
increase your quality - because you can handle others and more demanding
tasks without compromising your efficiency.
Remember the telco case from the previous section? A bot could provide
advice on roaming coverage at the customer’s cottage in a way that would
otherwise not have been possible - even for a customer service employee.
It is an indication that they now deal with an increased amount of complexity
that is handled automatically with a bot. The consequence is that such a
telecommunications company now can work with other business goals because
it now has a different pallet of possibilities.
The curtain van example also shows how AI enables capacity to be increased.
Bu using AI it is now possible for them to handle many conversations
automatically. Again, it means adjusting their business goals because such
automatic customer management provides a new set of opportunities that
they can utilize.
So your ability to design AI solutions that both solve a specific problem
(Phase 1) and challenge the requirements for how you create value (Phase 2)
is the basis for whether you can create exponential value increases with AI.
As the examples you have presented here show, it is the understanding
dynamics of how AI can create a new kind of value by solving your tasks
differently, which is the criterion for obtaining the exponential value.
If you want to use AI for similar cases yourself, then such projects start
with reflections on how you create value and how you reach your output goals.
Whether there are parts of your value creation processes that can be challenged.
If they can, what would your business look like?
Consider the way your business is run. It may require an analysis assignment
from you before you can see the specific cases you may have. So, you may not
be doing it right now.

142
 STRATEGY AND AI

However, you can probably relate to the potential of reaching the stages
described in this section. Do you see AI as an operational tool? or something
that should have your closest strategic attention in your business going
forward?
Your evaluation of where you place the use of AI to create exponential value
is a part of the strategic assessment. If you like, you can find the tools to do so
on cognifirm.com/tools.

Scaling your business on feature, func on, or corporate

level?
You might have noticed that I continuously have moved the perspective on
your business upwards.
The operational section of the book was - operational. The previous chapter
showed you how adding a new level of strategic perspective could help you
see if the AI that you have created also could function as a driver for more
substantial strategic change in your business.
Now we take one more step in the upwards direction. The goal with this
chapter is to force you to reflect on if AI could drive an even higher level of
change in your business. If AI could change and challenge the very core of who
you are as a business?
The reason you should be interested in this perspective is that we have
so far seen that AI can create exponential growth on an operational level.
This happens when AI functions as a feature, like in Business Analytics. This
enables you to make as many insightful analytics as you like. We have just
seen that AI also could drive exponential growth on a function level too. That
was what happened in the red dress example.
Remember this illustration.

143
 AI FOR CEOS

The Business analytics case is at the lowest operational level in the Pyramid
above. The red dress is at a functional level 3. In both examples, we manage to
see how exponential growth could be created with AI.
This brings us to the purpose of this chapter. It is possible to use AI in the
two top stages of the pyramid too? Stages where the impact will be at the core
of what your business is about.
If a possible effect on the lower levels could create the prerequisite for
exponential growth, would that also be possible to achieve on levels of higher
strategic importance too?
Obviously, it is far more interesting to create exponential growth on factors
that relate to your business and corporate strategies. Because the effect of
such will be must more significant for your business.
So this is the sphere we move into now. As you will see, the dynamics and
consequences are quite different here, compared to the impact we saw on the
operational and functional levels.
Before we start this talk a warning: It is obvious that when we look at
factors that can automate the core of what you are as a business, then the
consequences and the impact can be brutal compared to how you run your
business today.
When you challenge the core of how you create value and how you run your
business, then there will be consequences. The level of change that we now

144
 STRATEGY AND AI

look into will have consequences for your employees and how your business
is organized.
So I understand if some people will read this section of the book with equal
fear and fascination. Because AI at this level will remove jobs at a large scale,
just at the way we see the word and how we are organized will be very different
from what we know today.
I will outline the effects as I see them, and that includes the organizational
consequences. One of the reasons I do so is for you to be able to reflect on how
far you want to take your business when it comes to the use of AI.
One of the ethical dilemmas with AI, in general, is that the technology can
be used in areas that we might not want to use it in, even if we could. It is not
normal for us to doze the level of tech that we use. The typical case is always
to get the best and smartest.
We know that the newest smartphone is always better than last year’s model,
and the better the technology we have, the better we are off. That is not
necessarily how you should look at AI. It is more like driving a Porsche 911
on a racetrack. You don´t win by using the cars power potential all the time.
You win by dosing it, so you use the power when you need it. If you don’t, you
won’t make it through the tracks first corner.
AI, at its highest corporate level, is like that. You could choose to use AI in
many different aspects. It might make sense, or it might not. You now have to
decide the dose of technology that you want to use in your business.
So, there is an ethical dimension as to how much you should work with AI.
There is another dimension too. The answer to the questions determines the
level of high-level AI use in your business. It is:
How much of your company’s intelligence should be artificial?
This question might be misinterpreted as a philosophical question. It is not.
It is the most fundamental AI business question that you will ever have to
answer.
The consequences of how you answer this question define your corporate
and strategic use of AI
which is why the rest of this chapter focuses on elaborating on the implica-
tions that a lesser or higher degree of artificial corporate intelligence in your

145
 AI FOR CEOS

business matters will have. Here is why:

You need to see your company’s artificial or - digital intelligence - relative
to your company’s overall intelligence. Your current intelligence is the sum of
what enables you to solve the tasks that are handled by your company.
Your employees’ knowledge, education, and experience help you solve tasks.
So does the knowledge that is incorporated in the way your processes and IT
works. So, you could say that your company has a total amount of intelligence
that allows you to function as a company and solve the tasks you need to
achieve to reach your goals.
The arrival of AI provides a new way of looking at intelligence in your
business.
In some areas, AI complements what you already do, so you get better. In
other areas, AI is superior to other forms of intelligence and therefore, will
take over tasks from them. These are some of the dynamics we dive into
this section. So you can understand the business weight of the changes that
technology can drive for you.
In other words, your choice on the level of AI use in your business is not
technical.
It is basically a profound fundamental choice on the core of what should
define your business and how you work with knowledge.

Today, it is the sum of the knowledge, you as a company has that more than
anything defines what you are as an organization.
So, if AI gets a more or less dominant role in your overall knowledge presence,

146
 STRATEGY AND AI

then it has the potential to have major consequences for how you as a company
function and how you define yourself.
This is the backdrop for what I call “The 5 AI Levels”. The scale describes
the consequences of ever more use of Artificial Intelligence at the enterprise
level.
It means that the scale spans from companies that use AI on an operational
level, that is bottom of the strategic pyramid, to companies where AI is at the
core of the business focus. So here, AIs are of the highest corporate importance,
and that places them at the top of the pyramid.

Why five levels?

The five levels describe the five different stages you may have in your overall
use of Artificial Intelligence.
One of the consequences of the highest level of AI use has already been
introduced to you. It is a company without employees.
At the other end of the scale, the AI use is at a feature level. So, minimal use
of technology. On this level, you use AI to solve an isolated specific problem.
Most such Use Cases are AI as an analytic tool because it is the most simple
and easy to use application of AI. This stage is where most companies start

147
 AI FOR CEOS

their AI endeavors.
You need to see the two examples as each end of a scale, and the five levels
are thus the steps that go from the simple feature level to the company without
employees.
The five levels are defined by the volume, complexity, and business impor-
tance of the tasks that the AI will manage in the different levels. The five are
defined in this overview:

I totally agree that you intuitively will place your business at Level 1 or 2 when
you see such an overview. And maybe also think that the last three levels are
Science Fiction, which has no relevance to you.
I want to challenge you on that through now. The last three levels say
something about the direction you might be developing in. It is your choice if
you want to go there or not.
And who knows, maybe they aren’t actually as far away as you think. You
will be introduced to an example of a real level 5 company later in the section.
So, for some companies, level 5 is their daily business, for others, it is Science
Fiction.
We start the journey through the five stages with a description of what
characterizes the different levels, and a reflection on what organizational
changes the different levels could drive in your business.

148
 STRATEGY AND AI

The content of the five levels

The five levels represent different maturity levels of how you use AI in your
business. The maturity level in each state will be described in two core
dimensions:
• Objectives: What is the business objective of the level
• Organization: What are the organizational consequences
So, one element that has to do with why you use AI from a business
perspective, and what you can gain. The other one relates to the consequences
of how you organize your work and design your business.
One of this book’s core objectives is the role AI plays when it comes to
handling of tasks. This approach continues in the section too. It means that
the stage walkthrough will describe what types of tasks you would be able to
solve at the various stages.
An essential consequence of the increased use of AI is organizational change.
There is a big difference between a set-up where you only use AI to perform a
single function, compared to a level where AI has made your business fully
automated.
The first stage however has few organizational consequences. It is called AI
as a feature.

Level 1: AI as a feature

The goal of using AI at this level is to use AI technology to solve one or more
specific and independent problems.
The most typical way to start using AI is as an analysis tool, where you use
AI to produce predictive analysis that helps you make better decisions.
For example, how to define the online campaign’s audience that will get the
best possible result such as leads, sales, or other positive customer responses.
Your use of AI at this stage is probably not the result of strategic considera-
tions to deliberately build an organization where AI is a knowledge parameter.
You might be using AI as an element of your anti-money laundering
surveillance processes, your marketing, and your financial management.

149
 AI FOR CEOS

However, there is no overall control or coordination on how the technology

is used. The consequence is that you treat AI as a feature used to solve one or
more specific tasks.
The consequence of such a setup is that you do not work with AI in a way that
can raise your company’s AI overall digital knowledge level. For example, the
tasks that you apply to AI in your finance department are solved independently
of the way your marketing department works with their AI.
So the answer to the question “What proportion of our total intelligence
should be artificial” is at this level:
Not much, because AI doesn’t mean a lot to us as a business.
If we return to the flow of solving tasks from goal to complexity, then the
role that your AI does for you at this level can be described in this way:

AI solves a task in parallel with the jobs, processes, and IT systems that you
normally use to run your business. Or, put another way. It would be annoying
if all your AI solutions disappeared. In some ways, you would be less capable of
doing your jobs well, but you could easily manage without AI in your business.
The fact that you could run your business without problems if AI disap-
peared also means that AI here does not have a significant impact on your
organization.

150
 STRATEGY AND AI

What you are experiencing here is that some employees get AI as a new tool
that they must learn to use. Either because AI is built in as part of the software
solutions, they already use, or because you have a specific AI-based solution
that solves a particular problem.
This is not the case at the next level. Now, AI starts solving more complex
tasks for you. It means at AI doesn’t just tell you what you can do to solve
a problem. At this stage, AI solves the problem for you, by talking to your
customers, by sending out emails, or by designing your product prototypes.
So it is a very different level of value that is created in level two, compared
to the first level of AI use. As you can see in level 1, AI is at the lowest level in
the strategy pyramid, the level called “operations”. The next level is step two
in the pyramid where AI manages a function in your business.

Level 2: AI as func ons

Your goal with AI here at this level is very different from the first level. Now,
your aim is to use the technology to perform functions and solve specific tasks
in your company by itself.
The curtain van is an example of a company in level 2. Here, AI performs
a function, namely booking meetings on the phone. So here AI performs a
service that would otherwise be conducted by employees.
In order to use AI to perform a function, it requires that your awareness of
AI on enterprise-level AI is more apparent than it probably was at the first
level.
You should look at it this way. You can use first-level AI to help you with
specific features, like analytics.
You use predictive analytics because it’s smart, but if the same solution could
be made without AI then you might as well have chosen it instead. Because you
chose a solution, not a technology, and for you at this stage, is no deliberate
choice that the products you use should be made with AI.
This approach does not work if the goal is to have AI to solve specific
functions for you, as it is the case in level two companies.
The difference between a feature and a function is that a feature is a task

151
 AI FOR CEOS

that is done for you, a function is a task that is done on behalf of you.
Thus, you begin to more actively relate to the question “What proportion of
our total intelligence should be artificial?” Maybe not consciously. But you
cannot solve functions at the functional level without having made thoughts
on how to organize your applications and knowledge so that you can solve
your tasks with AI.
You might not be able to answer the question “What proportion of our
company’s intelligence should be artificial?“. But now you are aware that part
of your company’s intelligence is artificial because it is a prerequisite to being
able to solve the tasks with AI that you want.
The big difference now is that you link AI’s various capabilities together.
Doing so is absolutely central to this level of AI use, and I will elaborate on
why later. For now, you need to focus on the fact that a Knowledge System has
three different dimensions:

• Assistant
• Management
• Communication

The curtain van example is a Knowledge System that is much more sophisti-
cated than just providing analysis of data.
Predictive analytics in the first level gives you the knowledge to make better
decisions. That’s it. Now, at this level, there is an element of automation build
in our AI solutions too. So, AI starts doing things on its own for us now. This
is why tasks now can be solved based on your company’s digital intelligence.
In the curtain van case, our Knowledge System performed both the task
of having a dialogue with the customer and at the same time booked an
appointment for them. It thus worked with both Management and the
Communication part of the Knowledge System. Thus, an interaction between
two of the three areas of a Knowledge System.
It is precisely because you have now built an extra dimension on the
intelligence of the entire company, that at this stage, that you have more

152
 STRATEGY AND AI

opportunities to solve more complex tasks.

When you start using AI to solve functions for you, then there is also an
organizational consequence.
Now you use AI technology in more direct competition with your jobs,
processes, and IT. Recall the photoshoot example or the curtain van. The
implementation of both solutions would have had consequences for the way
you would have otherwise solved your tasks.
This can be described with the graph showing the path from goal to
complexity as illustrated below.

