Marketers are Watching You : An exploration of AI in relation to marketing, existential threats, and opportunities

(1)

Marketers are Watching You

An exploration of AI in relation to marketing, existential threats, and opportunities

BACHELOR DEGREE PROJECT THESIS WITHIN: Business Administration NUMBER OF CREDITS: 15 ECTS

PROGRAMME OF STUDY: Marketing Management

AUTHORS: Sofia Lindstrom, Sebastian Edemalm, Erik Reinholdsson TUTOR: Jenny Balkow

(2)

Acknowledgments:

The warmest and greatest gratitude towards everyone who has contributed to this thesis and helped us by providing guidance from beginning to end. All those who have provided their time, effort, and support in the progression of this thesis, especially with an ongoing pandemic. Because without them, this research would not have been possible.

Our tutor, Jenny Balkow, provided a backbone for our thesis, continued to show us that she had our full support, and provided guidance for us. Through her extensive knowledge, she devoted time and energy to provide us and this thesis with valuable feedback and information on the subject through her readings and experiences. This thesis would not be what it is without Jenny! By this, we would also like to give a big thanks to the peers from our seminar sessions who have been providing us with salient input during the seminars.

To the interviewees that will remain anonymous, your stories, insights, and time have been the foundation to starting our thesis. Your time allowed us to expand our knowledge and find new insights. To Anders Melander that provided the guidelines and information necessary for the progress and completion of the thesis.

Finally, we would like to commend the founder of Microsoft teams and Zoom Communications for allowing us to communicate safely and effectively during the ongoing pandemic COVID-19. Without these platforms, some of the interviews may not have been feasible.

(3)

Bachelor Thesis in Business Administration

Title: Marketers are Watching You

Authors: Sofia Lindstrom, Sebastian Edemalm, Erik Reinholdsson Tutor: Jenny Balkow

Date: 2021-05-24

Abstract

Background: As of today, it is apparent that with the ever-changing demands and needs of customers, companies are facing enormous pressure to deliver the right value, on time, in the right way, and proper manner. To realize the full potential of Artificial Intelligence (AI), a careful plan and method need to be established in the development and deployment when incorporating the technology with marketing. Technology is evolving at a rapid pace and Artificial Intelligence (AI) can be found in a variety of applications. AI in marketing can provide valuable data clusterization and insights for personalized recommendations, customer segmentation, or even advertising optimization.

Problem: To date, a few studies have been made due to the rapid development of AI which has shown an opportunity for marketers. From this hype, companies are looking into speedy implementation where one can forget that this technology comes with risks and threats. “The

problem is that everybody has unconscious biases and people embed their own biases into technology” (Kantayya, 2021). Although machines can deliver personalized numerical

information, it cannot deliver new solutions such as products and services, nor classify different outputs with a cognitive mindset which could result in biased results. The objective of this research is to utilize the information and insights gathered from experts in the field of engineering and marketing to gain a holistic view of the current and future capabilities of AI in marketing.

Purpose: The focus of this bachelor thesis is to provide additional insights in regards to Artificial Intelligence in relation to marketing, taking into consideration bias, personalization, the black box, along with other possible implications of AI systems, also referred to as the dark side. To fulfill the researchers’ objective, qualitative interviews with practitioners and employees with

(4)

different roles within the field of AI and Marketing were conducted. The paper will be focusing on concepts, theories, secondary data, and interviews which will be further discussed and give opportunities for future research.

Method: To perform this research, a qualitative research design was applied, and 12 structured interviews were conducted with those who have knowledge and experience with AI, marketing, or both.

Results: The study elucidates the potentials and fallbacks of Artificial Intelligence in marketing. Where the findings suggest a mixture of human intervention and technology is needed to work against the perceptions, bias, and manipulation the technology can possess. The aims then guide towards the conclusion presenting the important cognitive and emotional skills that humans obtain that are currently lacking in AI.

This study finds several key areas both in terms of opportunities and risks. Such key areas involve the possibility of delivering new, unique personalized content to a mass audience at lightning-quick speed and at the same time presenting a handful of risks by giving machines the permission to make human decisions. Risks found in this study presented as the dark side include

the bubble, bias, manipulation, fear of losing jobs, lack of transparency creating the black-box phenomena. Therefore, this research is interesting especially for marketing managers in how AI

could be used both from an opportunity perspective and possible risks to consider.

Keywords: Artificial Intelligence; Marketing Automation; Machine Learning; Bubble; Bias; Personalization; Black Box; Manipulation; Marketers; Human intervention

(5)

List of Figures

Figure 1: Outline of research approach 10

Figure 2: Visual Presentation of the Conceptual Framework 21

Figure 3: A final Framework of How AI Affects Marketing 55

List

of Tables

Table 1: Interviewee Overview 16

(8)

1. Introduction

This thesis starts by tackling the background and elaboration of its emergence. The goal of this chapter is to address the problem of AI and Human thought and emotion, from this leading on to the purpose and research question of this study. The unfolding of the delimitations of the study.

1.1 Background

It is of human nature to justify decisions with supporting information. Questions ranging from which customers to cater to, their interests, or what approaches one can use to reach out to customers. In this competitive environment, technology and marketing are going hand in hand, to create a path for triumph (Dereli, 2015). In turn, controversies have been growing concerning the involvement of having Artificial Intelligence in making business decisions. This means the latter has become more important to marketers over the years, to bring in social knowledge and human touch to prevent the technology from scandals potentially ruining the brand.

Artificial Intelligence (AI) has been present for decades but lacked the fundamental means to reach its full potential regarding the access of data that would push AI to the next level (Haenlein & Kaplan, 2019). The mass of modern society has adopted technological advancements and has continued to evolve since the origin of the internet, which has introduced a tremendous amount of data. Giving AI developers the means to feed this historical data into these AI algorithms and that marketers can now seize opportunities to personalize content into enhancing the consumers' experience. The benefits of AI have opened doors for all entities and contributed to delivering specialized content to customers. In terms of marketing, implementing automation has helped businesses to become more viable in the hype of AI and marketing activities (Goyal, 2020).

