DI State of AI in The Enterprise-2nd-Ed

State of AI in the Enterprise, 2nd Edition
Early adopters combine bullish enthusiasm with strategic investments

Contents
Executive summary | 2
Activity, investment, and positive results | 3
To maximize value, early adopters should become

risk and change management experts | 9
Early adopters want more talent, and need a

better mix of it | 14
Enthusiastic early adopters can take the next step

by getting serious | 17
Endnotes | 21
1
Executive summary
F
OR THE SECOND straight year, Deloitte risks. Project selection and managing return on
surveyed executives knowledgeable about cog- investment are also critical.
nitive technologies and artificial intelligence, 1
3. Early adopters need the right mix of
representing companies that are testing and imple- talent—not just technical skills—to ac-
menting them today. We found that these early celerate their progress. They are short of
adopters2 remain bullish on cognitive technologies’ AI researchers and programmers but also need
value. As in last year’s survey, the level of support business leaders who can select the best use
for AI is truly extraordinary. Our analysis uncov- cases. To garner this talent, they are training
ered three main findings: their current workforce, but many feel the need
to replace existing workers with new people.
1. Early adopters are ramping up their AI Early adopters also may need a strategic ap-
investments, launching more initiatives, proach to talent that automates what machines
and getting positive returns. Cloud-based do best, while still capitalizing on human judg-
cognitive services are increasing adoption by ment and creativity.
reducing the investment and expertise required
to get started. These findings illustrate that cognitive tech-
2. Companies should improve risk and nologies hold enticing promise, some of which is
change management. This includes reducing being fulfilled today. However, AI technologies may
cybersecurity vulnerabilities—which can slow or deliver their best returns when companies balance
even stop AI initiatives—and managing ethical excitement over their potential with the ability to
execute.
METHODOLOGY
To obtain a cross-industry view of how organizations are adopting and benefiting from cognitive
computing/AI, Deloitte surveyed 1,100 IT and line-of-business executives from US-based companies
in Q3 2018. All respondents were required to be knowledgeable about their company’s use of
cognitive technologies/artificial intelligence, and 90 percent have direct involvement with their
company’s AI strategy, spending, implementation, and/or decision-making. The respondents
represent 10 industries, with 17 percent coming from the technology industry. Fifty-four percent are
line-of-business executives, with the rest IT executives. Sixty-four percent are C-level executives—
including CEOs, presidents, and owners (30 percent), along with CIOs and CTOs (27 percent)—and 36
percent are executives below the C-level.3
2
Activity, investment, and

positive results
A
YEAR LATER, AND the thrill isn’t gone. In
Deloitte’s 2017 cognitive survey, we were
struck by early adopters’ enthusiasm for
cognitive technologies.4 That excitement owed
much to the returns they said cognitive technolo-
gies were generating: 83 percent stated they were
seeing either “moderate” or “substantial” benefits.
Respondents also said they expected that cognitive
technologies would change both their companies
and their industries rapidly. In 2018, respondents
remain enthusiastic about the value cognitive tech- layers of abstract variables. Deep learning
nologies bring. Their companies are investing in models are excellent for image and speech recog-
foundational cognitive capabilities, and using them nition but are difficult or impossible for humans
with more skill. to interpret. New technologies are making it
easier for companies to launch deep-learning
projects, and adoption is increasing. Among
Higher adoption, multiple our respondents, 50 percent said they use deep
options learning, a 16 point increase from 2017—the
largest jump among all cognitive technologies.
Compared with their counterparts in typical • Natural language processing is the ability to
companies,5 our early-adopter respondents have extract or generate meaning and intent from text
high—and growing—penetration rates of key cogni- in a readable, stylistically natural, and grammat-
tive technologies: ically correct form. NLP powers the voice-based
interface for virtual assistants and chatbots, and
• Machine learning is the ability of statistical the technology is increasingly used to query data
models to develop capabilities and improve sets as well.6 Sixty-two percent of respondents
their performance over time without the need to have adopted NLP, up from 53 percent last year.
follow explicitly programmed instructions. Most • Computer vision is the ability to extract
cognitive technologies are based on machine meaning and intent out of visual elements,
learning and its more complex progeny, deep whether characters (in the case of document
learning. That includes computer vision and digitization) or the categorization of content
natural language processing (NLP). Machine- in images such as faces, objects, scenes, and
learning adoption was already high at 58 percent activities. The technology behind facial recogni-
in 2017, and it grew by 5 percentage points in tion—computer vision—is a part of consumers’
2018. everyday lives. For example, some mobile
• Deep learning is a complex form of machine phones permit their owners to log in simply by
learning involving neural networks, with many
3
looking at them, via facial recognition.7 Com- finance, video analysis in brand management, and
puter vision technology “drives” driverless cars trouble ticket analysis in customer service. The
and animates cashier-less Amazon Go stores.8 need for companies to develop bespoke cognitive
Computer vision has also gone mainstream with initiatives will likely decline as similar services
our survey respondents, 57 percent of whom say enter the market.
their company uses it today. Off-the-shelf can go only so far, however. Many
companies will likely need to develop customized
What’s behind the growth of cognitive technolo- solutions to meet their lofty expectations for cog-
gies among early adopters,9 especially the popularity nitive technologies. Here, too, there are tools to
of sophisticated technologies such as deep learning? accelerate adoption. Many of the big cloud providers
One answer is investment. Thirty-seven percent of offer AI through an as-a-service model: Instead of
respondents say their companies have invested having to build their own infrastructure and train
US$5 million or more in cognitive technologies. algorithms, companies can tap into the technologies
Another reason is that companies have more ways
FIGURE 1
to acquire cognitive capabilities, and they are taking
advantage. Nearly 60 percent are taking what is Enterprise software represents the
perhaps the easiest path:10 using enterprise soft- most popular—and easiest—path to AI
ware with AI “baked in” (see figure 1). Respondents who report their company
More respondents gain cognitive capabilities uses this method of acquiring/developing AI
through enterprise software, such as CRM or ERP
systems, than by any other method. These systems Enterprise
59%
have the advantage of access to immense data sets software with AI
(often their own customers’ data), and can often be
used “out of the box” by employees with no special- 59%
Codevelopment
ized knowledge. 53%
with partners
The cognitive tools available through enterprise 59%
software are often focused on specific, job-related
tasks. While this can make them less flexible, they
Cloud-based AI 49%
may be impactful nonetheless. For example, Sales-
90%
force Einstein can help sales reps determine which
leads are most likely to convert to sales, and the 59%
Open-source
optimal time of day to contact those prospects. development tools 49%
Moreover, vendors continually develop advanced 59%
tools, which are gradually integrated into the soft-
ware. Salesforce recently developed an advanced Automated
46%
machine learning
NLP model for handling multiple use cases that
typically require different models.11
59%
The “easy path” will likely become even more
Data science 44%
attractive as software vendors and cloud providers modeling tools
develop AI offerings tailored to business functions. 59%
Google recently announced a set of prepackaged AI
services aimed at contact centers and HR depart- Crowdsourced
39%
development
ments.12 SAP’s AI capabilities, which it collectively
calls “Leonardo Machine Learning,” also include
specific solutions such as cash management in Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.
4
they need right away, and pay only for what they use. Cognitive technologies are
According to a recent Deloitte study, 39 percent of a necessity, not an option
companies prefer to acquire advanced technologies
such as AI through cloud-based services, versus Many early adopters are investing in cognitive
15 percent that prefer an on-premise solution.13 technologies to improve their competitiveness.
Indeed, the appeal of the AI-as-a-service model is Sixty-three percent of surveyed executives said their
reflected in its annual global growth rate, which is AI initiatives are needed to catch up with their rivals
estimated at a remarkable 48.2 percent.14 or, at best, to open a narrow lead (see figure 2).
Cloud-based deep-learning services can give And the linkage between adept application of AI
companies access to immense—and previously and competitive advantage appears to be growing
costly—computing power necessary to extract stronger. Eleven percent said that adopting AI is
insights from unstructured data. They can also of “critical strategic importance” today, but 42
manage large data sets and accelerate app develop- percent believe it will be critical two years from
ment with pretrained models. now. This is a small window for companies to hone
While there are myriad ways for companies to their AI strategies and skills, and they believe their
access ready-made AI or develop their own, many success depends upon getting it right. Executives
also seek outside expertise. Fifty-three percent of are becoming more realistic about the time this will
respondents codevelop cognitive technologies with require, however. In our 2018 survey, 56 percent
partners, and nearly 40 percent use crowdsourcing of respondents said cognitive technologies would
communities such as GitHub. transform their companies within three years, down
Through cloud services and enterprise soft- from 76 percent last year. The same was true of in-
ware, companies can try cognitive technologies dustrywide transformation: 37 percent of our 2018
and even deploy them widely, with low initial cost respondents think it will happen within three years,
and minimal risk. The growing number of cloud- 20 points lower than in 2017. We believe executives
based options may explain the spike in pilots and are acknowledging the complexity of using cogni-
implementations between 2017 and 2018. Fifty-five tive technologies to drive change across lines of
percent of executives say their companies have business, without despairing of attaining that goal.
launched six or more pilots (up from 35 percent in
2017), and nearly the same percentage (58 percent)
Earning while they learn
claim that they have undertaken six or more full
implementations (up from 32 percent). Many companies’ AI goals extend well beyond
ROI. Positive ROI, however, can build momentum
for future investment and generate support for ex-
FIGURE 2
AI helps organizations keep up with the (Dow) Joneses