The big difference between the first level, and now is that Knowledge Systems
are now starting to handle specific tasks or functions that would otherwise be
performed through processes or jobs.
Not only can the curtain van case take over tasks that are currently handled
by IT, jobs and processes. It can also happen in a way that supports exponential
scalability.
The curtain bookings bot has unlimited capacity. It makes no difference
whether it handles 1 or 1000 customers per hour. How to reach such a state was
highlighted in the strategy section in the section on exponential scalability.
So, it is at this stage that the capacity constraints in jobs and processes are

153
 AI FOR CEOS

seriously being challenged by AI.

It will be quite obvious which candidates are at risk to be replaced by the AI
function that you have developed. If it is the curtain van case, then it would be
the customer service staff who handle new bookings. All other elements of
your organization will probably continue relatively untouched.
Here, keep in mind a point from the previous chapter on exponential
scalability. Namely, what AI gives you is another way to reach your output.
It is not a goal of AI to eliminate jobs or processes. It is a likely consequence
because you can now reach your results in a different way by using AI.
You use Computer Generated Images to get product photos faster and
cheaper. You use bots to serve more customers. Both cases have consequences
for your old processes or jobs, but the goal of the effort was not to remove the
job as such. The goal was to create output in a new and better way.
What you may find is that AI could perform an increasing number of such
functions.
If you have a chatbot for your customers, then it is suddenly easy for you to
create a new version for your employees because the groundwork has been
done when you introduced your first bot. So the barriers to solving other
functions with AI become smaller because you already have many the elements
you need to solve them. And the more AI functions you create - the more
advanced your knowledge system becomes - the more obvious candidates for
AI will emerge.
Steadily increasing the use of AI will challenge the appropriateness of the
way you as a business do business and organize your tasks. That’s is what
happens at level three.

Level 3: AI as business func ons

You will reach a level three business when your AI is capable of managing a
complete business function for you. In other words, AI runs a part of your
business for you.
A business function is a set of complex high-level task that is required to
handle the core elements of your company. Our two examples can help explain

154
 STRATEGY AND AI

what this means.

In the red dress example, Computer generated images was a feature; it
created a picture. The creation of a picture is a part of a more complex
task of creating promotional materials. That is a function. The creation of
promotional material is again a part of your marketing efforts, that can be
categorized as a business function.
It is the same way with the curtain van. To manage bookings for customers
is a function, to manage your customer service is a business function, in which
the booking part is one of the tasks.
So, you can see tasks as a hierarchy that consists of features, function, and
business functions. In the curtain this hierarchy looks like this:

Remember the difference between a function and a feature is that a function is

a task that is done for you? In this example, the chatbot talks to the customers
for you. A feature is a task that is done on behalf of you. In the example it
is booking management. Booking management is a task that among other
things, involves talking to new customers, finding the right time for a meeting,
and sending out an appropriative calendar and e-mail invites.

155
 AI FOR CEOS

A way to understand the relationship is the “why” test. You start on the
feature level, in our example, the chatbot. You then ask yourself why we have
a chatbot? The answer is: Because we need to book an appointment with new
customers.” You then ask again, “Why do we need to book an appointment
with new customers? The answer is “To get new bizz sales”. So, the Why helps
you understand the context in with a feature or function exist and thereby
outlining the hierarchy between them.
The aim at the level three companies is to move the maximum AI capabilities
up from both a feature level (Chatbot), a function level (booking appointment)
to a business function level. At a business function level, the area itself is
managed and run by AI. It this example, the business level is “New bizz
management”.
A Business function is much more valuable for you to have managed
automatically than a function. The reason we want to do this is that we have
proven that on each level that can be managed by AI, we can create a foundation
for exponential growth.
The core driver for exponential growth is the unlimited use of AI at almost
no cost. So, in our case, unlimited and free New bizz management is obvi-
ously much more valuable than unlimited and free chatbots or appointment
bookings.
So, your task at this level is to design your organization in a way where AI
can manage all the tasks that define the business level that you are looking at.
This principle is shown here in two cases:
Case A:

156
 STRATEGY AND AI

Case B:

In case A, we have managed to solve all subtasks with AI and though have
created the foundation for scalable growth on a business function level. In
this case, you can provide an unlimited and free customer service for your
customers.
In case B, you only use AI for a specific function (newsletters), so the value
of the scalability is limited to managing and distribution of newsletters, which
is a feature.

157
 AI FOR CEOS

Your success in this level is tight up to your ability to redesign your

organizations so that you can solve all underlying tasks with AI too.
Let’s say that we now have succeeded in making our customer service
automatic and we what to see if we can achieve the same with our VIP service.
We can see from the illustration that we are ok with the Newsletters, but that
booking is still a manual function with two underlying features.
We now want to investigate if we can convert the booking area to be managed
by AI too. If we succeed with this part, then our VIP section will be free from
constraints, and this will enable unlimited use for our customers.
So, each of the underlying features and function must be addressed to see if:

• Our current AI can solve the task as it is now?

• If not, can better AI then solve it?
• If not, can they be rescoped or redesigned in a way so that AI can solve the
problem?

This principle is illustrated in his picture:

So the idea is first to investigate if your Knowledge System can solve your task
as it is now. If that is not possible, then you should evaluate if the goals of the
business area can be achieved in different ways that can be implemented with
your current level of AI.
You saw this principle in action in the previous chapter. The idea is that

158
 STRATEGY AND AI

AI can create a different kind of value and solution to your problem, and this
means that you might be able to redesign how you achieve your goals. If not,
then AI is not a feasible way, and the conclusion will then be that your use of
AI in this part of your business stops at the feature level.
This principle will show you if it is possible for you to enable an AI-driven
management at your business levels, or not.
Let’s say that you can use AI on a business level in your business. A
consequence is that AI will substitute jobs in a much more significant scale
now.
We saw that in the first chapter of the strategy section. It is a precondition
that any feature, function, or business function that is built with AI only
can achieve free and unlimited use if it is designed to run without human
involvement.
So there is a new dimension to this stage too. It might be possible for you to
achieve exponential growth on the business level. But you have to consider if
you are willing to pay the price of letting your business be run by AI and not
your employees.
It falls back to the question “how much of our company’s overall intelligence
should be artificial”. Should it be at much as possible, or as much as is
appropriate for us?
That is a management dilemma with ethical and social consequences that
you, as a manager, will be required to make decisions on.
Let say that you could and would increase your level of AI to a stage where
business functions are managed by AI. What would then be the next step? The
answer is “The fully automated business”. What this means, is the topic of
the next section.

Level 4: The fully automated business

The answer to the question: “How much of our company’s overall intelligence
should be artificial?” is for the fully automated companies: as much as it takes
to run our company automatically.
Your business goals are now defined so that your output can be achieved in

159
 AI FOR CEOS

ways that do not require people to be involved in the process.

Everything in your business is automated. All that you create is a result of
an automated production process—your products, how you marked them, the
administration and management of your business. Everything is done by AI
for you.
Utopia? Maybe, maybe not. I will let you judge yourself after reading about
how it could be done in practice.
An example of a fully automated business could be: An automatic dry-
cleaning service. The business model could be for customers to hand in their
clothes in a vending machine located in front of a dry-cleaning shop.
Then robots sorted the clothes so that they were assigned the right treatment.
The cleaning also happened automatically, as did the preparation of the
cleaned clothes. And finally, the clothes were sent back to a vending machine
where the customer could pick it up at their convenience.
In addition, a Knowledge System worked to optimize the sales through
online media, so that the physical capacity utilization of the dry cleaning was
as close to 100% as possible. Robots, controlled by Knowledge Systems would
perform all business functions.
The processes and tasks that such a business would consist of could look
like this.

160
 STRATEGY AND AI

What enables full automation is that all the arrows, which are tasks and
processes, can be solved by either AI or IT. By doing so will create an
organization that solves the operation of a relatively complex business set up
namely a dry-cleaning business, fully automatic.
However, not all tasks can be performed automatically. The store needs to
be renovated continuously, when operating errors occur, they need to be fixed.
So, there will be a number of features and tasks that are not automated after
all.
However, the core business model is automated, and thus scalable. Because
once you have one ready-made model for running such a company, then it
probably makes it possible for you to run 100 or more dry cleaner outlets
(automatically) almost as easy as to run just one.
So there’s a potential built-in exponential scalability in the business model
itself, and that’s where the great value of reaching this stage lies.
It is a significant point to keep in mind. To scale the business model
exponentially.
Remember that scalability matches the level that you are at. If you are at
level 1, you can scale your analysis. You can get as many analytics reports as
you want automatically. At level 2 you scale a function. For example, parts of

161
 AI FOR CEOS

your customer service. Here you again create exponential possibilities for the
use of the value you create once automated—this time on a function level.
And at this level 4, it is your business model itself that can be scaled.
Obviously, the value of a scalable business model is much more interesting
than what can be achieved in the initial stages.
As always, your achievement of scalability is a balance between setting your
business goals and the tasks you want your business to solve to achieve its
goals. So, achieving exponential growth on the business level means setting
goals that have the potential of being done automatically.
To understand this, we return to the way you create value, as seen below. In
this context, you must define goals and outputs in a way that can be solved by
AI. For that is the prerequisite for reaching this stage.

In the dry-cleaning example, it means that you could have the choice of having
customer service staff in the dry-cleaning outlets. But, by designing your
customer service around a (good) online experience, you can have a set-up
without staff, thus creating a better ground for the scalability of your business.
Because your customer service business goals will be achieved through your
digital solutions.
The consequence of reaching this level is that you have automated ev-

162
 STRATEGY AND AI

erything that can be automated in your business. So, the basic premise of
your existence is that “we are an automated company, but there are certain
functions that cannot be automated, so we have the employees to perform
such tasks”.
There may be a few new jobs where AI and humans work together to solve
specific tasks, such as product development or other expert tasks. But it is not
a goal in itself to have that type of job. The only reason they exist is that you
can’t use AI alone to solve that type of task.
You probably think this way of running a business is exciting (or scary), but
there is a long way to go. You can probably also think of many good reasons
why such business models would not succeed.
I promised you an example of a real business model that works completely
without human interference. A large part of what Amazon.com does, for
example, actually follows the way the last level works. It’s called the
employeeless company.

Level 5: The employeeless company

The answer to the question “How much of your company’s overall intelligence
is artificial?” is now: Everything. Thus, an employeeless company. A company
where all tasks are performed by AI and where AI makes all decisions.
Surely many will think. It is never going to happen that a business in itself
is run with Artificial Intelligence, completely without employees.
But, can you be sure of that?
The following two factors explain why the companies that were at the
previous level still had employees, that is, not 100% AI-driven:

• The goal of the way their business should be run requires a set-up where
employees were a necessity for a few tasks
• The quality of their AI was not high enough for all the tasks to be solved
by AI, and therefore there were still jobs for employees

But what if one could design a business model where both these conditions

163
 AI FOR CEOS

were addressed?
In a way, where AI actually did run the entire business. Where AI’s ability to
deliver value matched the business goals that were for the business. Then you
would have a self-managing business with no employees.
The commercial aspect of a self-managing business is obvious:
If you can create such a business in Denmark, then you can copy the set-up
to all other countries as well. And maybe you can take the model from one
market and copy it into another market - or create versions with different
products or services.
Before you get introduced to a particular case describing how an employee-
less business could be designed, then the prerequisites for what is required
will be outlined to you.
As you may remember from the section on Artificial Intelligence and the
potential evolution that technology is facing over time, scientists are debating
whether at some point we will reach a stage where AI becomes more intelligent
than humans.
If that happens, then AI will be able to fill all the roles that humans can today
- plus much more, since AI is already superior to us in a number of ways.
So technology alone can bring us to a stage where all companies reach a
level where they become self-managing, and thus employeeless.
As you may remember from the discussion about Super Artificial Intelli-
gence, it can potentially take many years before we reach this technological
stage (and we might not even do so).
Therefore, it is far more likely that business models are adapted to a less
sophisticated AI level will be the driver that gets us to this level. In other
words, business models that are designed to enable AI to handle the entire
value creation of a company.
Here is such an example of how to design a company with the AI technology
that is available today.
Thus, a company whose business goals and requirements for all the products
and services can be produced based solely on the use of AI at its current level
of sophistication.
The business model that I have in mind is a certain type of webshop that can

164
 STRATEGY AND AI

be automated to a state where no employees are needed. There are, of course

some special preconditions that need to apply. This also means that not all
webshops can become employeeless. But, as you will see now, some can.
There is a type of webshops that is called Drop Shops. They are unique in
the way that they sell other companies’ products without even having them in
stock themselves.
For example, if you buy a book on a website that operates as a Drop Shop,
they do not have the book physically in stock when they sell it to you. Their
business model is that if you buy a product from them, they will get another
company that they partner with to send you the book to you directly.
You could say that the webshop acts as a virtual sales agent. Parts of the
sales on Amazon.com are following such a model, so it is not an unknown
way of doing online business. When you buy a product on Amazon.com, it is
often an affiliate of Amazon.com that sells you the product and sends it to you
directly. So as an Amazon customer, you are getting serviced by Amazon, but
the logistics behind is run by a partner shop36 .
The tasks to run a Drop Shop business can be described as follows:

36
The book “DROPSHIPPING: Discover how to create a Standard e-Commerce on Shopify Amazon
and eBay” let you in on the concept in more detail. It is written by Christopher Miller.