“Artificial intelligence has dominated popular culture for years; and soon may dominate marketing” (Connick, 2017). With the ever-changing societal demands and needs of customers.

The dynamic and abundant amount of information allows AI to come into play by simultaneously creating the demand for personalization, clustering data, and tackling the dark sides of technology. In relation to this, a study by Freed (2020), states that AI and human

(9)

thoughts and emotions should synchronize with each other. Which then unfolds the research question “How is Artificial Intelligence influencing Marketing today?”

Based on this research question, two themes will be discussed in this study: ❖ What are some of the possibilities of implementing AI?

❖ What are some of the dark sides and possible consequences of implementing AI?

1.2 Problem Discussion

Demands are constantly changing and technology is constantly progressing in a dynamic environment. This puts pressure on marketers to meet current demands and adapt to the latest technology to stay competitive in the fierce challenge of staying on top of the market. The move from traditional marketing to digital platforms has put pressure on firms to stay competitive and new tools have emerged for businesses to utilize in the best way possible. AI is one of those cutting-edge technologies that quickly have become very powerful and very popular, especially within marketing. Many things can be automated with AI however, a tool cannot fully replace the need for human intervention. Rushing to things may have consequences, therefore this study is taking a closer look at how AI could be used within marketing by interviewing experts in the field for valuable insights and connecting it to previous academic research.

1.3 Research Purpose

As marketers adopt these new technologies to stay competitive and meet customer needs there will always be consequences. Therefore this study takes a new perspective of how AI could be used and implemented for marketing management and the possible outcomes of doing so. It is meant to give guidance to marketers on how AI could be used and the possible dangers to avoid. Budget restraints are becoming a norm and the need for a more effective customer approach makes marketing managers look to automate certain processes. This gives the research a purpose to highlight the key areas that marketers need to take into consideration before properly implementing AI. This topic was selected based on the high relevance of Artificial Intelligence in today’s society as well as the researcher´s interest within this area.

(10)

1.4 Research Question

How is Artificial Intelligence influencing Marketing today?

1.5 Delimitations

This research concerns three main delimitations. The first part is that the research is based on a perspective of individuals’ expertise and involves a sample consisting of multicultural interviewees with a profession in marketing and/or Artificial Intelligence. It was primarily employees from the Swedish telecom company Ericsson with a variety of roles from marketing, commission management to telecom communication engineers around the globe. In addition to this, a few other interviews were conducted outside the scope of Ericsson and those companies will not be disclosed. This sample was chosen not based on population criteria, gender, geography, nor age, but experience and competence.

Secondly, the study only investigated the employer’s perception and experience of the phenomenon in relation to their everyday business activities and knowledge within the field of AI and Marketing. Third and finally, the research and interviews were conducted solely online during spring 2021, where the researchers have faced an ongoing outbreak of COVID-19.

(11)

2. Methodology and Method

This next chapter addresses the methodology and method of the research. The objective of this is to explain the reasoning behind choosing to do qualitative research and justify the means of the method used for the thesis. The chapter will begin with a description of the research approach, how interviews and data were conducted, followed by the sampling technique and tools used. Three documentaries were used to contribute additional insights. Finally, the process of transcribing, use of quotes, ethical considerations, and credibility of the study is discussed.

2.1. Research Approach

As for the research, following an inductive approach where the chain moved from a bottom-up approach, with our observations and data gathering first, then using this to find patterns, and then to a theory (Bradford, 2017). This slightly unusual approach by beginning with the interviews and from there building the conceptual framework, instead of the other way around, was done in the search for a scope. As the topic concerning AI was new for all of the researchers, the purpose of beginning with the interviews was to get hands-on experience from industry specialists. This made it possible to assess the different topics within this thesis.

Figure 1: Outline of research approach

Qualitative studies are usually expressed through words, concepts, and patterns. The whole process of this study will be going on: formulating research questions and analyzing through interpretation, summing up, or categorizing patterns and themes. Qualitative research relies on the data the researchers have gathered and obtained. It is research that does not intend to gather statistical information or other types of data quantification, usually, the outcomes of this time of research are attained by analyzing practical situations (Golafshani, 2003, p. 600).

(12)

The objective of this study is to provide guidance and an understanding of Artificial Intelligence in the role of marketing managers. Leavy (2017, pp. 19-20) describes the interpretive paradigm as a philosophical belief system emphasizing people's subjective experiences in the context of social science. Furthermore, the author mentions patterns of interaction and interpretive processes referred to as the social construction of reality when researchers try to get a deeper understanding from subjectives’ experiences and circumstances. Hence, deep insight is considered more valuable than the probability of unbiased opinions. Thus, the research will emphasize keywords, themes, and patterns.

2.1.1 Observation Through Documentaries

During the time of research, three documentaries were used: Coded Bias (Kantayya, 2021), The Social Dilemma (Orlowski, 2020), and The Great Hack (Amer et al., 2019). Secondary sources are beneficial for the research because it can potentially increase efficiency, as it provides additional data acquired by others that could potentially contribute for additional insight to the research (Sekaran & Bougie, 2016, p. 38). The documentaries used for this thesis are critically analyzed and referred for the purpose to contribute with opinions originated from experts that possess expertise within Artificial Intelligence and data gathering.

The documentary Coded Bias (Kantayya, 2021) exemplifies our power of free will. Which embraces people’s rights to learn about the technology and people’s rights to shut it down. The film explores Machine Learning algorithms and how it has shown existing race, gender inequality, and social class differences. People believe that AI is the new answer for the future.

“However, because AI is based on the data we fed them, the data is just another reflection of our history.” (Buolamwini, 2021). Which is just another reality for how history is catching up with

us. It uncovers the fallout of an MIT Lab researcher, Joy Buolamwini, and her discovery of racial bias in the facial recognition programs.The documentary then proves that Artificial Intelligence is not neutral and in need of governmental intervention (Flores, 2020).