Relative to competitors, respondents say their company’s adoption of AI has allowed them to . . .
16% 20% 27% 28% 9%

Catch up Stay on par Edge slightly ahead Widen a lead Leapfrog ahead
Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

Deloitte Insights | deloitte.com/insights
5
ecutive champions of AI, and the technologies seem tions and delivering superior customer experience.
to be delivering. In our survey, 82 percent said they Netflix found that if customers search for a movie
have gained a financial return from their AI invest- for more than 90 seconds, they give up. By using AI
ments. For companies across all industries, the to improve search results, Netflix prevents frustra-
median return on investment from cognitive tech- tion and customer churn, saving US$1 billion a year
nologies is 17 percent. Some are more adept than in potential lost revenue.17
others at turning investment into financial benefits Robust returns are not limited to tech companies.
(see figure 3). Both established manufacturers and innovative
While these returns are estimates based on self- startups are using AI to make manufacturing more
reported data, they show that executives across efficient. For example, industrial firms, such as GE
industries feel they’re getting value from cogni- and Siemens, are taking advantage of the data in
tive technologies. Tech companies are spending “digital twins” of their machines to identify trends
significantly on cognitive, and getting a strong and anomalies, and to predict failures.18
return. They are also the driving force behind cog- Companies such as these are using AI to improve
nitive technologies, developing them for a market business processes, which are prominent benefits
already estimated at US$19.1 billion globally.15 companies seek. In fact, our survey findings suggest
This includes giants such as Google, Microsoft, and that companies are placing increased emphasis on
Facebook, and literally thousands of startups. AI 16
internal operations (see figure 4).
has also generated returns by improving opera-
FIGURE 3
Everyone’s winning, but some industries are winning bigger

AI investment and ROI: Relative landscape of industries
22% Low AI investment/high returns High AI investment/high returns
20%
Industrial products and services
Technology/media and
entertainment/telecommunications
Return on investment
18%
Professional services
16% Financial services and insurance
Consumer products
14%
Government/public sector
(including education) Life sciences and health care
12%
Low AI investment/low returns High AI investment/low returns

10%
Lower investment Median Higher investment
Note: The dotted lines in the graph represent the median ROI and median AI investment for all respondents, cross-industry.
6
FIGURE 4
AI’s leading benefits are enhanced products and processes—

and better decisions
Respondents rating each a top-three AI benefit for their company
2017 2018
Enhance current products

51%
44%
Optimize internal operations

36%
42%
Make better decisions

35%
35%
Optimize external operations

30%
31%
Free workers to be more creative 55%

36%
31%
Create new products

32%
27%
Capture and apply scarce knowledge

25%
27%
Reduce headcount through automation

22%
24%
Pursue new markets

25%
24%

This shift toward internal operations has been Health care and life sciences companies are in-
accompanied by a somewhat reduced emphasis on vesting in AI but, according to our data, have less to
integrating AI into existing products and services, show for it. Certainly, some health care “big bang”
although that remains the most popular objective. projects have disappointed thus far. However, ad-
In fact, operational change is often required before vances in fields as diverse as radiology and hospital
such integration can take place. Our respondents claims management show that AI offers substantial
may be realizing that they should make operational potential for value in health care,19 despite some
changes first. high-profile stumbles. For example, in a recent
7
study, deep-learning neural networks identified Earlier, we noted that eight in 10 surveyed exec-
breast cancer tumors with 100 percent accuracy utives claim positive ROI from their companies’ AI
by analyzing pathology images.20 Such advances, efforts. However, we should view ROI claims with
however, are thus far only in the lab and will take a bit of caution: Less than 50 percent of surveyed
time before entering clinical practice. companies measure key performance indicators
necessary for gauging financial returns accurately.
These indicators include critical elements such as
(Mostly) rational exuberance
project budget/cost, ROI, and targets for produc-
Despite the hype AI generates, many executives tivity, cost savings, revenue, and customers (such
are excited—not wallowing in a trough of disil- as satisfaction and retention). This lack of mea-
lusionment. That’s translating into investment. surement gets to the heart of a significant problem
Eighty-eight percent of companies surveyed plan with cognitive implementations: They are often not
to increase spending on cognitive technologies managed with the same rigor that companies use
in the coming year; 54 percent say they will boost with more mature technologies.
spending by 10 percent or more.
8
To maximize value, early