165
 AI FOR CEOS

In our example, the Drop Shop thus establishes the contact between their
customer and the business that has the book in stock. The reason why
this model works well is that Drop Shop sites are usually good at sales and
marketing, and may also have a large customer base that they can sell to.
Therefore, they partner up with corporations such as the store they are
selling from (the supplier store) who cannot reach the same customer base as
well as the Drop Shop can. Amazon, as an example, has a large customer base
that stores selling through Amazon network never otherwise would be able to
reach.
A Drop Shops business model is thus 100% digital. They do online marketing
that attracts customers to their website. They sell the products online and
manage the part that ensures that other shops send their goods to customers.
They thus create value by performing a number of digital workflows.
Therefore, in order to become an employeeless Drop Shop, it is required that
AI can control all the workflows that exist in their business. It is also required
that AI can make all decisions that need to be made about their business. Every
AI technology component that is required to do so is already available today.
Let’s start with the marketing part. As you now know, there is already a
rapid development in online marketing and the use of AI. We have already

166
 STRATEGY AND AI

elaborated on this a few times here in the book.

There is already a service available today that can automatically adjust sales
communication and the ads placement to make it work as well as possible.
Among other things, they work by training algorithms on which (personal)
messages work best and which automatically placed banner ads have the most
impact37 . This means that AI can efficiently ensure that you get the right
potential customers to a site.
Drop Shops profits from the difference between what it costs them to do
marketing and the profits they make on their product sales. AI can also learn
the management and selection of which products are most profitable. The
ordering and administration of the products and the publication on their
websites are already automated, so there will be no new tasks for AI in this
part of the flows.
So, what you have now seen is an example of a complete set of business
processes, workflows, and decisions that can all be automated. The machines
learn how to sell the products and which products sell the best. It decides
which products to sell on the site and does all the marketing automatically.
The consequence is an employeeless company.
Of course, it is a very simple business model, and some will argue that it is
not even a real business.
However, that can be a dangerous attitude. Keep in mind before judging such
a model that it can be extremely profitable. If the business model as described
here works, then it scales super well. Once the algorithms have discovered a
profitable way to sell, it will automatically scale out in many equally profitable
directions.
The strength of AI is that it masters both the complex overview and volume.
So, the automatic Drop Shop can easily handle 10,000 daily orders. Or 100,000.
Capacity in this business model is not an issue. If it finds a new market, it can
continue to open similar new ones up. It will do so as long as it can make a
profitable match between customers and products. So, a very simple business
model can become extremely profitable.

37
One of the better solutions on the market is Albert. You can read more on https://albert.ai/

167
 AI FOR CEOS

The point of the example is that it is already possible today to find a level
where business requirements, the complexity to be handled, and AI’s technical
level all match.
You have to think of the Drop Shop case here as a business Proof of Concept.
If it works for that type of business, then we all know that you can build
better solutions that can handle even more complexity, and thus do more than
just sell products online. You must, therefore, see it as the beginning of an
evolutionary process.
The Drop Shop’s evolution process will be the opposite of the one you’re
likely to look into if you’re working in conventional business, and you are,
considering how, and how much, you can use Al in your line of work.
In a typical business, you will reflect on how far up the five levels you would
be able to pull your business (Starting from level 1 or 2). That’s what I’ve been
asking you to consider when you started reading this part of the book. How
far can you get towards level 5?
The drop shop starts at level 5, and their journey is the opposite. They
will work with increased complexity management while maintaining their
employeeless state. So, they work in a direction to compete in more mature
markets with more sophisticated products and solutions, without compro-
mising the benefits their employeeless conditions gives them. They become
more advanced but still employeeless.
Whether it is the adaptation of business requirements or technological
development that is the driving force, it will be the same situation you end
up with. Namely, the company’s Knowledge Systems can run the business
without any human involvement.
AI is the foundation of this type of business. Because they have been designed
to reach an end goal that enables exponential value creation through an
organization without capacity constraints.

168
 STRATEGY AND AI

The stage is seen in the illustration above. The short version is that all
tasks that must be solved in the company are solved by a Knowledge System.
Therefore, there are no jobs or processes.
The situation now is that AI is your business. You probably use IT in
conjunction with AI to send emails, to run your website, communicate with
your partners, etc. But everything is controlled and managed by AI.

Why aren’t you at Level 5?

The Drop Shop example illustrates that it is possible to create a business based
100% on AI.
What you now have to consider is how far you can go in your business
towards level 5. The logical (and easy) approach will be to reflect on where
you are now, and how far you can possibly go. But I would like to challenge
you on this approach.
Instead, the (slightly provocative) question you could ask yourself might be:

Why are we not at level 5?

If you agree with the book’s considerations of AI and the value creation concept,

169
 AI FOR CEOS

then you can choose to see the world like this:

As a manager, you set the goals for your business. It is your choice which
output you want to try to achieve, and it is also your choice which degree of
complexity you want to deal with. It is all business choices that define the way
your business is run.
You can choose a way where the pieces fit together like this:

170
 STRATEGY AND AI

It is a strategic choice that you will produce your value in that way. For many,
this is probably also the only choice (for now).
But, as the Drop Shop example showed, one could choose another business
design. Namely this:

Where value creation happens with by solely using AI, and then there’s
everything in between, which we have just reviewed.

171
 AI FOR CEOS

But - is not a technical choice - that inhibits you from being at level 5.
You are not at level 5 because you have decided on a business design in a
way where you set your goals so that they can only be solved with a set-up
that requires the use of employees, processes, and IT.
You as a manager and the choices on how you run your business are the
reason why you are not at level 5.
I know it is a provocative statement, but I do so because I often experience
that the resistance to change is always rooted in two conditions.
Either in the way you have organized the way you solve tasks today, where
the change of operations is not open for discussion. Or to approach the subject
with an attitude that technology is not good enough to solve our problems and
therefore it should not be used.
In the short run, conflicts and difficult choices are also avoided by choosing
such approaches. So, it is an easy position to take. However, the price may
prove to be expensive.
Because by not having a realistic picture of how your business can change
with technology, you make yourself vulnerable. Your competitors may not
think the same way. This topic is so important for you to understand that we
spend the last third of the strategy section to help you get a grasp on the field.
But before that, a reflection of the organizational implications through the
five journeys. For it is also clear that what has just been described to you
will come with a hefty price that you also need to consider - the faith of your
employees.

Organiza onal adjustments through the five levels

Before we wrap up the section on your various application levels of AI, it

makes sense to stop and look at the overall change across the five levels. It is
presented as a table below.

172
 STRATEGY AND AI

This is how you read the table. The size of the bar in each cell indicates the
effect of the change and the arrow indicates the direction. In this way, pointing
to the right here means increased usage, and to the left means a reduction.
As an example, the only change in the first level is that Knowledge Systems
emerge on a smaller scale.
The overview of the sum of changes is quite interesting.
What you can clearly read from the table is that there is a ketchup effect of
change as you move from the use of a few single-level AI solutions in level
two, to the dominant role that AI gets when you hit the third level.
So, you can expect a process that starts relatively “harmless” if you take
an employment approach. But, when AI reaches the critical mass that allows
it to create business-level change as it does from Level 3 onwards, then the
significant changes will evolve.
So this is also why one should have the time perspective in mind when
talking about the effects of AI and jobs. There are some who would argue that
AI is not a threat to our jobs. The argument is that AI alone will make us better
at what we do. Others will claim that AI removes 40% of all office jobs.

173
 AI FOR CEOS

The point is that both views are probably right. They just looked at two
different time perspectives. The job pessimists - if you can call them that
- look at the effect of levels 3,4 and 5, where the “carefree” look at what
happens when AI just acts as an assistant to us in levels 1 and 2.
The five stages and the model on which it is based are my personal
suggestions of what could happen. So, you can always discuss the validity of
the conclusions.
My position on it is that I am quite convinced the mechanisms around the
dynamics of the effects I describe are right. Thus, the way in which there will
be changes in IT, jobs and processes driven by increased use of AI.
What can be a more open point of discussion is the realism of a rapid
adaptation of AI, Especially the evolution from a few standalone AI systems to
a scene where AI becomes a dominant technology.
So, the transition from level 2 to level 3. The crucial point in this regard is
whether AI can really be an alternative to IT, jobs and processes, and if AI is
truly capable of managing complex and critical elements in our business. And
if we are willing to let AI make important decisions on behalf of our business.
Now, however, are we opening the last section of the strategy part of the
book. Its focus is on a topic that many mangers find quite interesting, namely
competition and AIs ability to change your competitive landscape.
Competitive areas, such as product design and market agility, have already
been discussed in the operations section of the book. But at that point, we did

174
 STRATEGY AND AI

not weigh in how your competitors might act in a market where AI can change
how you compete.
The next section will teach you how to find the right commercial focus when
it comes to AI and how to foresee how your competitors might use AI. Knowing
this can help you make the right set of appropriate commercials actions when
the marked driven AI changes start to emerge in your line of business.
This and much more are some of the areas that we will cover in the next
section.

AI as a compe ve driver

So far, the focus has primarily been on how AI can create value for you in
various ways. But what AI can do for you, the technology can do for others as
well.
It means that you need to take into considerations how AI might change
the way you as a company compete because AI has the potential to change the
premise of your competitive situation.
You might be able to use the technology aggressively to change the premises
to your advantage. However, the opposite could also occur. Others could use
the technology to make them better too. Maybe even in a way that disrupts
your industry. So, there are two competitive angels that you need to keep in
mind:

• Where could the change happen in our industry?

• What effects will the drives behind the changes have?

The change question consists of two elements. First of all, you need to be able
to identify where AI would make changes in your industry. The second part
has to do with the relevance of the potential changes. Not all changes will
have an equal impact on your business, so you need to understand the level of
impact that they could have for you.

175
 AI FOR CEOS

The last element has to do with the driver of change. You need to know if
the underlying force behind the change can change the competitive landscape
on a more fundamental level, or if you can treat it with lesser significance.
So we start by looking at the change that could happen in your industry. To
see if you have areas that are of strategic importance to you can be changed by
AI.

Where could the change happen in our industry?

We will now start looking at the five areas and the type of change that you can
expect from AI. You know them form the operational section of the book:

When you look at the bullets with a competitive mindset, then it is clear that
they all have a commercial angle to it too. Meaning that changes in each of
the area have the potential to impact the way you compete.
If one company in an industry manages to increase its capacity because of

176
 STRATEGY AND AI

the smart use of AI, then it means that they can produce more at the same cost
base. This gives them one kind of competitive advantage. If another company
manages to make new products with new smart features made by AI, them
they will get a different kind of competitive advantage in their market.
So, for you to work with the five areas, you need to ask yourself three
questions:

1. Can AI change the area that we are looking into?

2. Would it give us an advantage if we made an AI-driven change in the
area?
3. Would it be a threat to us if others made an AI-driven change in the area?

We will get back to how you do this very soon. First, you get a reminder of why
you should think this way.
Remember, at the beginning of the strategy section of the book; I mentioned
that in Greenland they have a saying that is it not the Polar bear that you are
watching that will end up eating you?
This is what we are trying to help you avoid now. I am trying to help you see
where the real and commercial substantial AI-driven change in your business
could emerge.
To help you spot the change drives. We start this by returning to how value
is created in your business because all of the above bullets affect and changes
how value could be created in your industry:
• Productivity
• Service
• Complexity Management
• Product development
• Marked Agility
So, the starting point is the where in your business AI can create value:

177
 AI FOR CEOS

The three areas in the goal section make it possible for you to set more
ambitious goals for your business. You can thus create better products,
increase your service, or access the market in better ways.
The productivity part makes you more efficient in how you create value
in business, and finally, the complexity area allows you to solve more
complicated tasks.
So far, we only relate to the effect of what would happen if you got better
at the points from your point of view. But that’s usually just one part of the
equation. Your competitors can do the same with greater or lesser strength.
There are 1,000 examples of how technology has changed the terms on which
industries compete. And you may have also considered whether AI is going
to be a factor that will change the rules of the game for how your competitive
situation can change.
The big questions are: How do you know what’s happening in your industry?
And how do you think your competitive actions would be?
Should your approach be aggressive and challenge the existing way of
conducting business? Or should your strategy be more reactive, letting others
take the hard learnings first?
Changes in each of the five areas can have different effects on your competi-
tive situation.

178
 STRATEGY AND AI

An example explains how. Let’s say you are a Pharma company that makes
cancer drugs. Your critical success factor is probably closely related to your
ability to create products that are better than the ones on the market today.
So, product development is essential for you. For you, inventing new
products is probably harder than producing them. So, your commercial success
is tied to your ability to create new products.
Pharma 1: example: Product development is central

But other Pharma companies will rank their priorities differently. If you sell
generic products such as headache pills, then your sales, marketing, and
production costs would be the most critical competitive parameters.
In this case, it is essential that customers know your brand, your products
must be available in the channels where customers buy over-the-counter
drugs, and your products must be priced at an attractive level. So here,
productivity and perhaps market access would be the key areas for you. And
product development is perhaps unimportant.
Thus, the complete opposite prioritization of how your colleagues from the
cancer manufacturer company would prioritize their five areas.
Pharma 2: Example: Productivity is the key area

179
 AI FOR CEOS

So, the first thing to do is to judge the areas based on their importance and
change potential. The goal is to ensure that you are focusing on the areas that
are of the most significant value to your business.
Remember the three questions that I told you that we need to get answers
to:

1. Change potential: Can AI change the area that we are looking into?
2. Advantage: Would it give us an advantage if we made an AI-driven
change in the area?
3. Threat: Would it be a threat to us if others made an AI-driven change in
the area?