The Social Dilemma (Orlowski, 2020) unfolds the increased awareness of the dangers of social networking. Tech experts are arguing about the impact of social media's extensive purpose to

(13)

media to be able to keep users engaged on the platforms are because of the data collection and controversial technologies that analyze people's behaviors. Therefore, as a tool, Machine Learning proves to be powerful in predicting people’s behavior by gathering data and analyzing it. As a result, The Social Dilemma contained additional insights regarding the use of Artificial Intelligence in social media and how humans might be manipulated by all content presented through the platforms.

The Great Hack (Amer et al., 2019) is set around the Cambridge Analytica scandal that was revealed in 2016 by overcoming data of 87 million users of Facebook. The documentary emphasizes how valuable data is and what could happen when used inappropriately. Data can be skewed and biased therefore trust is difficult to obtain and it requires proper policies to avoid risks such as manipulation. It is important to consider the intentions of data gathering, what the purpose is, and how it is being used.

Disclaimer, the mentioned documentaries were used for new insights and additional perspectives. Hence, not used as a primary source of information as it could contain skewed bias depending on the unknown intentions and purpose of these documentaries.

2.2 Method

2.2.1 Sampling Approach

Sampling techniques refer to the method used to narrow down the different elements in collecting data from a subgroup. A sample consists of a few chosen elements which enable researchers to save time, costs, and human resources and are more likely to provide more reliable results, thus fewer errors (Sekaran & Bougie, 2016, pp. 235-241).

Furthermore, a sampling design: non-probability sampling was used. Non-probability sampling means that each element is of subjective judgment of the researcher rather than random selection. This is more common when time is a constraint. For this research, a combination of purposive and snowball sampling was used.

(14)

❖ Purposive sampling design, more precisely a judgment sampling was selected to continue

this study. Judgment sampling involves the choice of individuals who are the most suitable to provide the appropriate information (Easterby-Smith et al., 2015, pp. 78-81). ❖ Snowball sampling is the continuous loop of purposive design, the people who first were

interviewed based on the researcher's judgment were able to provide access to other suitable individuals meeting set criteria. This works well for closed networks where industry specialists can be very difficult to find and contact (Easterby-Smith et al., 2015, p. 82)

2.2.2 Process of Selecting Candidates to Interview

The sample design of chosen elements was evaluated from two criteria. ❖ Whose profession involves, experience with artificial intelligence (AI) ❖ Whose profession involves, experience with marketing

Or involves both AI and marketing

The selection of interview objects was purposely selected through a network based on one of the authors’ family members, whereby that person had access and connection with several people who fitted the criteria of this research. As a result, the majority of the interviews came from a line of work with a close relation to Artificial Intelligence and marketing on a high level of expertise. By that, it became natural to use that channel since it generated rich knowledge regarding the topic of the research.

2.3 Conduction of interviews

2.3.1 Interview Questions

There are two primary methods to consider when planning the interviews, unstructured and

structured interviews. Unstructured interviews mean that no questions are organized in advance

and are ideal for finding solutions to unknown problems and the discussion evolves with time. Albeit, considering the interpretive paradigm of the interview is to grasp underlying resonances from subjective interpretations, opinions, and attitudes (Leavy, 2017, pp. 19-20). A structured

(15)

interview was chosen for this study, prepared with a set of questions that was organized the same way for each interview.

To successfully explore areas of interest in interviews, it is valid to formulate appropriate questions. “When writing the guide, ask yourselves what do you need to know in order to answer

the research question?.” (Bryman & Bell, 2011, p. 486). This structured interview was done with

open and probing questions. Easterby-Smith et al. (2015, pp. 214-217) mention that the aim of qualitative interview questions is to attain an understanding from respondents' perspectives that not only includes viewpoints but also the reason, to capture the meaning and interpretation for these viewpoints. Hence, this structured interview consists of 26 prepared questions to capture these insights. (Refer to Appendix A, Interview Part 2: Questionnaire).

Structured interviews were done for two reasons. Firstly, to keep consistency within the research and secondly, avoid risks in getting potential bias results. Another way that helped deal with the broad concept of the research was to find different ways to get different meanings. In other words, breaking down big ideas into multiple and more tangible questions. Some questions were also highly selective in terms of the answers the researchers desired. For example, to get a deeper scope of the darkside of AI. Two questions were directly worded in relation to the dark side of AI and the other in relation to marketing. In juxtaposition, to then highlight the potential benefits of AI and marketing, three questions were formulated on this.

2.3.2 Guide to Interviews

As mentioned in section 3.3, structured interviews are conducted with the help of an interview guide (Bryman & Bell, 2017, p. 454). Initially, access was given to key personnel at Ericsson, an industry-leading company in telecom. Hence, at first, the questionnaire was written from this perspective. Although, it should be noted that no one was interviewed as an employee representative but rather for their knowledge within the field and not because of employee titles. During the process of interviewing, the chance to interview other individuals outside of Ericsson ultimately led to small changes in a few questions. The only changes in those questions were words replacing “Ericsson” with “your company”, leaving the remaining sentences completely untouched. (For consent form/approval of interviews, see Appendix B).

(16)

2.3.3 Interviews Through Video Conference

Face-to-face interviews can be one of the best approaches in comparison to virtual ones. Depending on the scope of the research factors such as body language, facial expressions, etc. can be taken into consideration. Risks such as interruptions, issues with an internet connection, poor volume, or other technicalities are then avoided. Bogner et al. (2018, p. 663) argue the idea that face-to-face is a prime alternative when conducting interviews. With an ongoing pandemic Zoom and Microsoft Teams were pre-eminent alternatives. Additionally, some of the respondents were geographically dispersed around the world.