adopters should become risk
and change management
experts
B
USINESS AND TECHNOLOGY leaders con- low levels of experience with them, it’s unsurprising
front an array of challenges as they seek to that this was the most-cited challenge. Integration
create business value with artificial intel- into the business is a challenge for technologies
ligence. Many respondents cited implementation, in general, but it may be particularly problematic
integration into roles and functions, and measuring with AI given the impact it can have on knowledge-
and proving the business value of AI solutions as worker tasks and skills.
top challenges of AI initiatives (see figure 5). Imple- Companies sometimes struggle in AI projects
mentation can be a challenge with any technology, to navigate the “last mile” of behavior change.21 An
but given the relative newness of AI tools and the example we have seen is an organization that built a
FIGURE 5
Many early adopters struggle with the basics

Top challenges for AI initiatives: Ranked 1–3, where 1 is greatest challenge
Ranked 1 Ranked 2 Ranked 3 Ranked top three
Implementation challenges 13% 14% 12% 39%

Integrating AI into the company’s
roles and functions 14% 13% 12% 39%
Data issues (e.g., data privacy,
accessing and integrating data) 16% 13% 10% 39%
Cost of AI technologies/
solution development 13% 12% 11% 36%
Lack of skills 11% 10% 10% 31%

Challenges in measuring and
proving business value 10% 11% 9% 30%

9
machine-learning system to support the sales team system. (It is possible to automate this analysis, but
by predicting which prospects were likely to convert that would be an AI project of its own.)
and which customers were likely to churn. Though Getting the data required for an AI project,
the system worked as planned, the sales team was preparing it for analysis, protecting privacy, and
initially unprepared to accept its recommendations. ensuring security can be time-consuming and costly
The team had not been closely involved in the devel- for companies. Adding to the challenge is that data—
at least some of it— is
Twenty-three percent of respondents often needed before it is

even possible to conduct
ranked “cybersecurity vulnerabilities” as a proof of concept. We

have seen companies
their No. 1 overall AI/cognitive concern. that, because they had
not fully considered the
opment of the solution and neither understood nor difficulty of obtaining the data they need, decided to
trusted the results it produced. One way to avoid shelve projects and disband teams until they were
this problem is to involve business owners closely able to lay the proper data foundation.
throughout the development process so they can Some organizations also struggle to articulate
better understand what is being delivered. a business case or to define success for AI projects.
Anyone following business news about AI knows This may be because AI is viewed as experimental.
about the critical role played by data. Survey re- Sometimes it is because machine learning—one of
spondents consider “data issues” as one of the top the most widely used AI technologies—is inherently
challenges for their companies’ AI initiatives. There probabilistic, meaning that a new system’s ultimate
are numerous reasons for this. Some AI systems, performance can be difficult to estimate precisely.
such as virtual assistants to enable customer self- And sometimes it’s because the group charged with
service, require data from multiple systems that developing an AI solution is unaccustomed to devel-
may never have been integrated before. Customer oping business cases to justify its work.
information may reside in one system, financial
data in another, and virtual assistant training and
Managing risks of AI
configuration data in a third. AI creates a need for
data integration that a company may have managed It is a fact of life that novel situations often
to avoid until now. This can be especially chal- present new risks. The same is true of emerging
lenging in a company that has grown by acquisition technologies such as AI. Executives are concerned
and maintains multiple, unintegrated systems of about a host of risks associated with AI technologies
diverse vintages. (see figure 6). Some of the risks are typical of those
Another challenge for companies is that the type associated with any information technology; others
of data required for some AI projects is different are as unique as AI technology itself.
from the data with which they’re accustomed to
working. For example, some solutions depend on CYBER RISK
access to significant amounts of unstructured data Chief among the AI risks that concern executives
that may have been retained for record-keeping are cyber risks, which ranked as a top-three concern
but was never intended for analysis. In one virtual for half of our survey respondents (see figure 6).
assistant project we know of, the team needed to In fact, 23 percent of respondents ranked “cyber-
review thousands of recorded phone calls to identify security vulnerabilities” as their No. 1 overall AI/
common themes with which to derive rules for the cognitive concern. This apprehension is probably
10
FIGURE 6
Cybersecurity heads the lists of AI-related concerns

Potential AI risks of top concern to companies: Ranked 1–3, where 1 is greatest concern
Ranked 1 Ranked 2 Ranked 3 Ranked top three
Cybersecurity vulnerabilities
of AI 23% 15% 13% 51%
Making the wrong strategic
decisions based on AI 16% 13% 14% 43%
Legal responsibility for decisions/
actions made by AI systems 11% 15% 13% 39%
Failure of AI system in a mission-
critical or life-or-death context 13% 14% 12% 39%
Regulatory noncompliance risk 12% 15% 10% 37%
Erosion of customer trust from

AI failures
11% 11% 11% 33%
Ethical risks of AI 10% 12% 10% 32%

well placed: While any new technology has certain AI has also been used recently to create fake
vulnerabilities, the cyber-related liabilities sur- photos and videos of celebrities and politicians.
facing for certain AI technologies seem particularly While there are also techniques for identifying fakes,
vexing. it appears that technologies may fuel an arms race
Researchers have discovered that some ma- of fake image development versus detection. Given
chine-learning models have difficulty detecting the prominence of AI-based image recognition, this
adversarial input—that is, data constructed specifi- area is likely to be a cyber-risk battleground in the
cally to deceive the model. This is how one research future.
team fooled a vision algorithm into classifying as a There is evidence that cyber-risk concerns are
computer what appeared to be a picture of a cat.22 slowing or pausing AI projects at some compa-
The process of training machine-learning models nies. In addition, one in five respondents said they
can itself be manipulated with adversarial data. decided not to launch AI initiatives due to cyberse-
By intentionally feeding incorrect data into a self- curity worries (see figure 7).
learning facial recognition algorithm, for instance, Executives are commonly concerned about the
attackers can impersonate victims via biometric safety and reliability of AI systems as well. Forty-
authentication systems.23 In some cases, machine- three percent of respondents rated “making the
learning technology may expose a company to the wrong strategic decisions based on AI/cognitive rec-
risk of intellectual property theft. By automatically ommendations” as a top-three concern (see figure
generating large numbers of interactions with a 6). Nearly as many cited failure of an AI system in
machine-learning-based system and analyzing the a mission-critical or life-or-death situation. Placing
patterns of responses it generates, hackers could re- strategic decisions or mission-critical actions en-
verse-engineer the model or the training data itself. tirely in the hands of an AI system would certainly
11
FIGURE 7
Cybersecurity threats are giving some companies pause