This is what we are doing right now.

The way you should approach it is this way. First, you evaluate if the area
has any change potential and if AI can make an impact. If that is not the case,
then it does not make sense for you to reflect more on the matter.
If you think that AI can make a commercial difference, then you need to
answer two questions, first, if you can use the technology to create a market
advantage. Secondly, to judge the impact on your business if others did the
same.

180
 STRATEGY AND AI

Now we turn our focus to what you can expect to the change in the five
different areas from a competitive perspective.

Produc vity

The overall idea with AI-based productivity-based is for you to achieve one of
two things:

• Increase your capacity

• Become more efficient

From a competitive perspective, this means that you will strive to create
a competitive advantage by achieving more efficient production than your
competitors. The focus is on economics, rationalization and efficiency
improvements, all measures that ensure that products can be produced
cheaper and more rationally.
There will be two different types of change that will be driven by productivity.
Either from existing players - your competitors - who can make better use of AI
technologies, and thus can create their products and services more efficiently.
The second type of disruption is driven by new players who are not con-
strained by a historical way of organizing work and who may, therefore, be
able to take full advantage of AI technology’s capabilities.
If you remember the drop-shop case, it is just one example of a company
that uses the technology to create a product (webshop) in a way that is radically
cheaper to operate than normal webshops because it can be operated without
employees.
From the perspective of an existing organization, there is a tendency to
be less open for new radical ways to achieve technology-driven efficiency
improvements. Culture, management focus, attention to the daily operations,
limited budgets for new development can be some of the constraints that have
to be overcome to implement technologically driven change.
The consequence of such considerations may be that you do not become
aware of how new ways of solving things can be made more efficiently. So if

181
 AI FOR CEOS

productivity is a significant competitive parameter, then as an existing orga-

nization, you run the risk of not paying enough attention to AI’s opportunities
to increase efficiency.

Market

Now, the goal is to know more about your customers. You achieve this through
a better understanding of their needs, and to offer solutions that are a better
fit for them. This is possible because AI can help you understand what drives
your sales by identifying the best customer, product, and market condition
match. The result of your AI-driven market focus is that you could:

· Create better digital products

· More efficiently sell your digital product online
AI-driven market change can happen if the current product offers are not in
line with the customers’ needs.

An example based on Kitesurfers behavior can explain how this could happen.

Let’s say that there are patterns in Kitesurfers’ behavior that tell us that they
have a particular need that is not met by the insurance companies that exist
today. There is no specific insurance product for Kitesurfers gear when they
travel by air.

However, the target audience is big enough to be profitable, and the customers’
price sensitivity means that they would probably be willing to pay premium
price such a new product.

And what are their needs then? It’s that they get their equipment destroyed
in planes when they travel to new Kitesurfing locations. And who knows
this information? The airlines, as they receive the claims for the damaged
equipment.

182
 STRATEGY AND AI

So instead of just dealing with the complaints, our airline in this fictive
example is now also making a new insurance product.

They know that kite surfers are traveling to Spain in high numbers and that a
proportion of them will end up damaging their equipment during their travel.

But it is not the insurance companies’ AI that can see this pattern. It’s the
airlines. They see a new market for which you can make a unique product: A
new Kitesurfing insurance product.

What is characteristic of this market is that it is well-identified: Kitesufters

via real market data and that it is what one can call a Micro market. Thus, a
small well-defined, digital, and probably profitable market.

The airline’s Kitesurfing product may be sold in Norway and Sweden, Germany
and the US: Or it can be adjusted so that the same mechanisms apply to tourists
who bring their bikes on a plane too. Or, the airline can make additional sales
to the Kitesurfers, now that they know more about this segment preferences.

By understanding their customer base well, the airlines can use their customer
base to sell a new product with a customer, market, and product fit that would
normally be offered by an insurance company.

So the dynamic here is very different from the previous area of productivity.
Now you should only have one focus, and that is to look for new areas where
you can create AI-driven advantages.

Product

It is few companies that produce a physical product where the quality and
possible features of the product do not have strategic commercial significance.
The consequence of AI-driven product development is that:

183
 AI FOR CEOS

• AI makes it cheaper to develop new products.

• AI gives you an innovative advantage.

So, if a product’s physical characteristics are important, and if AI has the

potential to create better features on your product, then this is a likely area
that can shift market share.
AI-assisted product development can change the premise of how to create
new products.
The design may be different; you may be able to use new materials; you can
quickly design prototypes that can be tested by real customers in other words.
There are a lot of elements in the creative part of producing products that are
subject to restrictions other than they are today.
All such element has a potential commercial effect on how you run your
business.
Let’s say you sell shoes. The process is that you design them yourself, and
get an Asian supplier to produce them for you. You might ask how can AI be
used to do it differently, and better?
Let’s continue with the example. AI can be used to change the features that
your product has, and maybe also the way your products are produced. This
means that AI in a shoe design process gives you fewer creative constraints
than you would otherwise have. For example:

• Design a sole with minimal weight and maximum shock absorption (so
you can make lighter shoes)
• Use new materials, qualities and colors that give you new design possibili-
ties
• Suggest designs based on input from you. For example: Design a leather
sandal with laces and wooden heel. Your AI assistant we then come up with
suggestions for products based on the requirements and prerequisites you
give it.

In addition, in your product design, you will also be able to use AI to help make
your production process more efficient. For example, it could be by telling you

184
 STRATEGY AND AI

how to cut the leather so that you get the least waste (lower production costs).
Or, by telling you how many more boots you can have in a standard container
if you change the design of the boots.
The commercial risk you run by not using AI is that your competitors may
discover new product features that make their product far more attractive. Or
that your competitors change the premise of how products in your marked
should be perceived.
This means that they might use AI to solves your customer’s problem with
a completely different solution than what you offer, which may be better for
customers, or more rational for your competitor.

Service

Remember that AI drives service changes has two consequences:

• Replace customer-oriented processes

• Offer automated and personalized customer advice

Services as an AI competition parameter are most interesting in industries like

telecommunications, insurance, finance. That means marketplaces where
companies with many customers are offering a relatively similar product to
their customers. In such cases, customer relationships typically mean a lot
when consumers do not know which product to buy. Therefore, AI-based
service can provide an edge over competing companies.
It goes without saying that the companies that manage to exploit the
opportunities for as close customer contact as 1-1 digital dialogue will be able
to offer a great customer benefit in many areas. They can connect customers
closer to their brand; they can provide a service in all areas of their business
where there is customer contact at a level that is not possible today.
It would also be reasonable to assume that AI-based service that has the
potential to influence companies’ brand, sales potential, and quality of support
will be able to move sales from competitors who do not match such initiatives.
In the mobile service industry, it is often seen that new providers buy into

185
 AI FOR CEOS

an existing cellular network and build a business by being keen on serving

a specific target group well. The claims can be low priced, attractive online
services, or extraordinary offers. Such business models could use AI-based
service as a new angle to enter new markets.

Complexity

The analysis part is probably not where you will find that there will be
commercial battles. Remember that the value of complexity is to enable you
to make more qualified decisions based on complex information.
The analysis aspect means that you will make more professional decisions,
but from there and to obtain a commercial advantage is a bit far-fetched.
However, that’s not the case with the other part of handling complexity,
namely, Product development. Not as it is done as a prototyping tool, but
where the complex analytics part is the core of the product that you create.
The Stanford example of cancer screening is an example of better manage-
ment of complexity. Here, AI can identify 12 different parameters that can
indicate cancer. Contrary to the standard 9 parameters in the medical science
approach. If Stanford were a traditional commercial enterprise, this handling
of complexity in the use of AI would give them a great competitive advantage.
They would be able to make a better product and perhaps faster than
competitors to find potential cancer cases.
However, trying to use complexity as a competitive parameter falls into the
High risk, High return category. Stanford didn’t know in advance that their
AI would end up being better than how humans work. They succeeded, and it
gave them a potentially great competitive advantage.
The Stanford example also shows that competitors who are creating new
products with data could be so radical that it could be categorized as disruption.
So, you can risk AI-driven disruption in your business from both new and
know competitors.

186
 STRATEGY AND AI

Next steps

We have now reached the end of the strategy section of this book. I hope it
gave you some new perspectives.
As you remember, the goal of this part was to prepare you to manage your
use of AI on three different strategic parameters.

• The opportunity to create exponential value with AI

• Uncover how many tasks AI has the potential to solve for you
• Help you see how AI can influence the dynamics of your competitiveness

Hopefully, this section has given you some reflections on how you could use
the technology from a strategic perspective.
Now you should be able to know the extent to which AI should be included
in your strategic considerations.
You might now be a point where you know that AI must be a force that you
must include in your business on a larger scale. To do with success, you need
to know of one more element, and that is:

AI and Leadership

For it is your ability to execute the management of AI in your business that

ensures that you achieve the goals that you are aiming at are met.
The challenge is that the premises for how you work with AI are different
than anything you have previously worked with. Some of the reasons are:

• Creating an AI solution is not the same as making an IT solution.

• The projects needed to implement AI solutions are different
• The way your people work with AI is different than any form of collabora-
tion they know

187
 AI FOR CEOS

So, the way to reach your AI goals requires a different kind of leadership than
what you have practiced until now.
The reason is the same as what allows AI to create a new type of value.
Namely, that AI redefines the way you work.
You know now that AI is something very new and something very different,
that AI is not IT. AI is not a job or process. AI is a new component of your
business that redefines the dynamics of the way you create value. Therefore,
you must deal with a new management aspect too. This is the topic of the next
section of the book: The leadership and management of an organization that
uses AI.

188
 5

Leadership and AI

AI changes the basic premise of how you perform

leadership

T
he three previous parts of the book have shown us that when you
introduce AI into a business, then it becomes an organization that
must sole new kind of tasks. The distribution ratio of tasks between
man and machine changes, you may also be organized differently now. You
probably also manage your business with different goals.
So there are changes in many dimensions that all require appropriate and
different leadership and management attention to succeed.
You now know that you can do other things with AI than you can with IT;
the way you create new AI solutions is completely different from the way you
would have to build IT software. It is also clear to you that AI changes the
premise of your organization’s ability to handle complexity, and that means
you can do something different with AI than you can with IT.
Working with AI also changes a fundamental relationship in the way we as
humans solve problems. This means that way we will interact with technology
are on new and different terms.
Among other things, this is why AI changes the premise of how management
and leadership should be performed. Both at a project level and in day-to-day

189
 AI FOR CEOS

operations.
We now dig into the consequences of what you as a leader would have to
deal with. So, you can start your preparation on how to lead and manage a
business that has a higher degree of AI use.
To help you do so, you will be introduced to five different dimensions of AI
and leadership.
The five areas are included in this section of the book because they represent
essential and new management perspectives. Each of them also represents
cases where the management requirements differ from how a manager
traditionally would operate.

• AI and leadership
• How should AI be managed?
• AI is not IT
• How do we work with AI?
• How to develop a Knowledge System?

The common denominators for all five parts are people, technology, and
management. But the angles and perspectives are very different. Some of
them aim at giving you a new perspective on the premise that you now must
perform leadership under. Others have a more practical focus on what it takes
to run, implement and execute AI projects in your organization.
So, the aim is that when you have read this last part of the book, then you
would know how to outline the roadmap it would require to start working with
AI in your business. Including how to how to design and scope your projects
on the new terms that AI requires.
At the same time, you would be able to see and reflect on how the increased
use of AI in your business would transform the premise on what it takes to lead
and manage a business that relies more and more on the use of AI to succeed.

