Artificial intelligence in marketing: Systematic review &

future research direction

Author:- Nisarg Darji

Abstract

Technologies such as the internet of things, big data analytics, blockchain, and artificial
intelligence have changed the ways businesses operate. Of all the disruptive technologies,
artificial intelligence (AI) is the latest technological disruptor and holds immense marketing
transformation potential. Practitioners worldwide are trying to figure out the best fit AI
solutions for their marketing functions. However, a systematic literature review
can highlight the importance of artificial intelligence (AI) in marketing and chart future
research directions. The present study aims to offer a comprehensive review of AI in
marketing using bibliometric, conceptual and intellectual network analysis of extant literature
published between 1982 and 2020. A comprehensive review of one thousand five hundred
and eighty papers helped to identify the scientific actors’ performance like most relevant
authors and most relevant sources. Furthermore, co-citation and co-occurrence analysis
offered the conceptual and intellectual network. Data clustering using the Louvain algorithm
helped identify research sub-themes and future research directions to expand AI in marketing

Keywords: Marketing, Artificial intelligence, Bibliometric analysis, Intellectual

structure Conceptual structure.

Introduction
Technological disruptions such as artificial intelligence (AI), internet of things (IoT), big data
analytics (BDA) have offered digital solutions for attracting and maintaining the customer base
(Anshari, Almunawar, Lim, & Al-Mudimigh, 2018; Bolton et al., 2018). Emerging technologies
provide a competitive advantage (Rouhani et al., 2016; Spring et al., 2017) by facilitating the
customers’ product and service offerings (Balaji & Roy, 2017; Khanagha et al., 2017; Liao, 2015). In
the current business scenario, the cut-throat competition and technological disruptions have changed
the way organizations operate (Gans, 2016).Globally customer-centric approach focused on customer
needs plays a pivotal role in organizational growth (Vetterli, Uebernickel, Brenner,
Petrie, & Stermann, 2016). Artificial intelligence (AI) is a widely used emerging technology that
helps organizations track real-time data to analyze and respond swiftly to customer requirements
(Wirth, 2018). AI offers consumer insight on consumer behavior essential for customer attraction and
customer retention. AI incites the customer’s next move and redefines the overall experience
(Tjepkema, 2019). AI tools are useful to deduce customer expectations and navigate the future path
(Shabbir,2015) Artificial Intelligence find its applications in different context in today’s business
scenario. Practitioners and academicians believe that Artificial Intelligence is the future of our society.
With the advancement of technology, the world has become a web of interconnected networks.
The technology implementation lead to investment in Artificial Intelligence (AI) for big data analytics
to generate market intelligence. Artificial Intelligence applications are not limited to only marketing;
rather, it is widely used in other sectors such as medical, e-commerce business,
education, law, and manufacturing. AI is continuously getting implemented to benefit many different
industries. As the organizations move forward towards Industry 4.0, Artificial Intelligence & other
emerging technologies are also evolving parallelly. However, the implementation of AI in all sectors
has not been possible due to many constraints, but scientists are working on systems that cater to the
theory of mind and self-awareness of the artificially intelligent systems.Nowadays the people interact
with some form of AI in daily activities. For example, the user enjoys the automatic e-mail filtering
feature. In the smartphone, the user may probably fill out a calendar with Siri, Cortana, or Bixby. The
user of the new vehicle gets assisted by a while driving. Artificial Intelligence can automate the
business process, learn insights from past data, and generate consumer and market insights through
the program-based algorithm (Davenport et al 2020).Technologies like Machine Learning (ML), Deep
Learning, and Natural Language Processing (NLP) train machines to handle big data for the
generation of market intelligence (Davenport et al 2020). As the adoption of AI in marketing is in
nascent stages, there is a dearth of systematic literature review exhibiting the in-depth research pattern
in the
AI-driven consumer market and lead to research questions like
RQ1: What are the applications of AI in Marketing?
RQ2: How marketing can optimally utilize the AI technologies for
maximizing customer satisfaction, market share and profitability?
RQ3: What are the trending topics and future research directions for
the adoption of AI in Marketing?
This paper attempts to fill that research gap through a systematic review of literature on AI in the
marketing research domain. Bibliometric analysis of more than 1500 articles (published between 1982
and 2020) offered scientific actors’ performance like most relevant authors, most
relevant sources etc. Co-citation and co-occurrence analysis based on the Louvain algorithm helped to
map the research domain’s conceptual and intellectual structure. In the subsequent sections, literature
review, research methodology, findings, discussion and conclusion are presented.