Effect of cybersecurity concerns on companies
Moved ahead with AI initiatives despite cybersecurity concerns

36%
Experienced a cybersecurity breach relating to AI initiatives within the last two years
32%
Slowed an AI initiative in order to address cybersecurity concerns

30%
Decided not to start an AI initiative due to cybersecurity concerns

20%
Canceled or halted an in-progress AI initiative due to cybersecurity concerns

16%

entail special risks. Entrusting AI systems with such and operational risks associated with these systems.
responsibilities remains rare today, however. A Complicating matters are questions surrounding
prominent exception is the use of AI in autonomous who can be held liable in the event of an AI-related
vehicles: The technology has been implicated in crime or mishap. How liability is assigned in these
several accidents, some fatal, during testing.24 cases is a topic of ongoing discussion.26
Another element of cyber risk that companies Two themes are particularly salient when
should consider is how much data—and what it comes to AI and regulatory risk: privacy and
kind of data—they are willing to put into public explainability. Because data is so critical to AI, com-
cloud environments, allowing them to use cogni- panies seeking to apply the technology are often
tive technologies to analyze much larger data sets hungry for the stuff. Privacy regulations governing
than private clouds. Analysis of sensitive customer personal data may dampen their appetite, though:
and financial data can yield valuable insights, but The General Data Protection Regulation (GDPR),
companies should weigh the perceived risks with which has recently come into force in Europe, sets
the benefits. A recent Deloitte study found that the privacy rules that require careful implementation.
more experience organizations have with cloud GDPR also mandates that companies using per-
computing, the more comfortable they are putting sonal data to make automated decisions affecting
sensitive data into public clouds.25 people must be able to explain the logic behind the
decision-making process.27 Guidance published by
LEGAL AND REGULATORY RISKS the US Federal Reserve (SR 11-7) affects US banking
Products and systems of all types, including IT similarly: It requires that the behavior of computer
systems, present a range of legal and regulatory risks. models be explained.28 What makes these regula-
As a result, it is unsurprising that four in 10 survey tions challenging for some AI adopters is the growing
respondents indicate a high degree of concern complexity of machine learning and the increasing
about the legal and regulatory risks associated with popularity of deep-learning neural networks, which
AI systems. Because not all methods of validating can behave like black boxes, often generating highly
AI systems’ accuracy and performance are reliable, accurate results without an explanation of how
companies will need to manage the legal, regulatory, these results were computed. Many tech companies
12
and government agencies are pouring resources Some of the ethical risks that resonated with
into improving the “explainability” of deep-learning our respondents are linked to the aforementioned
neural networks.29 cyber-safety and regulatory issues: unintended
consequences, misuse of personal data, and lack of
ETHICS AND REPUTATION explanation for AI-powered decisions. But there is
For most of our respondents, ethical risks are one concern that has achieved special prominence
not a top-of-mind information technology concern. in recent years, and ranked second among our re-
And while ethical risks ranked at the bottom of risk spondents’ ranking of ethical risks: bias.
concerns in our survey, about a third of executives Today, algorithms are commonly used to help
did cite them as a top concern. make many important decisions, such as granting
In a deeper look at potential ethical risks, sur- credit, detecting crime, and assigning punishment.
veyed executives revealed a wide range of concerns. Biased algorithms, or machine-learning models
At the top of the list is AI’s power to help create or trained on biased data, can generate discriminatory
spread false information. This may be due to the or offensive results. For example, one study found
attention that social-media-driven “fake news” re- that ads for high-paying jobs were shown more
ceived in the 2016 US elections. often to men than to women.30
13
Early adopters want more

talent, and need a better
mix of it
D
O EARLY ADOPTERS have the talent to need more skilled people. Thirty percent said they
develop and deploy cognitive solutions? face a major (23 percent) or extreme (7 percent)
The overall survey results suggest both a skills gap. Another 39 percent said their gap is
considerable amount of talent already and a strong “moderate.” Interestingly, the most advanced com-
demand for more. “Lack of AI/cognitive skills” was panies in our survey feel the skills gap acutely.31 The
a top-three concern for 31 percent of respondents— limitations of their technical skills may be exposed
below such issues as implementation, integration, as they launch more AI solutions, and as those solu-
and data. A skills shortage was identified as the tions increase in complexity and scale.
biggest challenge in moving from prototypes to Some skills are needed more than others (see
full production deployments for only 8 percent figure 8).32 Respondents report the highest level of
of respondents. need for AI researchers to invent new kinds of AI al-
Companies generally feel that they have sub- gorithms and systems. This suggests an aggressive
stantial AI capabilities. About four in 10 executives level of ambition for the technology. In addition, 28
report their companies have a high level of sophis- percent said they need AI software developers, 24
tication in managing and maintaining AI solutions, percent need data scientists, and roughly similar
selecting AI technologies and technology suppliers, percentages need user-experience designers,
integrating AI technology into the existing IT en- change-management experts, project managers,
vironment, identifying valuable applications of AI, business leaders, and subject-matter experts. Sixty-
building AI solutions, and hiring and managing one percent are already training IT staff to deploy
technical staff with AI skills. An additional 41–46 AI/cognitive solutions, and 54 percent are training
percent say their companies have familiarity with developers to create new AI/cognitive solutions.
these activities. This suggests that they don’t have a Given the level of enthusiasm and aggressive
severe talent shortage. adoption of AI/cognitive that we have seen in re-
In addition to internal resources, many compa- sponses to other aspects of this survey, the findings
nies are pursuing broad approaches to the “talent on talent are perhaps unsurprising. They suggest
ecosystem.” Ten percent of our survey respondents that while talent is not the most important concern
said they get talent from companies they have ac- of the moment, there is an ongoing need to hire
quired, invested in, or partnered with. And as we’ve and train highly capable AI experts. Companies
seen, companies effectively outsource some of their with this kind of commitment to AI/cognitive tech-
talent needs by using either AI-as-a-service capa- nologies will probably be living with skills gaps for a
bilities, or crowdsourced development communities long time.
such as GitHub and Bitbucket. Although AI projects often flounder because the
Despite the sophistication of their internal teams relevant technology skills are in short supply, or-
and access to outside talent, executives feel they ganizations should recognize that success depends
14
FIGURE 8
Companies need a broad range of skills for their AI initiatives