190
 LEADERSHIP AND AI

AI and leadership

I think it should be clear to you now, that AI is a technology, but the successful
utilization of AI does not relate to AI being a technology.
Success with AI is about management, leadership, and achieving the
business value that AI can create for you.
We will get to that part soon, but before we get there, you need to understand
a fundamental factor of what it takes as humans to work with AI. It will require
that we adopt a totally new mindset.
Why this is so, and how you can perform management under such conditions
is the topic of this section. You are about to learn the answer to the question:
How can we, as humans work with Artificial Intelligence?
The reason why you need to know this is that the way you create value with
Artificial Intelligence is significantly different from the way you work with
conventional IT tools.
The reason starts bit abstract and can be traced back to the very first tools
we as humans invented. In general, you can say:
A tool makes us better at our job - on our terms.
A bow makes us better hunters because it is designed to kill animals that
we might not otherwise be able to catch. So, the bow makes us better at our
work. A bow is designed to maximize its killing ability based on a grown-up
man’s strength, and his ability to tighten the bow and shoot with it. So, the
tool works on our terms.
You can use the same argument for all other tools we have invented ever
since (Cars, the computer, boats, pens, etc.)
AI will be the first tool to change that basic premise.
AI will make us better at our job. But not necessarily on our terms.
The last part is very important for the way we work with AI. Therefore, much
of this section is about how to deal with the premise of working with AI.
The first step in unfolding the consequences of this relationship is to look at
how AI is different from IT. So, we start with how AI is different from IT and
why it matters.
In this book, I choose to name the two approaches to technology as either a

191
 AI FOR CEOS

Knowledge Systems or a rule-based systems. AI is a Knowledge System, and

IT is the rule-based approach.
You might think that is just an academic discussion, and that the difference
is not such a big deal. Are the two approaches not just two versions of the same
overall technology, one might be tempted to think?
No. The reason why you should be interested in the difference between the
two is not technical. It has to do with how we, as humans work with our tools.
IT solutions, as you know them today, are all rules-based, which in short
means that the way they work is determined by a series of rules that are coded
into the system. The problems that the software is designed to solve are all
based on these rules.
This is not the way a knowledge-based system work. A Knowledge System
performs automated actions that are based on decisions made by Artificial
Intelligence.
Knowledge Systems have gained knowledge by learning how patterns in
data can explain a given behavior. Based on experience, the systems will be
able to propose the best solutions, thus solving the task for which they were
designed. So, they are based on knowledge, not rules.
A non-technical person might think. So, what if a piece of software is
designed to it looks for patterns and is not programmed the way you usually
do. To me, that seems like a feature that is only of interest to an IT architect.
This level of system design detail cannot have much significance for me.
Nothing could be more wrong. The consequences of the transition from
rule-based to knowledge-based software solutions will be the greatest social
change that we will all experience in our lifetime.
The reason is that the tasks that you can ask a knowledge-based software to
solve are much more complex, and intellectually more demanding than the
functions that rule-based software can be designed to solve.
We can create much better software that can help us with areas that you
would not be able to solve today, with the way software typically is made.
Knowledge-based software will be the foundation of what is known as the
Fourth Industrial Revolution, that we are steadily progressing in to now.

192
 LEADERSHIP AND AI

According to Forbes38 , the new form of society will make our planet ten times
richer than we are today. One of the three drivers of change in the revolution
is AI. The other two are new financial ecosystems and new forms of energy.
So, AI is a key component that will drive much of our global process to enrich
our planet. This is why AI vs IT is not a technical discussion of the best solution
on how to design software. What we look into is the foundation for a whole
new global form of progress.
As exciting as this talk might be, we will not discuss more on the matter now.
If you are more interested in the subject, then I have written articles on the
topic on the cognifirm.com blog. Now we go back to the difference between AI
and IT.
An example from the real world explains what the difference between the
two approaches means. There is a lot of focus on self-driving cars right now,
so that’s an excellent example to start with.
There are many companies working on self-driving cars, and the problems
they have illustrate the difference between knowledge-based and rule-based
IT solutions quite well.
You could choose to build the software needed to make self-driving cars as
rule-based software. That is software that describes all the possible outcomes
that a car could face. The software then describes how the car should react in
any given situation. When all possible outcomes were described, you had the
software needed to make the car run. Very roughly speaking.
They tried such an approach at Uber when they started making self-driving
cars, but they gave up39 . The reason they gave up was that the method would
require them to take into account all the exceptions. For example, what would
the car do if someone dropped a grand piano on the road? They had to make a
rule for all such cases, then coded it and then test them to see if they worked.
The consequence of working with a rule-based approach would be that Uber

38
https://www.forbes.com/sites/goncalodevasconcelos/2015/03/04/the-third-industrial-
revolution-internet-energy-and-a-new-financial-system/#6ff9da41271a
39
https://www.ted.com/talks/kevin_kelly_how_ai_can_bring_on_a_second_indus-
trial_revolution?referrer=playlist-what_happens_when_the_robots_take_our_jobs

193
 AI FOR CEOS

would have to uncover and describe all possible scenarios that an autonomous
car could face.
To put such a problem into perspective, you should know that a smartphone
has 12 million lines of code in its operating system today40 . Think about how
much code it would take to describe everything a car would need to be able to
do by and all the precise actions it should be able to handle.
That’s why Uber scrapped the rule-based approach to software development,
focusing instead on the kind of AI called Deep Learning. You already know
how that type of AI works from the first part of this book.
Working with Deep Learning means that the car’s software gets some goals,
such as that it must be able to run across a traffic light correctly. They try to
do that, maybe 10,000 times in a real car at real right traffic lights. From this
experience, the software then learns what the right behavior would be so that
it would know how to maneuver a car safely across the traffic lights, in a way
that also takes into account the surroundings and co-drivers, etc.
So Uber does not explicitly teach their AI what to do if a grand piano suddenly
appears on the road. But, it has learned which reaction is most appropriate if
there is suddenly a foreign object in front of the car. The software decides to
either drive around it or to slow down the vehicle.
And it will be the same way a real person as a driver in a car would make
decisions. In some ways, the computer would probably be able to make a
better decision. It would know if you could actually slow down, or what the
probability was of running into an oncoming car if you chose to run around
the obstacle. Because the software is faster and more rational and therefore
would be able to select the best solution in a given situation.
One million people are killed in traffic accidents every year on our planet.
It is generally assumed that human errors and misjudgments in the traffic

40
https://www.visualcapitalist.com/millions-lines-of-code/

194
 LEADERSHIP AND AI

account for 90% of the courses that lead to fatal accidents41 . So if software
could solve this problem, then it would have created a significant global impact
on that parameter alone.
The example illustrates the value one can gain with knowledge-based
software has a potential that rule-based software would not be able to match.
If you set out to make self-driving car software using rule-based software
alone, it would be a huge task to get the software coded, and it would certainly
not accommodate all exceptions. Therefore, it would probably be perceived as
faulty (and perhaps in this case dangerous too).
In other words, there is an upper limit to how good software solutions we
can make with rule-based software. The fact that we are now seeing AI pop
up in large and small software solutions is an indication that there are many
small and large needs that can suddenly be better addressed with this new
technology.
A logical reflection on such a development would be: If AI is so smart, why
don’t we use it for everything? And why don’t we do so as soon as possible so
that we can collect the rewards with this obviously better new technology?
The reasons are that they are built in a number of pitfalls. Pitfalls that
prevent us from exploiting this technology and its potential.
And the pitfalls are not, as many might imagine, that we lack the right people.
If we just had access to 10 Data scientists more, then we would almost be ready
to scale up our AI efforts, you might think.
To some extent, that is true. At some point in the process, you have to use
the right people and the right team to solve your AI tasks. However, as a leader,
you have to start somewhere else.
Why that is so, and how you do so is the topic of the rest of this chapter.
It is the way we manage people, and how we see the world that is our most
significant obstacle to effectively exploiting the new opportunities that AI

41
The US National Highway Traffic Safety Administration estimates that 92% to 96%
of all traffic accidents was caused by human errors. The data can be found here.
https://www.digitaltrends.com/cars/2016-nhtsa-fatality-report/. The global number of
leather traffic accidents was estimated to be 1.3 million, according to the WHO. Read more
here: https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries

195
 AI FOR CEOS

gives us.
As a leader, when you work with knowledge-based software, then three
things must be made clear to you:

1. Knowledge-based software must be developed in a completely different

way from the way we make software today.
2. The business value that can be derived from knowledge-based software
offers completely new possibilities
3. The way we organize people who work with knowledge-based software
requires a new mindset of all involved.

The journey towards all three areas requires that we, as human beings, have
to make significant adjustments. We would have to think and work differently.
We will have to work against our intuition and to work in ways that contradict
how we previously have been successful in creating value.
It may sound a bit philosophical and high-flying, but our human self-
perception and the way we have worked with IT so far will be the biggest
obstacle to realizing the true value of AI as a technology.
It all points back to the point from the beginning of this chapter that AI is a
tool that makes us better at our work - but not necessarily on our terms.
You can only truly create value with AI if you are aware of the fact that the
methods and our work processes must be different. And most central is how
we as human beings have until now designed our reality.
When you, like Uber, start working with AI, you come up with some locked-
in mechanisms in our behavior as humans. If you look at the task of making
self-driving cars as a project, then the easiest would be to do as you usually
do. Make a requirements specification, develop the software (the rule-based),
and finally launch it as a product. But, as the example shows, they would never
have succeeded with such an approach.
To understand what it really takes to be able to take advantage of AI, we
have to go back to our collective understanding of the world we live in.
For our worldview is the most significant limitation in the unfolding of the
value this new technology can help us create. Successful AI work requires

196
 LEADERSHIP AND AI

that we do things that conflict with our perceptions of how we as humans act
rationally. Why that is so, and what you can do to overcome it, is the topic of
the next chapter.

How should AI be managed?

As humans, and as a civilization, we are a product of our culture and heritage.

It may sound too intellectual, and you may wonder what does this have to do
with your business exploits of AI?
It actually has a lot to do with it. For you to succeed with AI, you have
to do things differently than what the last well over 5000 years of human
development have thought have taught us. It will require us to prepare for a
change in how we work.
AI, as it was described before, is a tool that makes us better at our work -
but not necessarily on our terms. This means that things have to be done in a
different way when you work with AI.
It is about control and humans as the center of our worldview. We are going
to challenge that premise if we are to be successful with AI in our companies.
The change process that we would have to go through when we start to work
with AI in our business will require adaptions in multiple levels of how we work.
The core reason is that the prerequisites for being able to work with AI conflicts
with the conceptual view that we have of how we work with technology.
To understand why we must go back to the two Greek philosophers, Aristotle
and Porphyry.
Porphyry introduced the idea of classifying all human knowledge into
categories42 . This mindset evolved into what became known as the Porphyry
tree. The tree explains how any given object can be categories as part of a
bigger, more universal understanding of the world.
The thinking of a tree as a way to categorize knowledge is what now is known
as a relationship hierarchy.
This means that you have some key components and a hierarchy of subcom-

42
Porphyry Introduction (Clarendon Later Ancient Philosophers), by Dan Pollack

197
 AI FOR CEOS

ponents. As you can see from the illustration below, the tree shows how the
relationship between several objects can be seen or perceived as a part of a
hierarchical structure. The objects are hierarchically divided with levels and
sub-levels that define the topic described in groups and subgroups.
In the example below, Plato as a human being is used as an example. But it
could as well have been used to describe fruit, fish or car types.

198
 LEADERSHIP AND AI

The point is that the way of seeing information as components in a hierarchical

199
 AI FOR CEOS

relationship has been a fundamental concept in how all branches of natural

sciences have evolved ever since.
This means that such a way of looking at the world has become a central
element of what one might call the rational human mindset. We are raised
to understand a complex context by breaking it down into components that
we can better handle. It is a natural way for us as human beings to act when
dealing with complexity.
As humans, we find structures and hierarchies that can describe the
complexity we experience in our lives. We gain a perceived control of a
situation if we can describe what we encounter as a hierarchical structure.
Take as an example, trees. If we look in a forest, we do not conceptual have
to deal with every single tree. We don’t need to know what kind of tree it is.
We mentally categories a large group of trees in a higher category, namely a
forest.
Such a hierarchy approach means that we do not have to spend so much
energy to relate to our surroundings. We understand our surroundings by
placing them in relation to each other. Our ability to do so is one of the core
intellectual dimensions that distinguish humans from the second smartest
species on earth, chimpanzees43 .
So, relating to our surroundings as a hierarchy is an ultra-central part of
being human. And AI challenges us on precisely those parameters. AI does
not see the world hierarchically. It is AI’s greatest technical strength, but also
where technology challenges us most conceptually. This factor can be the root
cause of many issues if you are not aware of how to address it.
Therefore, when leading AI projects or teams working with AI, this dimen-
sion should be kept in mind. It can be the basis of many of the possible conflicts
that you will experience in the journey towards realizing the value that AI has
the potential to create.

43
The scientific term is called Cognitive mapping, and this article from Nature explains
why humans are different form chimpanzees on that parameter. https://www.na-
ture.com/scitable/knowledge/library/primate-cognition-59751723/

200
 LEADERSHIP AND AI

Profuse and rule-based so ware

Now we return to software and management. We now know that we see the
world as a hierarchy. It is the way we understand the world, and that mindset
characterizes the way we act in many contexts. Including rule-based software
development and the way, we work with it.
We develop rules-based software, by decompressing the problems we want
our software to solve, by splitting our problems into a hierarchical structure.
Just like we do with trees and the forest, as shown in the previous example.
We describe the problem that our software has to solve, e.g. accounting in
a hierarchy. So, this will be the example that we now use: How to design
accounting software.
You would start by defining the components of the field of accounting.
Setting the scene means defining the meaning of the areas of core components.
Such as: What is an invoice? A debtor? an account? And how are the
relationships between them?
Thus, the entire subject area “accounting” described as it is possible in the
IT language that we use for is the task. Eventually, in the process, it will be
possible to formulate the problem that the software must be able to solve, as a
linear mathematical component. Such as this one:
If a new bill enters our accounting system and if we receive it from this type
of supplier, then the payment must be deducted from this account.
The more rules like this one we make, the more features our software will
end up having.
The process of designing software is to find ways where a problem can be
described as rules. If we can describe a problem as a rule, then we can make
software that can solve the problem. Of course, this is a somewhat rough
generalization, but it could be said that this is the way all software has been
designed so far.
One of the problems with such an approach is the code we end up with will
building will never be better than the human intellect that designed it.
If the mindset behind the knowledge tree that has defined our software has
been wrong or inadequate, then the software we develop to mirror it will be

201
 AI FOR CEOS

wrong too.
Our software, the way we build it, and the features we build into it are all a
result of the way we see the world. Our hierarchy, which we have chosen to
use to break down complexity, defines how our software is designed.
In this context, it makes sense here to refer to Uber’s self-driving cars once
more. Conceptually, their problem was precise that they couldn’t build a
version of the Porphyry tree that fits into the problem they were trying to
solve.
When they initially started the project and used IT to do so, they tried- and
failed - to map out all the outcomes that the self-driving car could come across.
The complexity of the world in which self-driving cars is a part of is way too
high. It does not make sense to describe such a world as a hierarchy.
The reason why Uber came up short was that the problem they wanted their
software to solve was impossible to describe hierarchically, with a structure
and with a fixed number of outcomes scenarios. The problem that software
should solve - running a car, simply makes no sense to try to describe it as
components of a hierarchy.
And this is where the real dividing line between knowledge-based and rule-
based software lies.
Can we - and does it make sense - describe the problems that the software
has to solve as a set of rules in a hierarchical structure? Or are there so many
aspects and nuances (complexity) of what we want our software to do for us,
that rules don’t make sense?
We might also encounter another derived problem from this line of thinking.
Because maybe there is a structure and rules for how to solve a given problem.
But we as humans are not sharp enough in our intellectual capacity to see what
they are.
Thus, in some situations, the problem is not that one cannot describe the
situation, but that the human intellect is simply the greatest constraint on
how far we come with this approach. Because we cannot see through very
complex contexts and describe the rules and hierarchy by which the rule-based
software must be designed.
That’s exactly what we saw in the Stanford skin cancer example.