2. Literature review

Unlike human intelligence, artificial intelligence (AI) is the intelligence demonstrated by the
machines. A system of intelligent agent machines that perceives the environment to successfully
achieve its goal represent the artificial intelligence. According to Russel and
Norvig (2016), artificial intelligence describes machines (computers) that simulate cognitive and
affective functions of human mind. The development of Artificial intelligence is phenomenal and
experts have worked tirelessly to advance AI concepts over the few decades. The work led to some
major innovations like big data analytics and machine learning applications in myriad sectors and
context. The term Artificial Intelligent generally leads people to think only
about automated robots who work for humans because the people have only seen the human-machine
interaction in the movies or any shows through robots only. Artificial Intelligence applies to any kind
the machine that needs to think like a human resulting in continuous learning
and problem-solving. These are the features of AI that make it unique. Sometimes people find a task
boring or dull, which is repetitive. However, with the help of a machine, people never have to
experience a similar job as boring. An artificially intelligent system does repetitive
jobs for humans continuously. Data ingestion is a very important feature of artificial intelligence.
Artificially intelligent systems work with huge amounts of data. The artificial intelligence system
collects data according to requirement and analyze the huge chunks of data. There is a huge amount of
data that organizations like google, amazon handle & that is impossible to analyze by humans. Also,
an artificially intelligent system stores multiple information about multiple people, multiple machines
from multiple sources. All of this appears on the system asynchronously or simultaneously.
AI-enabled systems are designed to observe and react to their surroundings. They perceive the
environment and take actions accordingly and keep in mind the situations that might come up soon.
For example, AI, with the help of historical data, can predict the breakdown time of
a machine. It can alert us for the action beforehand.

2.1. Advantage of machine learning over other technologies

Many technologies may do repeated work, but they can’t think independently. They lack to think
outside their code. On the contrary,
Machine learning is a subset of AI that aims to give machines the ability
to learn a task without pre-existing code. Here machines are fed with
some of the problems and examples through which machines learn for
certain tasks. As they go through these problems & examples, machines
learn and adapt their strategy to independently execute the activities.
For example, an image-recognition machine may be given millions of
pictures to analyze. After going through endless permutations, the machine acquires the ability to
recognize patterns, shapes, faces, and more.
In today’s scenario, the machine is only learning for the specific
repeated task, but the machines are trained to learn more than just a
specific task. The AI experts are working to make the machine able to
take what it has learned from analyzing photographs and using that
knowledge to analyze different data sets. Data scientists & programmers
are formulating general-purpose learning algorithms that help machines
learn more than just one specific task