According to Bogner et al. (2018, p. 663) interviews on the phone tend to be shorter than that face-to-face. The difference in time is crucial, longer interviews tend to obtain richer information. This never became an issue for this study concerning the two interviews that were over the phone due to video conferences not being possible at the time. However, the difference in time was not very significant, the average time for all the interviews averaged 51 minutes. A webcam was optional hence it enabled greater insight to see how the participants reacted with each question, the facial expressions allowed the research to determine if further elaboration was needed on a question (Sekaran & Bougie, 2016, p.120). Prior to using Zoom and Teams, setting up non-verbal communication via email allowed each party to discuss a bit about the thesis, language used, and our purpose for interviewing them, which strengthened the quality of communication. Finally, at the end of each interview, every participant was asked for possible follow-up sessions if more information was required.

2.3.4 Conducting the Interviews

Down below displays an overview of the conducted interviews (Refer to table 1). Each of the categories discloses the number of the participant, length of the interview, platform used, language, and profession. Due to differences in schedules, sometimes all three interviewees were present, sometimes two, and for interviews conducted over the phone, only one interviewee participated. All interviews were transcribed based on the recordings that were approved by each participant. This strategy ensured that all information was gathered properly to minimize the risk of acquiring biased results or some information was left out (Sekaran et al., 2016, p. 119).

(17)

Table 1: Interviewee Overview

All the participants having different national backgrounds, the interviews were performed in English; as a universal language. This was done to avoid any potential communication barriers between the researchers and the interviewed that could reflect in the results if questions were asked in different languages. Sekaran & Bougie (2016, p. 146) states that some expressions in English can be misunderstood since they could be interpreted differently in other cultures. Therefore, the questions used during the interviews were carefully designed with terminology within AI and marketing to minimize the risk of becoming misinterpreted.

All the interviews began with the same introduction, first explaining the purpose of this research and then their role in being interviewed (Refer to Appendix A)

(18)

❖ The first questions concerning background, interests, work positions, and working

experience of the interviewee

❖ Secondly, questions concerning marketing. This gave insight on the participants’

level and knowledge on marketing related.

❖ Thirdly the interviewer turned to questions regarding AI.

The goal for starting the interviews with this arrangement of questions was done for three purposes (Sekaran & Bougie, 2016, pp. 115-116).

❖ Allow the participant to become comfortable and ease into the subject. ❖ Allow the researchers to gain a better understanding of the respondent, their

competencies, and experiences.

❖ Finally, to find new aspects regarding the research topic.

As such, the focus was to be adaptable and flexible, if a specific interviewee wanted to discuss in Swedish. The possibility was offered, however, only one interviewee preferred Swedish, which was solved by having two of the researchers who are fluent in Swedish conduct that interview. It was clear that it allowed a natural flow and higher levels of in-depth conversation between the participant and interviewee. The interview conducted in Swedish was translated to English and eventually paraphrased in related topics. Interestingly, Bryman & Bell (2011, p. 488) also discuss that translation could possibly impose problems in the analysis of the material. However, with all three being bilingual and two interviewees native Swedish speakers, thus no major issues were faced during translation.

2.4 Interview Notes and Transcribing

Saunders et al. (2012, p. 394) propose that recording and taking notes simultaneously is a method that can benefit researchers and increase precision. By doing this simultaneously for this research was to achieve two things; firstly, to be able to take notes during the live session, ask follow-up questions, and interact with the text itself. Secondly, allow the researchers to go back and listen to the recordings, reflect, and assess as a team. The group split up roles, one or two would be taking notes, and the other one was to focus solely on conducting the interview. This method was

(19)

used to give the researchers both a chance to focus on the interview and its questioning and listening (Saunders et al., 2012, p. 396). The transcribing process itself was executed by the researchers moments after the interviews were finished or before the next interviews were conducted.

2.4.1 Themes, Patterns, and Selection of Quotes

Throughout chapter 4 in the analysis and findings selected quotes from the interviews will be highlighted. Some participants were quoted more than others. This is due to a couple of reasons, first the quality of the participants' answers and how valuable their information is in terms of relevance to the topic. Second, the quotes were carefully selected for the most explanatory of the researchers’ interpretation of the data collected. Finally, all the content which is not mentioned in the findings is not because it was less important or interesting for the study. But instead the true value and support it had as well as correlation with other respondents. Some respondents conveyed the same patterns and themes, which were distinguished during the transcribing process. These patterns and themes were clustered in different topics from the interviewees regardless if it was one interviewee who represented a pattern or theme, it will still be taken into consideration as it is a qualitative study. The final themes and subcategories/patterns are discussed in the analysis and findings. (Refer to Figure in section 4.6)

2.5 Ethical Considerations

Before conducting our very first interview, a guide for what to say and mention in each interview was formulated. This was to ensure integrity and consistency throughout the interview process. All the respondents were first contacted via email and informed that the interview was completely voluntary, although everyone who was contacted agreed to the interview. A small summary was also provided to give insight on the topic and what could be expected from the interview, in addition, that any question that felt uncomfortable was acceptable to skip (neither did this occur). It was carefully disclosed that any information that was given throughout the whole process would be used to support the analysis, finding themes, patterns, but once the thesis is complete, the information would be deleted. Important to remember that are never any amiss

(20)

answers, it is up to the researchers to ask the appropriate questions and reach for an adequate response.

2.5.1 Anonymity and Confidentiality

The data collected was not shared with anyone outside of this research or unauthorized. Easterby-Smith et al. (2015, p. 146), states that asking for permission to record an interview is crucial to be done before the recording begins during the interview. Therefore, at the beginning of each interview, a small introduction was proposed to explain their rights and ask if it was okay for us to take notes as well as record, which every respondent agreed upon. However, in order to ensure this and avoid any chance of misunderstanding between both parties, the script included a short presentation about us, the topic, and our purpose. The information mentioned above is to facilitate anonymity and confidentiality for the participants of this research during each interview.