202
 LEADERSHIP AND AI

As you remember at Stanford, they had trained an AI model to identify

dangerous birthmarks that could develop into skin cancer. Their AI ended up
having 12 different criteria that it rated the birthmarks after. The dermatol-
ogists had 9. The machine has thus found three new factors which were an
indication of skin cancer, which the best dermatologists were not aware of
having a significance.
I know nothing about dermatology. But I think the nine criteria that they
work with have been found through a long process where specialists around
the world have built up a mutual understanding of skin cancer and the signals
that can help identify it.
The knowledge and processes on which they are based probably go back
20-30 years, and follow a fairly established process that all the world’s
dermatologists follow. This would be how such domains usually would be
handled in medical science.
.
But the AI models aren’t limited - or hindered, some would say - by our way
of defining how to identify skin cancer, so that’s why they look at the world
differently. They see different patterns and interpret them differently.
It is precisely in such cases that AI becomes a super-powerful technological
tool that can help us with insights we would otherwise not be able to get. But
the way to get there might require that we can set ourselves free from our
usual way of dealing with complexity. And that part has everything to do with
leadership, and nothing to do with technology.

AI or IT? Know how to make the right choice

You are now going to learn why AI is not IT and why this is great if you need to
develop software that can help us in new and radical ways.
We start with the very premise of why anyone will make software, and that is
to solve a problem. You have been presented with this point a couple of times
now. We exploit in more detail why the difference is so vital to understand.
We start to do so with some examples of software that solves problems;
Let´s say that you need to be able to control how packages are distributed at

203
 AI FOR CEOS

a warehouse. So, you make a piece of software that helps solve this problem.
You want to be able to send electronic mail, so you make an email client to
help you do so.
The illustration shows examples of problems that (maybe) can be solved
using software: “Buy the best shares for me”, “pay my bills” “help my
customers”.

The picture shows an exponentially rising curve in a two-axis coordinate

system. One of the axis describes the extent to which the problem you are
looking at is complex or not. “Pack my goods” is a lesser complex problem to
solve than “Drive my car”.
The second axis describes the extent to which it is possible to describe the
problem as rules. As we recall, the basic premise of coding software is that
rules can describe the problem that you are trying to solve. Otherwise, the
problem cannot be solved by conventional software.
It is the management of complexity that is the limiting factor for software.
The more complex the problem, the harder it is to decode complexity in a way
that it can be expressed as rules.
Let’s say that you want to create software that helps your customers get
started with your product; you may need the software you develop to be able
to:
• Uncover the real needs of your customers

204
 LEADERSHIP AND AI

• Understand your product

• Provide qualified feedback to customers.
Such requirements are hard to describe with rules. This is because we are
now in a similar situation to Uber when they tried to make self-driving cars
using conventional IT software. The scope of possible events is too great, and
there is too much complexity built into our requirements. A requirement such
as “Helping customers” can cover a lot of outcomes, meanings, and individual
interpretations.
That is why the shape of the curve is exponential, and not linear. Because at
some point you will hit the wall for what is practically possible to describe with
rules. And once you’ve hit the wall, problems that are twice or three times as
complex can’t be solved with rule-based software either.
The challenge of Knowledge Systems is different. A Knowledge System
works according to goals, not rules. Ask a Knowledge System to find the best
way to Mars, find the best translation, drive your car home from work. You
can (almost) always describe a problem as a goal that you want to have solved.
Therefore, the limitation of a Knowledge System is not the handling of
complexity. The limitation lies in the quality of the Knowledge System. The
difference between the two approaches is illustrated in the figure below.

For software, the process is this:

1. Can we describe what we would like to achieve using rules?

205
 AI FOR CEOS

2. If, yes, can we describe a system that contains the rules that define how
our problem can be solved?
3. Build the software in a way that solves our problem
It is the first point, which potentially gives us problems if we want to make
very advanced software that solves a complex problem.

Because the task of making the rules may be extensive, or perhaps impossible.
On the other hand, points 2 and 3 are relatively simple to manage if you have
succeeded in solving the requirements in the first bullet. If a case can be
described as rules, yes then you can build software that can solve the problem.

The illustration above illustrates this principle. We have a problem. This is

broken into rules and based on these; the software is created which solves our
problem. The critical point for rule-based software is whether we can break a
problem into rules. If that is not possible, then it will fall short.
The process of knowledge systems is different. Here are also three steps.

206
 LEADERSHIP AND AI

The problem is formulated as a goal. These goals are broken down into sub-
goals in a knowledge system. In the breakdown, you define different goals for
the AI algorithms that may need to work together (such as one part of or AI
must understand Japanese, another one must be able to speak Japanese, etc.),
and eventually, the system is trained to solve the required task. So, there are
these three steps:

1. Define a goal
2. Define a number of sub-goals
3. Train the knowledge system
This principle is seen in this figure.

Here you can see the three flows. It will (almost) always be possible to
formulate a problem as a goal to be solved. The problem with knowledge
systems is their ability to solve the goals you wish to solve.
So here the issue is the quality of your AI (and data). It is easy to formulate a
problem as a goal, “Fly a rocket to Mars”, but if the Knowledge Systems that

207
 AI FOR CEOS

you wish to solve the goal for you do not have the required level of quality to
reach the goal, then this is where you get stuck.

Absolute vs. Rela ve constraints

The reason for the difference in the potential lies in the factors that are holding
them back. Whether it is absolute or relative, it means everything.
The difference between the two approaches is that there is an absolute
limitation on what tasks rules-based software can solve. If the problem cannot
be formulated as rules, then it cannot be solved. No matter what you do, rule-
based software never gets better than what rules conceptually restrict them
from being able to do.
So, no one will ever be able to make rule-based software for a self-driving
car, simply because the complexity of the problem you want to solve is too
extensive for it to ever be described with rules. Thus, the restriction becomes
absolute. It cannot be challenged because one will encounter a fundamental
contradiction between a problem of high complexity and the ability to describe
the problem with rules.
The constraint in Knowledge Systems, on the other hand, is relative. As
we recall, the limitation of a Knowledge System lies in its quality. We can
formulate extreme goals, but if the Knowledge Systems cannot solve them,
then we will not be able to get a solution to our problem.
However, this limitation is relative. For the quality of a Knowledge System
can be improved - it lies in the basic premise of it is defined. It is a learning
system that potentially reaches increasingly higher levels of intelligence. The
system can also become better by linking multiple algorithms together, or
by giving it more data to work with. So, the level of task complexity that a
Knowledge System would be able to solve is continuously moving upwards.
This is precisely the thinking that is reflected in the future development
of AI that was presented to you in the first part of the book. Remember the
illustration of the three stages?

208
 LEADERSHIP AND AI

It is precisely because of the illustration above that it makes sense to state that
AI-driven technology has a growth potential that surpasses what is possible
with the technologies we know and use today. The sum of the collective
learnings of all the Knowledge Systems components is what fuels such trends
that move the technology from Narrow Artificial Intelligence towards General
Artificial Intelligence.
However, now our focus is on the management of AI and IT. We set out to
explore why you should tread AI differently from IT. And this is why. It works
with different premises and has the potential to solve many more valuable
problems. What’s equally interesting is how to manage companies that work
with Knowledge Systems.
For it goes without saying that if the premise of a Knowledge System is
fundamentally different from that of an IT system, so will the management
of such a system be. We have touched on the subject a couple of times in this
chapter, but we will elaborate on it much more in the following sections.
We start this by showing you the process of developing a Knowledge System
so that it can solve a business problem that you might have.

How to design an AI project

In the previous section, we briefly touched on how to build a Knowledge

System.
As you remember, the premise was that you develop a Knowledge System
by formulating a goal that was derived from a business problem. The way the

209
 AI FOR CEOS

goal was solved was by designing a Knowledge System. This was illustrated as
a design like this:

The key is thus to build a Knowledge System that can realize the goals that
have been specified.
The best way to understand how you can do this “in real life” is to return to
our example with the curtain van.
Remember that they implemented a Knowledge System that was designed to
talk to the customers on the phone. This made it possible to book appointments
to measure up for a customer’s new curtains.
The business problem that the case needs to solve is to free staff resources
from the task of booking new customer meetings.
The goal is thus to have a system that can handle telephone bookings without
human interference and handle all aspects of the meeting booking.
We start with the three elements, but this time with the goal of solving the
curtain van case. That’s why our drawings look like this:

210
 LEADERSHIP AND AI

The next step is to look at which of the three main areas of a knowledge system
must be applied to solve the goals you have.
We gain this knowledge by breaking down our goals into a number of
sub-goals. Each of these sub-goals is then categorized in one of the three
categories that a knowledge system consists of (assistant, management, or
communication).
In our example, it makes sense to work with these sub-goals.

211
 AI FOR CEOS

What we will see is that our Knowledge System will be focused on communica-
tion and management and that it can be drawn like this:

212
 LEADERSHIP AND AI

In this case, there are 8 AI algorithms, each with their own purpose, which
combined can be used to solve the goals you have set out to solve.
As a starting point, you work at a corporate level with Knowledge Systems.
You do this because you might already have an architecture with algorithms
that have been developed for another purpose that may be reusable.
That is precisely the fundamental idea of working with a Knowledge
System. That you build shared knowledge on many dimensions so that your
organization can create a unified intelligence for your business.
This means that if you have already made AI solutions before the curtain van
case project, then you may have AI algorithms that can be reused for our new
purpose. If not, then you start from scratch with the ones that need to be made.
And they should all be made so that they can be used in another constellation
for other purposes at a later date. This is how you grow the company’s overall
digital intelligence.
Some of the algorithms that you would have to use can also be purchased as
a service. Both Google, Amazon, Microsoft, and IBM provide AI services that

213
 AI FOR CEOS

can speak English (and many other languages). It will always make sense to
use as many of such types of services that exhibit general business logic. As
the alternative - developing them yourself - never gets better, or cheaper.
Each of the algorithms that must be built specifically for our purposes must
now be trained to solve their individual goals. When this has happened, the
entire knowledge system is trained towards the overall objective. When this is
done, you are ready to go live with the solution.
It sounds simple, and in a way, it is. However, I left out on one part, and that
is how you train your AI. The next section will teach you how to do this.

The art of training your AI

I have mentioned several times that you need to train an AI application to make
it work. I also told you fractions on what that meant. But you have not got
the complete picture of how to do so yet. Now you will, and there are several
reasons why this should be important to you.
The first reason is that it is a crucial process of how your work with AI and
thus building your company’s digital intelligence. Another reason for you to
have an interest is that this way of working is very different from how you
develop IT systems. So, if you are an old dog, this is where you must try to
learn new tricks - even if it conflicts with how you used to do things.
I use an example to explain the method. It is based on the Stanford case with
the identification of skin cancer that you already know by now.
In this example, we would also like to train our AI to be able to detect whether
or not a patient has skin cancer. I don’t know how Stanford did their AI training,
but it could have been something like this.
So, let’s say we too would like to train our AI model to do the same as they did
at Stanford. Namely, analyze birthmark photos, and use them to determine
whether or not they could develop into cancer.
Before AI was available for this task, the usual way to identify skin cancer
would have followed a process similar to this description:
You begin with a starting point. Probably one or more high-resolution
photos of birthmarks from potential patients. This is our starting point in the

214
 LEADERSHIP AND AI

illustration below.
From these photos, an action is made by an employee. In our example, this
means that an expert will analyze the images. This analysis will be completed
with a diagnosis. Either the patient is ok, or he needs further examination if
they cannot rule out the risk of cancer by analyzing the pictures.
The process follows a flow which can be illustrated as follows:

It is the same process we want our AI to conduct for us as well. So, our AI can
make the diagnosis instead of the expert.
This means that the starting point in the AI case is this one:

215
 AI FOR CEOS

Our problem is that we would like to identify patients with skin cancer. Our
goal is to be able to sort the patients in two groups: Potential skin cancer
candidates, and the rest (the known healthy patients).