2.2. Principle behind working of artificial intelligence

Artificial intelligence is that human intelligence can be transferred

to machines to execute tasks from the simplest to the most complex. The
objective of artificial intelligence is to learn, do reasoning & execute the
activities. As technology moves forward, previous standards that explain
artificial intelligence become outdated. There are three basic concepts
behind Artificial Intelligence. These basic concepts are machine learning, deep learning, and neural
networks. These concepts are leading to
further development of data mining, natural language processing, and
driving software. While AI and machine learning may seem like interchangeable terms, AI is usually
considered the broader term, with machine learning and the other two AI concepts a subset of it.
The mechanism of Deep learning is based on the principle of artificial neural networks. It imitates
neurons or brain cells. Artificial neural
networks were inspired by things we find in our biology. The neural net
models use math and computer science principles to mimic the human
brain processes, allowing for more learning & command to act. An artificial neural network integrates
the processes of densely interconnected
brain cells, but instead of being built from biology, these neurons, or
nodes, are built from human-made code.
Neural networks contain three layers: an input layer, a hidden layer,
and an output layer. These layers contain thousands, sometimes million
nodes. AI imitates human minds through the concepts of neural networks. It thinks the way a human
thinks & acts accordingly to solve
the problems. This is the uniqueness of AI. AI imitates human brain to
interpret the environment & act accordingly

2.3. Use of artificial intelligence in marketing

The authors undertook the literature review to comprehend the extent of research on enhancing
customer experiences through AI. Gacanin
and Wagner (2019) described the implementation challenges of autonomous customer experience
management (CEM). The paper also narrated how the intelligence network and critical business value
driver were established through AI and ML. Customer experience improved
through AI-driven chatbot with Natural Language Processing (NLP) (Nguyen and Sidorova, 2018).
AI and ML algorithms enabled efficient data processing, which allows us to formulate the correct
decision (Maxwell et al., 2011). Application of AI is required to analyze customer
habits, purchases, likings, disliking, etc. (Chatterjee et al., 2019). Customer Relationship Management
(CRM) functions benefited through Artificial Intelligence User Interface (AIUI) (Seranmadevi &
Kumar, 2019). AI and IoT converted traditional retail stores to smart retail stores.
The smart retail stores elevated customer experience and ease of shopping, and better supply chain
(Sujata et al., 2019). Besides brick-and mortar stores, AI guides the online businesses also. Sha and
Rajeswari
(2019) described the advancement of AI and demonstrated the AIsupported machine, which can track
the five senses (sight, hearing, taste,
smell and touch) of humans. The result showed a better consumer-brand
association and product-brand association in the e-commerce business.
A summary of some of the interesting research studies is presented in Table 1

3. Methodology

We used Rowley and Slack’s (2004) guidelines for conducting the literature review.
Methodologically, the literature review used a five-stage process described in the following sections.
Comprehensive review protocols helped in the identification of research themes and future research
directions.

3.1. Selection of bibliometric databases

Scopus and Web of Science (WoS) are the two most reputed bibliometric databases. We explored
both Scopus and Web of Science (WoS) databases to search the relevant literature. According to
Yong-Hak (2013), Scopus had broader coverage, and it includes more than 20,000
peer-reviewed journalsfrom different publishers (Fahimnia et al., 2015). Due to its wider coverage, we
preferred Scopus for data collection. Scopus offered advanced search filters and data analysis grids for
better data management

3.2. Defining keywords (search strategy)

The initial search string included words like “marketing” and “artificial intelligence.” Synonyms used
for artificial intelligence like machine learning, deep learning, natural language processing, etc., are
used with boolean operators like “OR” to get the universal set of papers. Boolean operator “AND” is
used to get the intersection set of paper covering marketing and artificial intelligence.
3.3. Refining the initial results (Inclusion and exclusion criteria)

Inclusion and exclusion criteria are applied to the search results.

With the help of inclusion and exclusion criteria, delimitation helped in
the extraction of the most relevant articles for the literature review. To
achieve the research objective, the search results limits to only articles
published in journals as they represent “certified knowledge” (RamosRodríguez and Ruiz-Navarro,
2004). Conference papers, book chapters,
commentaries, erratum etc., were excluded from the search results.