2.5.2 Credibility

One of the crucial aspects of establishing credibility for this research is the execution of conducting the interviews. The reason for this is that the relationship between the interviewees and the authors should be formal in order to receive credible and unbiased answers during the interviews. As stated by Sekaran and Bougie (2016), to obtain unbiased answers from an interviewee is to build trust by giving the interviewee a clear explanation about the purpose of the research and ensure confidentiality. Some of the participants have been related to one of the authors and therefore in those interviews, the author with no relationship with the interviewee has taken the lead during the interviews to avoid inequality. Otherwise, the majority of the interviews took place where the authors and the interviewees met for the first time. Additionally, Sekaran and Bougie (2016) state that biased answers can become evident if the respondent feels the pressure of being a representative of a particular organization and can generate a dishonest response by the fear of not agreeing with company standards. Therefore, as mentioned in section 2.5.1 at the beginning of each interview, the author ensured that the interview is solely and anonymously based on the field of experiences and personal opinions.

(21)

3. Conceptual Framework

The following chapter discusses and highlights the different aspects and presents the method used for the conceptual framework. An extensive elaboration of Artificial intelligence, the different concepts, and marketing is provided. The concepts are all linked to digitalization, personalization, and how AI is influencing marketing, which is the final concept of investigation in this chapter. Additional insights will be discussed such as the influence of technology on marketing and how technology is affecting social entities.

3.1 Method Adopted for the Conceptual Framework

The conceptual framework for this study was established around understanding the

concept of Artificial Intelligence and the related theory of marketing. For this study, Machine

Learning was one of the core characteristics of Artificial Intelligence. To realize the full potential

of AI, trust needs to be established in the development, deployment, and use of AI. Following factors will be discussed; digitization, personalization, the darkside of AI, were examined. For the conceptual framework, the method adopted is an inductive approach where there is external validity. External validity can be defined as the relations among existing variables as observed in a sample. The sample population will hold other samples of observations in the same population, often seen as the generalizability of results (Frey, 2018). To ensure the validity of these articles and journals, there were guidelines and criteria of conditions for the literature, namely (Thurén, 2013, p. 7-8):

1) Authenticity 2) Independence 3) Time relationship

The first criteria state that the source used should stand for what it says to ensure authenticity (Thurén, 2013, p. 7). The scope of the literature was based on the keywords which were used on the Primo database and Google Scholar, in regards to Artificial Intelligence and marketers roles. But thousands of results showed up which is why different keywords such as machine learning,

(22)

product recommendation, bias, and so forth were used to narrow down the results. From this, the

different articles, journals, were selected based on relevance to the abstract and headline.

The second criteria state that the source should not depend completely on other sources, but also be an independent first source. This was not strongly focused on with some of the sources due to the limitation in some concepts (Thurén, 2013, p. 7). However, the use of secondary sources is inevitable and works as a complement for this research.

The third criteria refers to the timeframe or relevance in terms of that collected data was not limited to any specific year, due to the wide scope of looking at the whole journey of AI in relation to marketing automation. Although the third criteria highlight our scope which was wide open for interpretation. In order words the relationship between the source of the story vs the actual happening of the event. Additionally, time was not a factor, thus the scope was open for interpretation from all timelines (Thurén, 2013, p. 8). Not to mention, the brief explanations from several interviewees who elucidated that AI has been around for decades.

3.1.1 Visual Presentation of the Conceptual Framework Figure 2: Visual Presentation of the Conceptual Framework

(23)

This figure displays a visual presentation of the different concepts and factors which will be taken into account concerning AI and marketing. The sole purpose of this figure is to give a clear and concise visual understanding of the conceptual framework.

3.2 AI What is it - The Key Concepts

3.2.1 Artificial Intelligence (AI)

Can be defined as the ideology as well as the development of computer systems that can undertake jobs typically requiring human intellect. Examples include the ability of visual recognition, voice recognition, easier solutions to decision making, risk management, and predictions, tackling problems, and translation of languages (Oxford University Press 2021). Artificial Intelligence contains algorithms that support programs and systems to increase performance by analyzing big data. It follows that AI establishes “intelligence” in the same way compared to human intelligence by using Machine Learning to make decisions and predictions based on big data analyses from collected information of consumer’s purchase history (Shankar, 2018).

The core principle of AI is the data that is being used. Therefore the biggest five tech companies, Facebook, Google, Apple, Microsoft, and Amazon, are holding an edge for the data it possesses about the worlds’ citizens (Lekkas, 2020). Due to the data being processed it is widely discussed how one’s integrity is protected and to what cost. From this, different documentaries have emerged to reveal behind the scenes how data is collected and how it can be used. In “The Great Hack” it is evident that data is not only about individual integrity, it can also be dangerous such as manipulating people amongst other things (Amer et al. 2019). Brittany Keiser, former Director of Business Development for Cambridge Analytica describes “the wealthiest companies are

technology companies, Google, Facebook, Amazon, Tesla and the reason why these companies are the most powerful companies in the world is because last year data surpassed oil in its value. Data is the most valuable asset on earth. And these companies are valuable because they have been exploiting people's assets.” (Amer et al. 2019). Cambridge Analytica came by the data of

87 million Facebook users without consent, this data can then be sold or used to hold an advantage against the user. An example of this could be for marketing but specifically, target the individual with content that it is most likely to purchase (Amer et al. 2019).

(24)

3.2.2 Machine Learning (ML)

The purpose of Machine Learning revolves around how humans can train or program computers to enhance the systems to increase efficiency or be able to make predictions. The predictions can be made from data sets that are given and the computer analyzes the data using algorithms. The accuracy of each prediction depends on the sample size of data because more information will enhance the ability to make more precise predictions. There are several methods e.g., supervised learning, unsupervised learning, and reinforcement learning used to create a learning environment for the computer in order to teach the computer about which approach is best for different situations. The end goal of Machine Learning is for it to become a product that is smart enough to make its own decision without human input (Chen, 2019).