Respondents rating each a top-2 needed skill to fill their company’s AI skills gap
AI researchers
30%
AI software developers
28%
Data scientists
24%
User experience designers

23%
Change management/transformation experts

22%
Project managers
22%
Business leaders to interpret AI results

21%
Subject-matter experts
20%
Note: Base = those who said that their company has moderate/major/extreme skills gap in meeting the needs of AI/
cognitive projects. Sample size = 752.
on more than technology talent. For example, data through automation” received the lowest ranking
scientists often struggle when they aren’t clear on among the list of “primary benefits from AI/cog-
the business problem they’re supposed to solve. The nitive technology” choices, with 24 percent of
result can be AI projects that go nowhere, and disil- respondents rating it among their top three choices.
lusioned data scientists who defect to competitors. 33
That said, there is evidence that many compa-
Subject-matter experts who can “speak data” to data nies plan to automate tasks and cut jobs. Sixty-three
scientists while “speaking business” to executives percent of respondents agreed with the following
can be valuable, yet only 20 percent of companies in statement: “To cut costs, my company wants to au-
our survey said they were needed. And given their tomate as many jobs as possible with AI/cognitive.”
struggles to implement AI solutions and manage In many companies we have observed, the business
projects, it’s surprising that only 22 percent stated case for some cognitive projects, such as chatbots,
the need for each of these skills. relies heavily on using AI to replace workers.
While our survey did not directly address the
magnitude of job losses due to AI, 36 percent of
The impact on jobs
respondents feel that job cuts from AI-driven auto-
Companies in our survey are searching for the mation rise to the level of an ethical risk. In some
right balance between using AI to automate tasks industries, such as financial services, executives
(cutting costs—and jobs) and to augment the ca- have been candid about their plans to automate tens
pabilities of its workforce. AI-driven automation is of thousands of jobs in the next few years.34 Perhaps
not seen as a top benefit of AI. “Reduce headcount some losses have already happened. The number is
15
likely to rise as cognitive technologies become more will make the remaining and newly added workers
firmly entrenched. better at their jobs, and happier at the same time.35
The threats to existing workers are not only from Seventy-eight percent of survey respondents agree
automation-related job loss. The great majority of that AI/cognitive technologies empower people to
survey respondents agree that AI leads to either make better decisions, and 72 percent believe that
moderate or substantial changes in job roles and AI will increase job satisfaction. Perhaps the biggest
skills, both already (72 percent) and in three years advantage could be new ways of working that
(82 percent). But perhaps of greatest concern for blend the best of what machines do with human
the currently employed should be executives’ pref- experience, judgment, and empathy; 78 percent of
erence for new hires with the required skills over executives believe that AI-based augmentation of
retraining and retaining current workers. Only 10 workers will fuel new ways of working.
percent of respondents stated a clear preference for This is no time for American workers to be com-
retraining and keeping current employees. Eighty placent. While automation is not a top priority for
percent leaned toward either “keeping or replacing many companies, it still looms as a growing threat.
employees in equal measure” or “primarily replace In addition, many companies are looking to acquire
current employees with new talent.” new AI-related skills with outside talent.
Despite the threats to current employees due to
AI, executives believe that cognitive technologies
16
Enthusiastic early adopters can

take the next step by getting
serious
F
OR THE SECOND year in a row, we’ve seen that • Fifty-four percent have a process for moving
early adopters are using cognitive technologies prototypes into production;
to effect positive change for their companies. • Fifty-two percent have an implementation road
Overall, their reactions to this new set of tools are map;
remarkably bullish. Though they face challenges, • Forty-five percent have appointed senior execu-
many of the companies we surveyed are having tives across the company as AI champions;
early success integrating AI within their operations • Thirty-seven percent have created a company-
and customer relationships—and getting economic wide center of excellence; and
benefits. They are enthusiastic about their successes • Thirty-seven percent have a comprehensive
to date, and about the potential for these technolo- strategy for AI in place.
gies to transform their companies in the near future. However, to ensure that cognitive remains a
We believe the excitement early adopters express top priority once the hype subsides, leaders of AI
about AI is warranted. We also think that early initiatives should ensure that costs and impacts are
adopters—and companies that want to emulate tracked carefully, and that successes are incontro-
them—could have a surer path to success if they vertible. This will help CFOs make the investments
take the following steps. required as projects—and budgets—get bigger.
Pursue execution excellence Address cybersecurity risks