3.4. Data analysis plan

The bibliometric analysis of data was carried out using R-software for
performance analysis of scientific actors like most relevant authors and
most relevant sources. The content analysis and performance analysis
of each scientific actor offered the intellectual structure of the research
domain. Two researchers analyzed the Scopus data for inter-rater validity.
Data analysis is structured in three stages. Stage 1 data analysis focused on scientific actors’
performance like most relevant sources and
most relevant authors in the research domain. Bibliometric analysis
based on total citations and citation index helped in the performance
evaluation of scientific actors. Stage 2 analysis used co-occurrence and
co-citation analysis for conceptual and intellectual network analysis. According to Chen et al. (2010),
research papers’ co-citation network indicates the intellectual structure, concepts co-citation
network indicates
conceptual structure, and the authors’ co-citation network indicates the
social structure of the research domain. Stage 3 analysis focused on
emerging trends and future research directions of AI in Marketing.
3.5. Identification of research gaps and future research directions

The articles relating to artificial intelligence in marketing were reviewed to understand the
theoretical evolution, methodological evolution, and emerging research themes. Thematic coding is
used for the qualitative analysis of data. Thematic coding is a form of qualitative
analysis that involves recording or identifying passages of text or images
linked by a common theme or idea, allowing data to index into categories for the development of
the thematic framework (Gibbs, 2007) An in-depth review of research papers in each theme offered
research gap insights and helped chart future research directions. Research gaps
are translated into research questions that future researchers can embark on to solve.

4. Findings

4.1. Descriptive statistics of bibliographic collection

A total of 1580 documents (1523 articles and 57 reviews) have been

published to date on this specific topic in 710 numbers of prominent
journals. 5780 number of keywords have been used to date in this topic,
and authors have used 5062 keywords to date. Table 2 presents the
descriptive statistics of extant research done on artificial intelligence in
marketing. The research data has displayed that the average engagement
of authors for each paper is 2.79 (collaboration Index)

4.2. Performance of scientific actors

4.2.1. Most relevant sources

Table 3 present the five most relevant sources based on the maximum number of articles published
in different journals. Most of the
papers on artificial intelligence in Marketing have been published in
Expert System with Application. The number of articles published in
the next two most relevant journals viz. Journal of Business Research
and Knowledge-Based Systems, are far behind from Expert System with
Application. Further, to understand the most influential source, the top
five most relevant sources were compared on H-index and G-index. Once
again, an expert system with applications topped the chart with both the
highest H-Index and G-Index. Even the total number of citations is maximum for an expert system
with an application. In terms of all indicators,
the expert system with applications is the most relevant source.

4.2.2. Most relevant authors

Table 4 presents the five most relevant authors identified based on

the maximum number of articles published, total citations and citation
indexes (H-Index and G-Index). Liu Y has secured the top rank among
all the researchers with 9 article publications. The other two researchers
Chen Y and Liu J have also shown their interest in the uses of AI technology in the marketing domain.
Further, the author impact is assessed
with the help of total citations (TC), number of articles published (NP),
publishing years (PY_start) and citation indexes (g and h indexes). Liu
Y has an h-index 5 with TC as 97 whereas, Casillas J has an h-index
7 with 203 TC. Casillas J has more citation records than Liu Y. It is
evident from Table 4 that the citation record and h-impact both are independent. Research on a
specific topic or a specific sector could impact
more, whereas research work might get less citation