Bolton et al. (2013) mention that marketing departments around the globe are facing increasing complexities such as consumer demographics, technology changes, cost restraints, and abundant quantities of data. This in turn is increasing the demand for new and improved solutions such as models with propositions to create value and differentiation to stand out (Bolton et al.,2014). Moreover, this study will treat AI as a general phenomenon and focus on the effects it has on marketing, therefore any specification of the technologies or algorithms involved will not be specifically analyzed. The aim is to uncover the different terminology and themes in relation to the purpose of the topic.

3.2.3 Narrow Artificial Intelligence (NAI)

The concept of Narrow Artificial Intelligence is about how an AI system can execute a task well enough that it could be equally as good in comparison to a human who possesses a high-level skill set for that particular task. The reason behind the coined term “narrow AI” comes from the fact that the AI can only execute one particular task at hand and has to be programmed in advance to make it suited to tackle the problem (Negnevitsky, 2011).

(25)

3.2.4 Artificial General Intelligence (AGI)

The idea of Artificial General Intelligence (AGI) is the level of artificial intelligence where the systems can solve complicated problems in different scenarios by utilizing their own experience based on trial and error. As a result, AGI can learn how to solve problems in the future by itself without any human interactions (Goertzel & Pennachin, 2006). In order for the AGI to gain experience, it has to use some sort of method from Machine Learning to understand the trial-and-error behavior, e.g Reinforcement Learning can be used for this. Reinforcement learning is about how to teach someone or something the consequence or reward of carrying out an action. The concept will not guide the learner about what to do, instead, it will discover what outcome each action will provide. The process will not only focus on isolated decisions, but it will also learn what outcomes can affect upcoming actions and the consequence or the reward (Kubat, 1999). Predict the outcome of a specific marketing campaign for instance.

3.3 Marketing Automation

In John DC Little´s presentation at UC Berkeley 2001, the term marketing automation was introduced, where he continues to discuss and refers to it as the act of automated marketing decisions support system through the internet (Little et al., 2001). The way Little et al., (2001) explained his theory, suggesting that analyzing the digital footprints of customer A, B, or C, could help tell retailer Z of what appropriate tools and models can be used. He phrased it as “What do we tell retailer Z to do when customer A arrives that morning”. Using a model can allow one to gather information, patterns, and data to back up and strengthen the retailer's argument, which should boost productivity and problem-solving. A study by Heimbach et al., (2015) explored marketing automation in relation to business and information systems. It was found that marketing automation supports better decision-making, productivity, greater returns on investment, and customer satisfaction.

According to Goyal (2020), Artificial Intelligence can create opportunities for companies to decrease the marketing budget by using AI to automate the marketing content. AI helps marketers to predict current customer’s behavior in terms of satisfaction and by that decreases the risk of losing customers. The marketers use AI to find targeted customers and develop

(26)

advertisements according to the result generated from the data analyzed by the AI (Goyal, 2020). This will increase the level of quality put into the work of each campaign that will increase customer satisfaction by giving them personalized communication between customer and company. Murphy (2018) discusses the use and understanding of automation in marketing, how companies should implement automation successfully. Furthermore, marketing automation can enhance the productivity and collaboration between marketing and sales for businesses. By outlining a guideline about the crucial steps a company must master in order to implement marketing automation properly. E.g., evaluate existing internal marketing processes, develop timely content, and allocate human resources properly.

3.3.1 Digital Marketing

A notable framework, by (Kannan & Li, 2017) can be connected to this research. Over time, the term digital marketing has evolved, using digital channels to market products and services. The umbrella term is the process of using digital technologies which facilitate an environment to create, communicate, and develop value for customers. It obtains customers, and determines customer preferences, promotes branding, and returns on investment (Financial Times, 2012). With such rapidly changing environments, digital technology allows ease for companies, as it can reduce information asymmetries between customers and sellers. Digitization can foster a healthy environment which can help with the rapid changes, reducing any lag time.

The definition of traditional marketing can be applied to digital marketing. It can be defined as the set of strategies and tactics that are executed through digital channels to achieve corporate goals with a defined period of time and limited resources (Minculete et al., 2018).

3.3.2 Social Media in Marketing

The study by Appel et al., (2020), illustrates the profound impact social media has on marketing and the benefits for the practitioners within the industry. All the social media platforms available today present a network populated by billions of people. This new universe connects all users worldwide and establishes an excellent channel for marketers. As a result, marketers possess the incredible freedom to select a specific market segment and for the exact purpose of exposing the segment to a tailored marketing advertisement through social media platforms.

(27)

3.3.3 Social Media's Use of Artificial Intelligence in Marketing

Artificial bots, also known as social bots in social media are used for several purposes, for instance, used to interact with consumers on behalf of companies. According to Ferrara et al., (2016), social bots can be exceptionally useful for customer care, because bots help companies to make quick responses towards the consumers in social media. Although, there are examples of situations when the attempt of using bots has failed or been misused e.g., in 2016 Microsoft launched on Twitter a bot named “Tay” to engage with millennials on the platform. Unfortunately, the experiment got terminated within 24 hours. The reason for this failure was the influence of the community on Twitter, which made the bot Tay write controversial statements on the platform (Hunt, 2019). However, Neff & Nagy, (2016), stated that the organized group of people on Twitter proved a point about how a well-developed algorithm can be tempered with and can be changed in a short period of time. As a result, the outcome became a catastrophe for Microsoft but likewise a big failure for Artificial Intelligence in social media.