216
 LEADERSHIP AND AI

As you can see the illustration above is a process with five different stages:

1. Define goals, start, action and success criteria

2. Collect data.
3. Train model
4. Let your AI make the diagnoses (what is OK and not OK)
5. Model feedback

What you do as you go through the five stages is to train your AI. You teach it
how to separate the birthmarks that are ok, from those that are not. What you
really do is train your AI to spot what differences in the data that may explain
why one birthmark is OK and another one is not.

The AI algorithm does this by building a model for the “ok” and “not ok” case
based on patterns in your data. When the AI analyzes a new birthmark photo,
then it will try to compare it with the two output cases “ok” and “not ok”. It
will then match the birthmark with the output case that holds patterns that
are most similar to the photo it is currently looking at. This is how it will make
its recommendation as to where to classify any given birthmark photo.

At some point, your AI reaches a point where the quality of its response
matches your requirements. You now have an AI model that can perform

217
 AI FOR CEOS

the job it has trained itself to handle. Sounds easy?

It can be. However, you can expect a great number of elements to happen
during the different training phases. The next couple of pages let you in on
what to watch out for.

First stage: Define goals

The prerequisites for your AI to take on a task is that it can be described as

a flow similar to what is as shown in this illustration. Thus, with a known
starting point, an action that has two outcomes. One outcome that is the
successful case we aim for, and one that is not.
In this example, there is only one goal, namely if the diagnosis is ok or not
ok.

Second stage: Collect data.

218
 LEADERSHIP AND AI

Your AI learns from data, and the more data you have, the better prerequisites
your AI has to learn. This one of the aspects that makes your AI fundamentally
different from IT. AI does not relate to the amount of data and the structure of
data in the same way an IT system does.
This fact is related to the knowledge pyramid. Remember the weather station
from the first part of the book? It was only when the measurements change
from data to information that you got value from the readings. Where ” 25”
was converted to “25 degrees”. Putting units on your data converted it to
information, and this was what enabled you to extract the meaning of the
content.
So, data turned into information. If you need to get results from an IT
system, then it must work with information, so that you, as a user, know how
to interpret the findings you get. So, IT needs data that has been converted to
information to be able to work properly.
AI is not interested in the meanings of data or information as such. AI looks
for patterns, and meanings in patterns of data. So, there is no need for AI to
know the connection between data sources. Or to know the unit of the data it
is looking at. What AI needs is the ability to find patterns in the available data.
So, the quality requirement for AI data is that it is reasonable to expect that
it can contain patterns that can explain a behavior. So good quality data for AI
is something other than good quality data for IT because the two technologies
use data for different purposes.

219
 AI FOR CEOS

Third stage: Feed your AI with data and answers

Now you start teaching your AI how to make decisions. You teach it to be able
to tell you whether the diagnosis is possible skin cancer or not. In our example
you do this by feeding your AI with outcomes, i.e. patient data. You then tell
your AI whether the data it is looking at is in the “OK” or “not OK” category.
The purpose of this is to help your AI understand the two different categories
“OK” and “not OK”. This exercise will help you AI identify the strongest
patterns that best can explain why one photo should be in the “OK” category
and why another should be in the “non-OK” group.
So, your AI will split your data into two groups. The data associated with the
users with the result “OK” and the data related to the “not OK” users. Your AI
will then look at the strongest differences between the two groups.
The AI will look for meaningful patterns that can explain the difference
between the two groups. You will get feedback from your AI based on the
quality of its findings. Usually, as a percentage describing the level of certainty
and accuracy of its findings. If you are happy with that level, you go to the
next stage. If not, it will require more or different data for you to be successful,
and that means you need to go back to the second stage.
Fourth stage: Let your AI give you answers

220
 LEADERSHIP AND AI

At this stage you have an AI model that can give you answers. If you feed it
with a photo, it will tell you if it thinks that it is either in the “OK” or “not OK”
category.
Now your AI begins to tell you what the outcomes are. Your AI will now
compare the data it is processing from you and compare it to the two pattern
groups it has already created, ie. the “OK” group and the “Not OK” group.
Your AI will then choose which of the two groups it believes has the strongest
similarities to, based on the patterns in data that it has found in the third stage.
So, you now get answers from your AI. It starts giving you diagnoses based
on the data from the cases it knows of.
So, your AI can produce an output. The next stage has to do with validations
of the quality of the outputs.
Fifth Stage: Give your AI feedback
Here you tell your AI if the answers you got in the fourth stage were right or
wrong. Remember that you feed your AI with test data, so this means that you
know what the correct output should be.
This is another dimension in training the AI that has to do with output
quality. You will tell the AI where on a matrix the answer would fit.

221
 AI FOR CEOS

If you have worked on clinical trials, or in the field of statistical probability,

then you may recognize this matrix as a false positive or a false negative
outcome.

Remember the focus now is on whether the quality of the diagnosis is OK. It
probably won’t be in the beginning. So, you continue the training. Either by
running several cases through your AI system, i.e. repetitions of stages three,
four, and five.

222
 LEADERSHIP AND AI

Your aim is quality improvement, and this means that you want to improve
is your AI’s ability to make better decisions. So, it becomes better in
understanding what defines the difference between “OK” and “not OK”. You
will also be able to go back to the second stage and give your AI more or different
data if your suspect inadequate data quality is the cause of low-quality output
from your AI.
However, at some point, hopefully, you will reach a quality level of the
answers that are so good that you are ready to unleash your AI and let it begin
to make diagnoses. At this point, you go live with your solution.

Three considera ons about the decisions AI can make for you

In the example that we have just walked through, the goal was to analyze
birthmarks. There are some things you should know when it comes to AI and
knowledge.
Let’s say we did succeed in building the skin cancer AI. Does this make our
AI a skin cancer expert? No. Our AI knows how to detect skin cancer in a photo,
but it has no knowledge of skin cancer as a subject. And it never will.
Just as self-driving car AI does not know anything about what driving a car
means. It knows what it takes to drive a car, but it does not know anything
about the concept of driving a car.

223
 AI FOR CEOS

AI can give us recommendations and make decisions because it can detect

patterns in (a large) set of data. But the AI does not understand the patterns.
This fact has three critical consequences that you need to be aware of:

1. AI knows nothing about your field - and never will

2. The basis for AI’s decisions is unknown
3. The better your Knowledge System is, the more abstract answers you can
get

AI does not become a skin cancer expert, even though it may be able to provide
diagnoses that are better than any a trained expert would be able to do. This is
the case for all subject areas that you use AI for.
What really makes AI great is that there are actually patterns that are strong
enough to separate two different groups “OK” and “Not OK”. So, it is the fact
that such patterns exist that make AI great. Not AI in itself. AI is just a tool
that makes it possible to find the patterns that are already there. We just don’t

224
 LEADERSHIP AND AI

have any other way of seeing and working with them than to use AI.
So, when I say that AI is a Knowledge System, it’s not because AI is a
knowledgeable system. It is because AI can make decisions based on structured
knowledge processing.
The other element that you need to be aware of is that nobody knows how
AI makes decisions.
That’s because it’s not explicitly clear what weight your AI assigns to the
patterns it sees. You can say that you get a rather mechanical response from
your AI. By this, I mean you will get the conclusion: “This patient does not
have cancer”, but the “why” remains unanswered. There is no reason or
elaboration of what lies behind the decision.
Many times it is no problem as long as the quality of the answers is high.
However, there may be an ethical dimension to this problem.
For example, let´s say you use AI to screen job candidates. One of your
criteria is that the likelihood of the candidates ending up working for you for
more than five years should be high. Requirements like this mean that you
could risk ending up with your AI using discriminating factors when it selects
candidates to you.
Graduates tend to change jobs more frequently at the beginning of their
careers, so they might be excluded deliberately by your AI. Your AI might also
choose to avoid women in their 30s because they might have children at that
age and from a purely mathematically point of view, young mothers might be
considered unstable in the eyes of your AI.
Of course, such factors can be corrected if you are aware of the problem. You
do this by requiring you AI do select an appropriate proportion of women,
graduates, etc. However, because the criteria are not known, you do not
actually know whether your AI uses discriminatory criteria unless it is actually
detected and perceived as a problem.
The last thing you need to be aware of is that the more mature your
Knowledge System is, the more abstract goals it will be able to solve for you.
This is important because there is potentially more value in being able to
solve an abstract goal than a more concrete goal. An abstract goal could be
“driving a car”, where a more concrete goal could be “Dispatch a parcel from

225
 AI FOR CEOS

a warehouse”.
In general, abstract goals has more business value, because the alternative
to letting AI do your abstract task is probably to have it done by humans.
What characterizes more abstract tasks is that it is more difficult to specify
their purpose.
The goal of a self-driving car is not to drive from A to B.
The goal is to get you from A to B in a way that is comfortable, safe, and in a
way that follows the law and many more of such less specific requirements.
Such conditions make it challenging to formulate goals and requirements in
specific terms.
So, the implementation of the abstract task requires a design of a Knowledge
System where multiple AI algorisms works in a joint and complex set up. So,
it is a demanding and expensive task that they will have to solve.

Your learning process will never stop

I would like to end this part of the book by sharing some findings from how
we managed to create AI in my company, Canecto.
The first thing you have to be aware of is the role structure of the develop-
ment team. Obviously, you need a Data Scientist to make the AI models, but
an equal part of the task lies in the analysis part of the project’s delivery.
Being able to break down business logic into goals and needs are classic
business analytics tasks. The overall quality of your end result depends on
how you approach these tasks. So, the business analyst role is as critical to
the quality of your delivery as a good Data Scientist is.
Another central finding is that we failed to create a well-functioning
Knowledge System in our first attempts. This is likely to happen to you too.
You will find that the first algorithms you make do not give you the results
you hope for.
However, they could give you something that is much more valuable to you:
Knowledge of how to design one that actually could be made to work. The
development process itself is a learning process for you. In Canecto, we started
by spending four weeks on our first algorithms and we can now make them in

226
 LEADERSHIP AND AI

2-3 days. In a significantly better quality.

The fact is that nobody knows what the actual output of your deliveries
before you start working with them. And even when you have a successful
Knowledge System to put into operation, you will always be able to get more
quality out of it. Because you gain a better understanding of what works and
because the system itself learns what works and adjusts the output accordingly.
The good thing about this situation is that there are built-in mechanics
that ensure continuous quality improvements to your product or service. The
downside is that you will never stop developing the system because it requires
resources on your part to be able to implement the improvement opportunities
that you can see.
So, in terms of management, the development of knowledge systems
fits really well into an Agile context, which works with continuous quality
improvements.
On the other hand, it will be a problem to approach an AI delivery as a project
that is dissolved after its deployment, as you risk missing out on the value of
the teams learning in the process towards project completion.

How to work with AI as a co-worker

As employees, we are used to working with IT and software. We use IT to solve
specific functions (Excel, email client, Word) and to help us as part of a process
using software from companies such as SAP or Salesforce.
So, as an employee. Isn’t AI just more of the same? Isn’t it just IT with a
twist?
No, it’s not. Remember the knowledge pyramid? It showed us that AI gives
us knowledge, where IT gives us data. And the consequences of the difference
between the two things are much more than words.

227
 AI FOR CEOS

The consequence of working with a system that provides knowledge and not
just data is that we have access to tools that allow us to do things we probably
would not have imagined possible.
For example, building advanced products in record time, targeting customer
groups with messages we had no idea they would respond to. Working with AI
can open up a new world of opportunities.
But it comes with a price. The technology will challenge us in the way we
work and the way we see our world.
This section introduces you to reflections on what it means to work with AI.
To do this, we have to go back to the concept of the knowledge system and its
three main groups:
An assistant, bots, and the automation element.

228
 LEADERSHIP AND AI

As an employee, you can work with AI in all three dimensions. However, the
most normal will probably be working with AI as an assistant to you.
Remember that you were introduced to Use Cases where AI, as an assistant,
was applicable. It was relevant in situations such as:

• Product Development: You use AI to create products and prototypes that

you would otherwise not be able to make. For example, drones or new
lightweight parts for aircraft.
• Predictive analytics: you would let AI find customer segments for you, or
suggest to you new promising medical trials cases.
• Virtual Assistant: Here, AI will do small tasks for you automatically and
keep an eye on factors that are important to your work. For example,
monitor your competitors’ behavior and tell you when they start doing
something important that you haven’t discovered before.

The common denominator for all the tasks is that you do something with AI.
You use AI as a tool that makes you better in the way you create value. You
work with AI.
You could also use AI in your job in other ways. AI might drive your car, or
talk to your customers, etc. This means that we are using the communication
and management part of the Knowledge Systems. Here is the consequence

229
 AI FOR CEOS

that AI takes over tasks from you, so you don’t really work with AI in such
cases.
This is why our focus now is on the assistant and what it means to work with
AI.

What does it mean to work with AI?