Early adopters should combine their experimen- The problems that early adopters have with cy-
tation and industrious—even frenetic—activity with bersecurity help to make their execution problems
better operational discipline. Despite its complexity clear. Less than half are building cybersecurity
and transformational potential, AI’s implementa- into their AI projects. This despite the fact that
tion resembles that of other technologies. To drive cybersecurity is the top risk cited by executives
change across lines of business, companies should in our survey: They fear that both the algorithms
focus on project management and change manage- that deliver insights, and the data that fuels those
ment. The fundamentals of fostering organizational algorithms, are vulnerable to attack. And recall that
change can get lost amid excitement around pilots, respondents’ biggest concern about the ethics of AI
grassroots experiments, and vendor-driven hype. is the falsification of images and bots that create
Some management infrastructure around AI is “fake news”—both of which are easier to propagate
being created; our survey results reveal the fol- when cybersecurity preparedness is low. The worst-
lowing indicators of structures and processes to case scenario—such as autonomous vehicles getting
improve execution: hacked—can have life-threatening ramifications.
17
We’ve seen that cybersecurity is having a nega- Buy (some) off the shelf
tive impact on some early adopters—32 percent have
suffered an AI-related breach. Some companies are
also slowing down or halting AI initiatives due to Enterprise software and cloud services give
cybersecurity concerns. Others are forging ahead companies expanding options for adopting cogni-
despite them. Neither is ideal: One leads to slower tive, without taking the tortuous path of having to
implementation and reduced competitiveness, the build everything from scratch. Although cognitive
other to unnecessary risks. No cybersecurity efforts technologies are still evolving, this evolution is
can prevent every attack, but early adopters can happening at a breakneck pace. Cloud-based CRM
improve their defenses by incorporating security and ERP software with cognitive capabilities are
from the beginning of the process, and making it a widely available, as are chatbots. Many big cloud
higher priority. providers are developing subscription-based AI
Advances are being made to reduce risks associ- services aimed at specific business functions. This
ated with AI—for instance, forensic technology is may be the easiest path to getting the benefits of AI
getting better at detecting manipulated images and into functions such as product design and sales and
videos, known as “deep fakes.”36 The explainability marketing.38
of deep-learning models will likely also improve, For companies that want to develop their
helping companies to avoid regulatory noncom- own solutions, tools such as automated machine
pliance and other risks associated with bias in learning can also boost the capabilities of “ordinary”
algorithms. Companies should stay on top of these programmers. AWS wants to democratize AI so that
developments and incorporate them when they are programmers without specific AI training can use
proven. it. Google and the startup DataRobot have similar
ambitions with their automated machine-learning
offerings. Surely, companies need expertise within
Apply AI beyond the
their “four walls,” but they should examine which
IT function capabilities they can get from enterprise software
The top three cognitive use cases in our study—IT and cloud-based platforms. This can lead to quick
automation, quality control, and cybersecurity—are wins, lower initial investment, and momentum.
largely focused on IT (see figure 9). These are im-
portant use cases—especially cybersecurity (in this
Staff wisely
case, tackling cybersecurity with AI). IT automa-
tion is showing early promise, according to some There are many potential holes in any com-
studies. 37
pany’s AI capabilities. But focusing only upon the
It makes sense that complex technologies, which hardest talent to attract and retain—AI researchers
require heavy involvement from the IT department, and programmers, and data scientists—may not
would be applied there first. But the transformative be the best strategy, especially for companies just
potential of AI will likely be reached only when it starting out. AI newcomers may want to see how
permeates an entire company and enables change far they can get using off-the-shelf solutions and
in multiple business functions and units. Cloud can cloud platforms. Partners and consultants can also
play a pivotal role in achieving those objectives via provide much-needed expertise and guidance, and
services that provide broad ranges of users with most of our early adopters are using them.
easy access to AI-based capabilities. Companies also need sound strategies for talent
development and acquisition. While many early
adopters say they are training their employees for
the new roles and skills AI requires, they prefer to
18
FIGURE 9
Top three AI use cases focus on IT

Respondents that report their company is focusing on these use cases with AI
IT automation
47%
Quality control/detecting defects

46%
Cybersecurity
41%
Predictive analytics
38%
Customer service (including virtual assistants)

37%
Risk management
37%
Sales optimization
34%
Decision support
34%
Workforce management
32%
Marketing optimization
30%
Connected equipment, devices, products