3.4 Personalization

Smarter With Gartner’s definition of personalization follows as a “process that creates a relevant, individualized interaction between two parties designed to enhance the experience of the recipient” (George, 2017). Furthermore, personalization is a frequently used term in marketing, which Misiak (2019) describes as a “method that utilizes consumer data to modify the user experience to address customers by name, present shoppers with tailored recommendations, and more”. Adomavicius et al. (2019) say that recommendation engines are meant to simplify and reduce the time spent on research by consumers and at the same time generate a diversified sale that builds trust and loyalty for businesses. The rumor says that Steve Jobs once uttered “people don't know what they want until you show it to them” (Smith, 2019), this is the enigma that recommendation engines, i.e. personalization could solve to increase revenue streams. Scenarios using AI can be found in a variety of applications such as Netflix, Spotify, and Amazon which will be further looked at in this thesis (Spotify Engineering, 2020; Baer & Ngahane, 2019; Netflix Research, n.d.). Furthermore, MacKenzie et al. (2013) talk about how AI has shaped users’ experiences and interests since the early beginning of the smartphone era. In addition,

(28)

Earley (2017) means that this was possible because more devices generate more data that is the foundation of all AI recommendation systems.

3.4.1 Product Recommendation

The study conducted by Bock & Maewal (2020) hypothesizes the idea of adversarial learning for a product recommendation, and the findings suggest that the recommendations produced by the different models can provide benefits for marketers and consumers.

“Product recommendation can be considered as a problem in data fusion—estimation of the joint distribution between individuals, their behaviors, and goods or services of interest” (Bock & Maewal, 2020). In other words, a system that collects and filters the data, clustering it to show items that the customers would most likely want to purchase. Despite it not being 100% accurate, it is a tool that helps one predict certain products and services to specific people. The systems that use recommendation algorithms are able to take in an abundant amount of data, which we as humans would not be able to progress. Iyengar & Lepper (2000) is phrasing the filtering process can be of importance “based on studies on marketing, it discusses that the concept of having too many choices can decrease consumer satisfaction and suppress sales.” Amazon was the first to use and develop the product recommendation engines in the early 90s that now have become a business standard within sales, more so with the use of AI in addition to filters(Smith & Linden, 2017). Hence, the product recommendation functionality has helped Amazon capture 40% of the entire e-commerce sales in the United States (Droesch, 2021).

On another note, a study conducted by Huang & Philp, (2020) examined the connection of AI and service failures. When a human employee had a service failure, customers shared a negative word of mouth based on the experience. On the contrary, when it was a system failure by AI, (same service failure and variables) customers were less willing to complain. This has proven some businesses using AI as a form of backup for service failure.

(29)

3.5 The Darkside of AI

As discussed above in section 3.2, the three main aspects of AI are Machine Learning, Narrow

AI, and AGI, which are all supposed to supplicate an artificial or human-like mind. For

businesses, employing AI as a part of the team holds great value for innovation and problem-solving. An example discussing how benefits, could raise uncertainty and risks:

“Its efficient ability is similar to human intelligence, allowing marketers to become more competitive and improving the customer journey. But even with these benefits, it can be

calamitous.” (Mahmoud et al., 2021).

A journal by Wagner (2021) states that AI can leverage dark traits. Highlighting that leaders and other employers that show personality traits like sadism, narcissism, or psychopathy can foster a basis of traits in others, including technology. Additionally, Spector (2017) also discusses that innovations might not always be for the greater good, the initial intentions could be but the result illustrating something else.

3.5.1 Recommendation Engines (Netflix & Others)

The struggle for companies in e-commerce was how to understand the customers' taste in either music, movies, or products based on users’ historical data. The recommendation systems have solved the issue to an extent where companies with the help of Machine Learning can gather the data and predict what the consumers might like to consume next (Adomavicius et al., 2019). In addition, Adomavicius, et al., (2019) discuss how these recommendation systems can be biased in terms of decision making and how that affects consumers. Because the recommendation algorithms could shape the customer into specific preferences and make the customer like new things that otherwise never would have happened naturally. This introduces the problems mentioned in the following parts of this chapter.

3.5.2 The Bubble, Transparency, and Manipulation

The concept of the bubble is a term used throughout this research to highlight the idea of being clustered into a certain type of box. “What it means, is that it clusters-specific lists of different

(30)

products which then are used as inputs and categories in the matching process.” (Lawrence et

al., 2001). This term refers to the idea of being confined to certain products and services. What the technology does is provide recommendations to match products to customers based on the expected appeal of the product or service based on behavior and history (Lawrence et al., 2001). This helps determine and identify relationships customers have with different products and categorizes them in this so-called bubble for future marketing purposes. The bubble creates the risk of missing exposure to new content and being stuck in a vicious loop. Schelenz et al. (2020) emphasize the importance of transparency in elements of AI personalization. Defined by the authors as “transparency is a practice of system design that centers on the disclosure of information to users, whereas this information should be understandable to the respective user and provide insights about the system”. Schelenz et al. (2020) further highlight how beneficial AI personalization could be in terms of delivering relevant content to users. Although, the lack of transparency in how the Machine Learning process is established has created ambivalence to how personalization may introduce the risk of bias and manipulation.

Dahl (2018) points out that there is no need to reveal secret details about an algorithm for the end-user. Rather than transparency means explaining plain and simple in a sufficient manner how the outcome of the algorithm came to be, what led to the results. When this influence is not transparent and instead subtly tries to persuade the person to take action, perhaps against their own goals, this then becomes a constitution of manipulation. Brittany Keiser explains how powerful data sets can be and that it further could be used to manipulate what she calls

“persuadable people” (Amer et al. 2019). She mentions that during the US election 2016, data

was used to make people vote in a certain way by targeting these individuals on social media with specific content that would persuade them to vote in a certain way in the election (Amer et al. 2019).