To help you understand what it means to “Do something with AI”, we turn
our focus on our old Greek friend, the philosopher Porphyry. And why so, you
might ask?
We do this because he can help us shift our focus from how we use to look at
thinks, and what it takes to change our perspective.
Remember the talk on how our view of the world is to make tools that make
us better, on our terms?
That is the premise we need to challenge when we work with AI. And this
is where Porphyry can help us. The Porphyry tree represents the way we as
humans perceive the objects we use in a general sense.
As humans, we seek to understand our world. If we can describe our world
as concepts, as formulas, and perhaps as mathematics, then it gives us a sense
of wellbeing and of control. We have a certainty that we understand the world.
Working with AI means we’re going to let go of that kind of control. We as
humans are not necessarily going to be in control. The technology will be better
than us in some areas, and there will be areas where we do not understand
why it produces the results it actually does.
The skin cancer example is a good representative of those circumstances.
Here, the AI algorithms found 12 indications of cancer. Three more than what
the doctors could find. We cannot explain why and how it found the last 3 cases.
But we can see that they are right and that they do contribute to explaining
skin cancer.
The consequence of a technology at this stage is that we now work with the
technology. It does not work for us. It becomes more of a colleague than a
tool.
To accept this, we need to let go of our expectations that we, as humans,

230
 LEADERSHIP AND AI

must be in control of any work situation. And it will be a difficult change.

The fact is that now AI is not just not a tool that makes us better at our terms.
AI can make us better, but the way it can happen may be defined by AI and not
by us. Because AI can tell us:

• That there are three other factors that are also indicative of cancer
• That you can build a drone better than how we do so now, or
• That there are more attractive customer segments than the ones we are
working with right now

The problem here is not AI. The problem is that this way of working conflicts
with thousand years scientific views of the word.
It is deeply rooted in humans that we want to understand our world. It gives
us as humans a sense of security that we can understand and explain how the
world around us works.
So, it is a basic substance of being a human we ask you to no longer do when
you work with AI. You have to let go of the perception that we must understand
the world around us and that it can be seen as an extension of us as humans.
For example, we describe nature with words and concepts that make sense
to us as human beings. Like a bird has wings. We define a mammal as an
animal that does not lay eggs, they feed their young instead. Our worldview is
a product of our human development, our culture, our language.
We describe the world as we see it, in a way that we can communicate,
and that makes sense to us. And so, it should be. It makes it easy for us
to communicate in such ways because it enables us to achieve a collective
understanding of our world.
We should not stop that. But we will be challenged by a new way of
understanding our world that does not have the same basis in our language,
culture, and history.
Our ability to coexist, with the new way of seeing the world as AI does it, is
what ultimately determines how successful we will be in our exploitation of
AI.
We will work with AI as an equal partner, and not as a technology that does

231
 AI FOR CEOS

things for us. Because AI as a basic premise does not perceive the world the
same way we do.

That is exactly why we are going to change our worldview. We cannot start
with the premise that human knowledge is the center of what we are looking
at.
What is new to us as human beings are that knowledge-based systems can
make decisions. And the systems can reach a level of sophistication where
they will be able to make decisions that are better and faster and with a smaller
margin of error than what humans would have been capable of in the past.
And these are the perspectives that have the potential to change everything.
We must learn to work with a system that is more qualified to make a given
type of decision than we will ever be able to do ourselves. And not only that.
We actually don’t know why and how the system makes the (right) choices.
The only thing we can see is that both in experience and in a direct comparison
between humans and AI will show that our AI makes better decisions than we
do.
So, this is where Porphyry’s decision tree comes into the picture again, for
we will be confronted with a conflict between two worlds here. As human
beings, we want, as in Porphyry decision-making rationally, to understand
why a decision is made. But we might no longer have this opportunity in the
Knowledge-Based systems.

232
 LEADERSHIP AND AI

As in the Stanford cancer screening case. We can actually see that the system
is better than us in making decisions. So, we have to accept a new world that we
cannot understand or explain rationally. We can see the Knowledge Systems
are better than us, but we cannot explain why, or understand how they have
come to their - right - conclusions.
We need to learn to ignore the dimension of control that is a core part of
how we act as humans.
I find the aspects and perspectives that have been uncovered in this section
fascinating. If you do too, then visit my blog for more articles on the topic.
However, there is also an aspect that has to do with actually getting things
done—executing on what AI can do for you. I have been doing this for the last
three years, and the last part of the book summarized my experiences with
working with AI.

My personal experiences with management and AI

Hopefully, this book has helped you set the direction for how AI could be used
in your business.
I would like to end the book by sharing some of my personal experiences of
running a company where AI is a huge factor as it is in my company, Canecto.
My first point to you is that AI is a learning process and that working with
AI should be seen as a continuous process.
Whatever goal you set with your AI project, you won’t reach it on your first
attempt. But trying to reach your goals better prepares you for getting there
the next time. You might also discover in the process that there are other and
better goals you can achieve.
So, you have to get started with your AI projects as soon as possible. And
you have to do so with built-in stamina in your projects.
For most people, it is properly a good idea to start to solve a simple problem
with AI. And then be ready to take the detours it will take before you reach a
goal with more value to you.
I mentioned earlier that we had gone from a development time of 4 weeks
to 2-3 days to make a new AI application. This journey is a consequence of

233
 AI FOR CEOS

many different conditions. We have designed our data structure in a way that
we can easily navigate it. We are better at formulating requirements for our
solutions that can actually be solved with AI. We can interpret the results we
get from a new AI application quickly. Thus, we can see if it makes sense to
refine it or whether it should be scratched.
So it is the combination:

• The team’s overall understanding of the AI models they work with

• The insights into their data, and not least
• Their ability to articulate specific business desires as goals that can be
achieved with AI

which enables us to achieve such a level of efficiency.

That is why an AI team needs time and stamina. Because it’s not going to
deliver on the same terms that an IT development team can. The premises for
AI development are different, and that means that you must make room for
this in the way you as a manager lead your AI projects.
For instance, your AI teams should not provide code as and follow a project
delivery roadmap the way your IT teams shall. Your AI teams must be measured
by their ability to systematize and process knowledge. Such requirements will
change the way they should be managed.
As a leader, you do not get results from your AI team by demanding deadlines,
milestones, and tangible results. Because they are not in a position where they
can promise you such management deliveries.
Basically, what your AI teams are delivering to you is enriching your
corporate intelligence.
What you get for your development investments is more digital knowledge
in your business. I agree that it should be placed in a business context, and
practical solutions should be the result. That will also happen; it will just
materialize differently than if it was an IT project.
Leading AI projects means that your requirements for what your team needs
to deliver are different. Your requirements for your team should be formulated
as goals their AI should be able to achieve. Examples of this could be:

234
 LEADERSHIP AND AI

1. Tell me how we can sell 10% more of product type X online

2. Tell me which customer group is the easiest to sell more new products to
3. Tell me how we can make our product 20% easier without reducing the
strength
4. Build a bot that can advise our sellers on the support of our latest product
If your teams have not worked with AI before, it will probably take a few
months to get an answer on the first point. But if you made the AI it takes to
get an answer to point 1, then you have pretty much the answer to bullet 2 for
free. And you will be able to get answers to many other similar questions very
quickly.
That’s why you need to get started with AI and then the learning will start to
grow. Because when you as a manager know that the answer to 1 and 2 is now
essentially free, then I am sure that you can find a lot of similar goals that you
would like to get answers to too. Because now AI is changing the premise of
what you can do as a manager too.
The same argument applies to the prototype example in point 3. If you can
make one prototype, then the next one is almost free. Because you have gained
the learning, it takes to work with technology in a new way. And the same
goes for the last bullet. If your bot can service your salespeople, then you’re
probably close to having another bot that can do something similar for your
support team.
You should also expect new options to appear for your business. As your
learnings from your experiences with AI starts to unfold, you will see new
opportunities. You now suddenly see ways to solve tasks and gain insights
that you would never consider being able to get earlier.
So as a leader, you have two goals with your AI teams. They must deliver
results just as is the case for any other team. Equally important is that they
must also build the competence that enables you to be able to solve tasks with
increasing complexity and business value.
And as a leader, you must not inhibit your team’s ability to do so, by being
too quick to make specific demands on your teams. By doing so, you risk killing
the knowledge creation process before it starts. You also risk making demands
to your team that they do not have the prerequisites to deliver to you.

235
 AI FOR CEOS

The consequence of both such requirements from you will be that your AI
efforts will die. Not because of a lack of business opportunities, or because of
your team’s competence. But because the premise of how their work should
be done and the success criteria you set up would never work for them.
Your team may not be able to see such consequences (how could they),
especially t if it’s a new team: So they won’t even be aware of such issues
until very late in the development process, at which time it may be too late
to fix. It is your responsibility as a leader to ensure the right conditions and
requirements for your AI team exist.
My next advice is to look at the (wo)man in the mirror. If you are a manager,
then it is very likely that you are the greatest obstacle to getting started with
AI.
I often hear that the reason why someone is not working with AI is that
they lack the right resources and therefore, do not have the opportunity to get
started.
In my opinion, that’s a terrible excuse, especially if your reason is that you
have too few Data Scientist who can develop AI solutions for you.
My experience is that 80% of an AI project involves tasks that do not require
that you work with actual data. So, if you really wanted to work with AI, then
any lack of Data Scientist should not be an excuse for not working with AI.
This point is especially true because you can easily source highly qualified
Data Scientist as an external subcontractor. Some of the best Data Scientist I
have worked is Bulgarian developers that I have never met in real life, as an
example.
In my opinion, the right reason for not starting is almost always that you do
not understand as a manager what it takes to start an AI project. Therefore it
is convenient to use the lack of Data Scientist as an excuse.
A test on your self and your possible AI project resistance could be your
answer to this question:
“If you had a team of competent Data Scientist available tomorrow, would
you be able to have an AI solution ready in 2 months?”
I guess that your answer will be “No”.
Your reason for this answer could be that you also would be missing the

236
 LEADERSHIP AND AI

people from your business, who understand the span of data and industry that
you operate in.
It is their work that makes up 80% of the project time that I mentioned
before.
Don’t get me wrong. I have the highest respect for the value that a Data
Scientist produces. But the real value of what you can create with AI is your
ability to solve your business problems with AI. And even the best Data Scientist
can´t do this for you alone. It requires in-depth knowledge of your business
and how you operate that they most likely do not have.
The good thing about that problem is that you probably have the people in
your organization who know this already. These are your business analysts,
your specialists, your project managers. The people who drive change in your
business.
It will be a new task for them to break down business requirements into
something that can be solved with AI. But if they have already tried to make
the requirement for the development of IT solutions, then they can also learn
to make AI requirements too.
So you probably already have the most important resource needed to make
AI. They just have to learn how to do it.
I think that is a huge problem. That you opt-out of AI for the wrong reasons.
It is also one of the reasons why I wrote this book. So those who need to take
the plunge into the AI world feel ready to do so.
Another big reason why I experience a lack of interest in working with AI is
risk.
Admittedly, there is a very real risk that the AI projects you start will end
in failure, especially the first ones. You risk not to reach your goals with your
AI projects, that you cannot find patterns in your users’ behavior, and thus
cannot make usable AI solutions.
So, you run a risk when you start an AI project. But you also run the risk of
doing nothing.
Imagine if there were patterns in your users’ behavior that would enable to
sell 10% more to your customers, or if AI could make a product prototype at a
sophistication level you never dreamed possible. But you just didn’t know it.

237
 AI FOR CEOS

You run the risk of missing out on opportunities to create something

amazing with AI by doing nothing.
And who knows. Maybe others than you are willing to take that risk. Which
means that their products sell well in your markets. That their new products
are better than yours and that they are more effective than you. Because they
took the risk, you weren’t willing to take.
The understanding of AI that you gained after reading this book will only
gain value when you transform it into actions. That is why I have been so
insistent on getting you to use the online tools this book has.
Because acting on AI is probably your biggest problem, and I would like to
challenge you by showing you that you need to be operational with AI. It’s not
just something to talk about at board meetings or seminars.
But, most of all, I hope it is the insight you now have gained AI gives you the
spark you need to get started even if you haven’t used the book’s tools.
When you do start, then you need to know working with AI is a journey.
No one can tell you in advance where the journey ends. But hopefully, it will
be a long journey where you will be able to create value in new ways—a tour
where you will learn what you can do with this new technology along the way.
I hope it has been worth your time to read this book and that you are now
getting ready to get started to creating the value with AI that awaits you.
If you have questions or thoughts on how you use AI, visit cognifirm.com,
which is constantly updated with more knowledge about business and AI. And
also remember all references from the books, including all reference URLs
can be found at coginifirm.com/book, and that the tools from the book are
available at cognifirm.com/tools.
You are also very welcome to follow or contact me on social media. All the
links can be found on the book’s website too. I would also appreciate your
honest review of the book.
Remember to do the difficult things while they are easy and do the great
things while they are small. A journey of a thousand miles must begin with a
single step.
(Lao-tse)

238
 About the Author

Per is an international keynote speaker, entrepreneur, and CEO.

He speaks on how AI can transform the foundation on which a business could

be built upon.

Per has been creating digital solutions for more than 20 years. He founded the
Copenhagen based AI startup Canecto in 2017 and Cognifirm in 2020.

The two companies run on an AI infrastructure design patent that he has

created.

Priorior to his startup endeavors, he spent ten years with IT management in

the finance industry.

So his insights in AI and business are rooted in two areas. The insights of how
AI works that he has gained as a startup CEO. And the management knowledge
on the challenges of how a large corporation runs.

239
The implications of the AI business potential, combined with his management
experience, have fueled his passion for how AI can redefine how a business
could be run.

You can connect with me on:

https://cognifirm.com/book
https://twitter.com/perdamgaard

Subscribe to my newsletter:
https://cognifirm.com/newsletter

240