4.3. Intellectual structure

Co-citation analysis offered the intellectual structure of the research

domain. The research domain is classified in different clusters with the
help of between centrality index computation. Fig. 1 presents the cocitation network analysis. The
cluster was prepared based on the strong
relationships among articles. Due to many papers in a cluster, the authors have selected only a few
with the most citation. The author has
selected a total of five clusters. In each cluster, the number of papers
varied from two to five. They then studied and discussed the research
focus & the suggestions of each cluster.
In cluster one, authors mainly focused on the trust factor that directly impacts selling and
distribution in both manufacturing and service organizations. The authors discussed that trust leads
to long-term
relationships between buyer & supplier, leading to counter-market uncertainty. The authors propose
here to continue the relationship & trust
between buyer & supplier irrespective of the industry segment to get a
competitive advantage. The authors suggest that the research be done
in making a marketing model considering the relationship in the future.
In cluster two, the authors discussed the linkages between market
orientation & business performance. The author has also discussed how
the market is evolving towards customer-centricity. The focus is also
shifting to intangible areas such as skills, knowledge & interactions. The author has given future
research directions to address the effects of additional factors on market orientation & the relation
between market
orientation & market share.
In cluster three, the author describes the value creation for the customer. To make long term value
to customers get a competitive advantage, the organization prepares structural equation models
based on
theoretical-methodological & statistical analysis. There is a tremendous
opportunity to apply those concepts to add more value, especially in the
retail sector areas.
In cluster four, the authors discussed the benefits of data science in
various fields such as finance, marketing, consumer research, and management. The authors also
deliberated about the role of typological theory on cause-effect relationships. The authors suggested
future research
work on predictive validity, not just on the fit validity, to address the
changing business environment issues.
Cluster five discusses consumer sentiment & word of mouth in online platforms. The data captured
through online platforms can be used
for dynamic analysis of an organization. This data will lead the organizations to take measures to get
a competitive advantage in the market.
The studies propose a framework to capture user-generated data. The
authors propose to use the data not only from product reviews but also
from textual communications

5. Conclusion

It is no longer a debate that companies who provides great customer

experiences will be winners in the Fourth Industrial Revolution where
intelligence will reign supreme. The Fourth Industrial Revolution has
been conceptualizing the company to have integrated data about customers and products across all
channels and products, using that data
to understand better its end customer experience and visibility across all
functional areas. In this context, AI and ML have played a crucial role in
big data analytics to anticipate and provide guided experiences to meet
customer expectations. Through this research, the authors provided a
holistic view of using AI to enhance customer experience. Leveraging
AI and predictive analytics is the key to offering customer experiences
that builds advocacy and customers for life. Event-based architectures
combined with AI and predictive analytics is the future. There is no end
state, but it is a journey all companies must begin as we enter the Fourth
Industrial Revolution.
Disruptive technologies such as internet of things, big data analytics, blockchain, and artificial
intelligence have changed the ways businesses operate. Of all the disruptive technologies, artificial
intelligence
(AI) is the latest technological disruptor and holds immense potential for
manufacturing, pharmaceuticals, healthcare, agriculture, logistics, and
digital marketing. Many practitioners and academicians worldwide are
trying to figure out the best fit AI solutions that their organizations can
utilize. However, there is a lack of bibliometric reporting that exhibit detailed research pattern of AI
in marketing. Therefore, this study aims to
aggregate the research studies about AI in marketing using bibliometric
analysis and co-citation analysis.
A five-step methodology for the systematic literature review suggested by Costa et al. (2017) was
used in this study. Initially, research
terms were defined scientifically along with Boolean operators, followed by a detailed search
procedure penned down. Data was collected from the SCOPUS database and saved in BibTeX format
and .csv for further analysis. Bibliometric analysis was done for the most relevant authors’
descriptive analysis, most frequently used keywords and most relevant research papers. Intellectual
structure synthesis was done with the help of co-citation analysis
References

 Anshari, M., Almunawar, M. N., Lim, S. A., & Al-Mudimigh, A. (2018). Customer relationship
management and big data-enabled: Personalization & customization of services.
 Applied Computing and Informatics, 15(2), 94–10
 Pitt, C. S., Bal, A. S., & Plangger, K. (2020). New approaches to psychographic consumer
segmentation: Exploring fine art collectors using artificial intelligence, automated text
analysis and correspondence analysis. European Journal of Market
 Tripathy, A., Agrawal, A., & Rath, S.K. (.2016). Classification of sentiment reviews using
n-gram machine learning approach expert systems with applications
 Shabbir, J., & Anwer, T. (2018). Artificial intelligence and its role in near future. arXiv
preprint arXiv:1804.01396
 Tjepkema, L. (2019). What Is Artificial Intelligence Marketing & Why Is It So Powerful.
Emarsys