29%
Forecasting
29%
Tax, audit, and compliance

23%

hire new workers from outside their organizations. in-house, companies should consider the mix of
Both approaches may be required, especially for talent needed to manage AI projects successfully.
scarce technical skills. We believe that there is more Focusing too much on acquiring high-cost, scarce
potential for retraining current employees to work talent that the tech giants are scooping up in a
alongside smart machines than the survey respon- bitter arms race39 may lead to frustration and dis-
dents seem to, although it’s important to start early appointment. Companies that are developing their
with such initiatives. own bespoke cognitive solutions need significant
But when considering which AI development technical talent. But they also likely need business
and implementation skills a company should have executives who can “speak AI” to data scientists
19
and understand the uses and limitations of data they need to recruit—and cut. It’s naïve to think that
analytics. AI won’t cost jobs, and some CEOs are becoming
more frank about admitting this.40 However, ex-
clusive focus on automation and cost-cutting could
Decide where to automate
hinder chances to use AI for transformative efforts
and where to augment that leverage the best of artificial and human intel-
Companies that automate simply to cut costs or ligence.41 It could also fuel distrust and fear among
improve efficiency are not taking full advantage of employees who may be waiting for the other shoe
AI. Yet there are clear use cases where automation to drop.
is simply better and more efficient than humans.
In these use cases, machines will likely eventually
Conclusion
replace people altogether. In many more instances,
machines will surface information, make pre- Our survey results clearly show that growing
dictions, and offer alternatives. Humans, using numbers of companies are becoming more sophis-
judgment, empathy, and business skill, should apply ticated in their usage of AI technologies. Now is the
this information to best effect. This is a matter not time for organizations to start selecting the business
simply of placing humans in the loop but of the loop use cases that can deliver measurable value through
being built to augment human decision-making. AI-powered capabilities. By using cloud services
Knowing where companies want “automation as a gateway, it’s never been easier to explore and
to replace” and where they want “intelligence to access AI’s potential—with minimal up-front invest-
augment” will likely help them be clear about how ment and a reduced need for in-house expertise.
they will change operations, and what kind of people
20
Endnotes
1. It would take a separate article to delve into the differences between cognitive technologies and artificial intel-
ligence. In general, they are considered as synonymous in the market. Therefore, we use the terms cognitive,
cognitive technologies, artificial intelligence, and AI interchangeably in this study.
2. All survey respondents can be considered early adopters when compared with their counterparts in an average
company. All of the respondents’ companies have implemented at least one cognitive prototype or full-scale
implementation. In addition, 75 percent of respondents say they have an “excellent” understanding of AI or
are experts.
3. In addition, 56 percent of the respondents represent companies with 500 to 5,000 employees; 44 percent come
from companies with more than 5,000 employees. In terms of annual revenue, 16 percent of the surveyed
companies are over US$5 billion, 36 percent fall between US$1 billion and US$5 billion, 31 percent range from
US$500 million to US$1 billion, and 17 percent range from US$250 million to US$500 million.
4. Thomas H. Davenport, Jeff Loucks, and David Schatsky, Bullish on the business value of cognitive: Leaders in cogni-
tive and AI weigh in on what’s working and what’s next, Deloitte, 2017.
5. A 2018 study showed that 22 percent of companies are using machine learning, and 14 percent are using natural
language processing/generation. Respondents in our study have much higher adoption rates: 63 percent and
62 percent, respectively. See Narrative Science, Outlook on artificial intelligence in the enterprise 2018, accessed
October 1, 2018.
6. Marty Swant, “Google’s new voice-activated analytics fueled by AI will simplify data queries,” AdWeek, July 18,
2017.
7. Computer Vision Machine Learning Team, “An on-device deep neural network for face detection,” Apple Machine
Learning Journal 1, no. 7 (2017).
8. Arren Alexander, “Computer vision case study: Amazon Go,” Medium, April 2, 2018.
9. It bears repeating that we do not claim that all companies are adopting cognitive technologies at the rate of our
early adopters.
10. The wide availability of AI-infused enterprise software is not a surprise. In a 2015 paper, Deloitte noted that most
AI-related venture capital investment was directed at companies targeting traditional enterprise issues. We said,
“In the near term, though, established industries can find significant opportunities in applying these technologies
to conventional business functions.” See David Schatsky and Ragu Gurumurthy, Cognitive technologies: Applying
machine intelligence to traditional business problems, Deloitte University Press, October 6, 2015.
11. Kyle Wiggers, “Salesforce develops natural language processing model that performs 10 tasks at once,” Venture-
Beat, June 20, 2018.
12. Larry Dignan, “Google Cloud rolls out prepackaged AI services aimed at business functions,” ZDNet, August 16,
2018.
13. Gillian Crossan, et al., Accelerating agility with XaaS, Deloitte Insights, September 17, 2018.
14. “The Artificial Intelligence (AI) as a service market is expected to grow at a Compound Annual Growth Rate (CAGR)
of 48.2 percent,” PR Newswire, April 26, 2018.
15. IDC, “Worldwide spending on cognitive and artificial intelligence systems will grow to $19.1 billion in 2018,” March
22, 2018.
16. CB Insights, “AI 100: The artificial intelligence startups redefining industries,” December 12, 2017.
21
17. Jyoti Nigania, “Machine learning improving the quality at Netflix,” House of Bots, August 22, 2018.
18. Keith Shaw, “What is digital twin technology?” Network World, June 11, 2018; see also Chris Middleton, “Digital
twin AIs designed to learn at the edge,” Control Engineering, June 10, 2018.
19. David Raths, “JASON report: Hype about AI in healthcare might be justified this time,” Healthcare Informatics,
January 22, 2018.
20. Jessica Kent, “How artificial intelligence is changing radiology, pathology,” HealthITAnalytics, August 3, 2018.
21. Jim Gusczca, “Smarter together: Why artificial intelligence needs human-centered design,” Deloitte Review 22,
January 22, 2018.
22. Gamaleldin F. Elsayed et al., “Adversarial examples that fool both computer vision and time-limited humans,”
arXiv.org, February 2018.
23. Dilip Krishna, Nancy Albinson, and Yang Chu, Managing algorithmic risks: Safeguarding the use of complex algo-
rithms and machine learning, Deloitte, 2017.
24. Timothy Carone, “Self-driving car accidents will keep happening. We need to learn from them,” CNN, March 21,
2018; Michael Hiltzik, “Self-driving car deaths raise the question: Is society ready for us to take our hands off the
wheel?”, LA Times, April 3, 2018.
25. Crossan et al., Accelerating agility with XaaS.
26. For a useful overview of this topic, see John Kingston, “Artificial intelligence and legal liability,” in Max Bramer and
Miltos Petridis, eds., Research and Development in Intelligent Systems XXXIII: Incorporating Applications and Innova-
tions in Intelligent Systems XXIV (Cambridge, UK: Springer Verlag, 2016).
27. J. M. Porup, “What does the GDPR and the ‘right to explanation’ mean for AI?,” CSO, February 9, 2018. For more
on GDPR, see Deloitte, The General Data Protection Regulation, 2018.
28. Seph Mard, “What’s model risk, and why does it matter?,” DataRobot Blog, January 16, 2018.
29. Jason Bloomberg, “Don’t trust AI? Time to open the black box,” Forbes, September 16, 2018.
30. Osonde Osoba, “Keeping artificial intelligence accountable to humans,” TechCrunch, August 20, 2018.
31. We analyzed companies in our study based on the number of implementations and their sophistication in imple-
menting and managing AI. We found that the segment that was the most sophisticated and had built the most
production AI systems report a more acute skills gap.
32. The following are full descriptions of the skills/capabilities listed in figure 8:
• AI researchers—to invent new kinds of AI/cognitive algorithms and systems
• Software developers—to build AI/cognitive systems
• Data scientists—to analyze and extract meaningful insights from data
• User experience designers—to improve the user experiences of our AI/cognitive-related products/services
• Change management/transformation experts—to implement strategies for effecting change, and to help people
adapt to cognitive/AI
• Project managers—to plan and execute AI/cognitive projects
• Business leaders—to interpret AI/cognitive results, make decisions, and take appropriate actions
• Subject-matter experts—to infuse their area of expertise into our AI/cognitive systems
22
33. Daniel Byler and Jeff Loucks, “How do you retain your data scientists? Keeping tech talent through a post-merger
integration,” Deloitte, May 31, 2018.
34. Laura Noonan, “Citi issues stark warning on automation of bank jobs,” Financial Times, June 11, 2018.
35. According to a recent study by the World Economic Forum, developments in automation technologies and arti-
ficial intelligence could create 58 million net new jobs by 2022. See World Economic Forum, “The future of jobs
2018,” accessed October 1, 2018.
36. Steven Melendez, “Can new forensic tech win war on AI-generated fake images?,” Fast Company, April 4, 2018.
37. Cliff Saran, “IT set to gain the most from AI and automation,” ComputerWeekly, April 3, 2018.
38. Ravi Akella, “What generative design is and why it’s the future of manufacturing,” New Equipment Digest, March
16, 2018.
39. Cade Metz, “Tech Giants Are Paying Huge Salaries for Scarce A.I. Talent,” New York Times, October 22, 2017.
40. Adam McCulloch, “Finance leaders say AI will soon replace half of banking sector’s jobs,” Personnel Today, August
24, 2018.
41. Thomas H. Davenport and Julia Kirby, Only Humans Need Apply: Winners and Losers in the Age of Smart Machines
(New York: HarperBusiness, 2016).
23
About the authors
JEFF LOUCKS is the executive director of Deloitte’s Center for Technology, Media & Telecommunications,
Deloitte Services LP. In his role, he conducts research and writes on topics that help companies capitalize
on technological change. An award-winning thought leader in digital business model transformation,
Loucks is especially interested in the strategies organizations use to adapt to accelerating change.
Louck’s academic background complements his technology expertise: He has a bachelor of arts in
political science from The Ohio State University and a master of arts and PhD in political science from
the University of Toronto.
TOM DAVENPORT is the President’s Distinguished Professor of Information Technology and

Management at Babson College, the cofounder of the International Institute for Analytics, and a fellow
of the MIT Center for Digital Business. He is an independent senior advisor to Deloitte Analytics, Deloitte
Consulting LLP. He collaborates with Deloitte thought leaders on all things related to business analytics,
from the potential of cognitive technologies to industry-focused explorations and outcomes. Covering
topics from emerging technologies to innovative business applications, Davenport’s Deloitte Insights
series reveals leading-edge thinking on analytics and cognitive technology.
DAVID SCHATSKY analyzes emerging technology and business trends for Deloitte’s leaders and
clients. His recent published works include Signals for Strategists: Sensing Emerging Trends in Business
and Technology (Rosetta Books 2015), Demystifying artificial intelligence: What business leaders need to
know about cognitive technologies, and Cognitive technologies: The real opportunities for business (Deloitte
University Press, 2014–15). Before joining Deloitte, Schatsky led two research and advisory firms.
Acknowledgments
The authors would like to thank Susanne Hupfer for her work on the State of AI in the Enterprise, 2nd
Edition. We could not have done it without your tireless and thoughtful contributions from the early
stages of questionnaire development through publication. Thank you to Scott Pobiner and Brijesh
Singh for your subject-matter expertise and thoughtful insights, Sayantani Mazumder for your support
with data analysis, and Karthik Ramachandran and Gaurav Khetan for your contributions to sec-
ondary research and analysis.
The authors would like to thank Kerry Iseman, Karen Petty, Alyssa Pharr, and Lara Warthen of
Deloitte Services LP for their marketing leadership and ongoing support of the survey; Alicia DiGennaro
and Anisha Sharma for their support with external communications; and Wendy Berk and Debra Floyd
for keeping us all aligned, on deadline, and in sync with our Green Dot agency.
24
Contacts
David Kuder Dave Rudini
Principal Chief analytics officer
Deloitte Consulting LLP Deloitte Analytics
+1 404 631 2300 Principal
dkuder@deloitte.com Deloitte Consulting LLP
+ 1 212 618 4686
Mic Locker drudini@deloitte.com
US Technology, Media & Telecommunications
Cognitive Advantage leader Paul Sallomi
Managing director US Technology Sector and Global
Deloitte Consulting LLP Technology, Media & Tele-
+1 203 423 4727 communications leader
miclocker@deloitte.com Partner
Deloitte Tax LLP
Dr. Jeff Loucks +1 408 704 4100
Executive director psallomi@deloitte.com
Deloitte Center for Technology,
Media & Telecommunications David Schatsky
Deloitte Services LLP US Innovation
+1 614 228 4285 Managing director
jloucks@deloitte.com Deloitte Consulting LLP
+1 646 582 5209
Nitin Mittal dschatsky@deloitte.com
Deloitte Analytics & Cognitive Offering leader
Principal
Deloitte Consulting LLP
+1 617 831 4128
nmittal@deloitte.com
25
Sign up for Deloitte Insights updates at www.deloitte.com/insights.
Follow @DeloitteInsight
Deloitte Insights contributors

Editorial: Matthew Budman, Blythe Hurley, and Rupesh Bhat
Creative: Kevin Weier, Anoop R, and Emily Moreano
Promotion: Nikita Garia and Amy Bergstrom
Cover artwork: Daniel Hertzberg
About Deloitte Insights

Deloitte Insights publishes original articles, reports and periodicals that provide insights for businesses, the public sector and
NGOs. Our goal is to draw upon research and experience from throughout our professional services organization, and that of
coauthors in academia and business, to advance the conversation on a broad spectrum of topics of interest to executives and
government leaders.
Deloitte Insights is an imprint of Deloitte Development LLC.
About this publication

This publication contains general information only, and none of Deloitte Touche Tohmatsu Limited, its member firms, or its and
their affiliates are, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other profes-
sional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a
basis for any decision or action that may affect your finances or your business. Before making any decision or taking any action
that may affect your finances or your business, you should consult a qualified professional adviser.
None of Deloitte Touche Tohmatsu Limited, its member firms, or its and their respective affiliates shall be responsible for any
loss whatsoever sustained by any person who relies on this publication.
About Deloitte
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its
network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent
entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. In the United States, Deloitte refers to
one or more of the US member firms of DTTL, their related entities that operate using the “Deloitte” name in the United States
and their respective affiliates. Certain services may not be available to attest clients under the rules and regulations of public
accounting. Please see www.deloitte.com/about to learn more about our global network of member firms.
Copyright © 2018 Deloitte Development LLC. All rights reserved.

Member of Deloitte Touche Tohmatsu Limited

DI State of AI in The Enterprise-2nd-Ed

Uploaded by

Copyright:

Available Formats

DI State of AI in The Enterprise-2nd-Ed

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DI State of AI in The Enterprise-2nd-Ed

Uploaded by

Copyright:

Available Formats

State of AI in the Enterprise, 2nd Edition

Early adopters combine bullish enthusiasm with strategic investments

Activity, investment, and positive results | 3

To maximize value, early adopters should become

Early adopters want more talent, and need a

Enthusiastic early adopters can take the next step

Activity, investment, and

AI helps organizations keep up with the (Dow) Joneses

16% 20% 27% 28% 9%

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

Everyone’s winning, but some industries are winning bigger

22% Low AI investment/high returns High AI investment/high returns

16% Financial services and insurance

Low AI investment/low returns High AI investment/low returns

AI’s leading beneﬁts are enhanced products and processes—

Enhance current products

Optimize internal operations

Make better decisions

Optimize external operations

Free workers to be more creative 55%

Create new products

Capture and apply scarce knowledge

Reduce headcount through automation

Pursue new markets

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

To maximize value, early

Many early adopters struggle with the basics

Ranked 1 Ranked 2 Ranked 3 Ranked top three

Implementation challenges 13% 14% 12% 39%

Lack of skills 11% 10% 10% 31%

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

Twenty-three percent of respondents often needed before it is

ranked “cybersecurity vulnerabilities” as a proof of concept. We

Cybersecurity heads the lists of AI-related concerns

Ranked 1 Ranked 2 Ranked 3 Ranked top three

Regulatory noncompliance risk 12% 15% 10% 37%

Erosion of customer trust from

Ethical risks of AI 10% 12% 10% 32%

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

Cybersecurity threats are giving some companies pause

Moved ahead with AI initiatives despite cybersecurity concerns

Slowed an AI initiative in order to address cybersecurity concerns

Decided not to start an AI initiative due to cybersecurity concerns

Canceled or halted an in-progress AI initiative due to cybersecurity concerns

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

Early adopters want more

Companies need a broad range of skills for their AI initiatives

User experience designers

Change management/transformation experts

Business leaders to interpret AI results

Enthusiastic early adopters can

Pursue execution excellence Address cybersecurity risks

Top three AI use cases focus on IT

Quality control/detecting defects

Customer service (including virtual assistants)

Connected equipment, devices, products

Tax, audit, and compliance

Source: Deloitte State of AI in the Enterprise, 2nd Edition, 2018.

25. Crossan et al., Accelerating agility with XaaS.