3.5.3 Bias

Coded Bias is a documentary that highlights how some technologies are biased to certain groups of people. Asking the daunting question, “what happens when technology inhibits our liberties?” (Flores, 2020). Businesses are as of today committing more to AI algorithms to do

(31)

According to the Cambridge Dictionary (n.d.), bias is defined as “the action of supporting or

opposing a particular person or thing in an unfair way, because of allowing personal opinions to influence your judgment”. People are influenced every day by different sources and different

scenarios. The difficulty is to identify the known or unknown intentions of those who are trying to influence someone else's behavior or convince them to take action. It is not only people who can influence each other, it is also possible that AI can be biased in terms of how data are interpreted or shown to users. Kantarci (2021) describes “AI bias is an anomaly in the output of

machine learning algorithms. These could be due to the prejudiced assumptions made during the algorithm development process or prejudices in the training data.” AI is built on data that could

be used in different ways to alter the results depending on how it is interpreted, lack of data is one way the results can be altered. Such biases are possible to be found in scenarios like the

sample, exclusion, measurement, recall, observer, racial, or association bias (Lim, 2020). For

instance, if an algorithm is made to make decisions based on a given data set it is necessary that the data is genuine and transparent, or hygienically clean as Lim (2020) describes it. The data could be biased before it is even used for the algorithm. Hence, the importance that no data is excluded or that any specific data is given favoritism in order to affect the outcome.

Richards & Gummadi (2018) speaks of algorithmic decision-making to be non-discriminatory and transparent to be unbiased. An algorithm could be trained to make as few errors as possible to predict who is most likely to recidivate or re-offend in the near future of the entire population. It is about training the algorithm to identify patterns based on historical data and previous records of a criminal offense. The result could be discriminating if one sub-group of the population is overrepresented, a subgroup could be based on socio-economic factors or race. The interpretation of data is always a trade-off when determining the objectives of training AI algorithms. Bias could be measured by the outcome or the procedure itself. On the other hand, human decision-making can be very biased, the difference is that the human can be asked about the intentions and give an explanation to his reasoning. An AI algorithm will not be able to tell the intentions or explain why or how a certain outcome came to be, also known as the black box phenomenon described in 4.5.3 In such case it would be a question of the intention from the programmer to set up such algorithms (Richards & Gummadi., 2018). As Danks and London

(32)

point out: “Although some algorithmic biases are neutral or even desirable, many are

problematic and should be mitigated” (Danks and London 2017).

3.5.4 Human Thought and Emotion

In conjunction with the section above on bias, findings by Danks and London (2017) discuss the different types of algorithmic bias which can raise different types of issues depending on the norms and standard one has. The findings also put out the point that situations like this are extremely complex especially because people are also complex. According to Dick (2019), the early automation attempts were based on human intelligence, where the goal was to replicate conjoining intelligent human behavior in the technology. Additionally, Wood (2015) states: “I

truly hope that no marketing automation tool will ever replace the human touch.”. Ultimately,

automation or AI should not purely replace any human interaction a business has with its customers. Racial or clustered threats through bias and recommendation engines could be of an illegitimate method, this highlights the distinction between the context of technology and the context of human justification. (Freed, 2020). The book AI Human thought and emotion by Freed (2020), walks through the middle ground of humans and technology co-operating. It discusses the idea that humans can share insights to provide philosophical and cognitive input into the technology in order to avoid flaws in the systems. Flaws like racial or gender bias. Humans can find errors and connections so that the technology learns to work alongside emotions and have a better cognitive understanding such as bias (Freed, 2020).

3.6 The Influence of Technology on Marketing

The heavy influence of technology has made marketers change from a traditional approach towards advertisements and realized that the competition has grown stronger. Because consumers are doing research online about products before deciding to purchase a product or service (Heil et al., 2010). As a result, all information online could affect companies in good or bad ways, since word of mouth still has a strong impact on whether a consumer will consider purchasing goods from certain companies. According to VanLaer et al., (2010), the impact of word-of-mouth has on the competitiveness among companies pushing to focus on keeping up with all the comments that are made from customers and how to make quick and proper responses in order to

(33)

satisfy the customer base. Therefore, companies have allocated resources to monitoring blog posts, social media content, etc., to compete with the digital market and take control of their brand equity.

3.6.1 Technology Affects the World

Bessen (2018) discusses the effect of AI on jobs, and the fluctuation in the nature of demand by the world is continuously changing, ever so more now with technological advancements. Additionally, Pettinger (2020, pp. 1–3) mentions the term industrial revolution when there have been significant changes in innovation that affect the global economy and the global society. E.g., In 1914 World War I and in the 1980s the launch of the World Wide Web. Pettinger (2020, pp. 1–3) explains the signs of a new industrial revolution because of the increasing development in technological infrastructure by making communication easier using digital platforms, the use of Artificial Intelligence and digitalization. Due to this, the borders between the physical world and the digital world are starting to emerge (Pettinger, 2020, pp. 1–3). The new era of technology that provides automation by using Machine Learning and Artificial Intelligence can be seen as a threat to job opportunities. According to Frey and Osbourne (2017), 47% of the jobs that can be automated in the United States are most likely at risk to become replaced by Machine Learning in the future. Likewise, Frey and Osbourne (2015) state that 53% of the jobs in the EU are at risk of becoming automated. However, the digital society creates opportunities for self-employment by using the digital market and removing geographical barriers. It can be argued that self-employment is one of the jobs that has increased since this industrial revolution. Because the generations growing up during- and into this era of technologically advanced society are already adapted towards self-employment by using social media that has enhanced the opportunity to become self-employed by creating content. The perception of job scarcity and machines taking over people's jobs has been an ongoing dilemma and people have been discussing this for centuries. The idea of being replaced by machines is not a new phenomenon. E.g., In 1985, Microsoft launched Excel and made standard bookkeeping easier to manage for everyone. This innovation decreased the number of bookkeepers. Microsoft Excel increased the desire for expertise from individuals that could manage the new technology and the importance of accountants became greater (Ip, 2017). This actively demonstrates that technology could replace

(34)

certain human jobs or processes but on the contrary, it could make people adapt to a different approach within that area.

3.6.2 Cases of Companies Implementing Artificial Intelligence

Mckenzie et al., (2013) mention that 35 percent of all the purchases done on Amazon.com in 2013 were from recommended products initiated by sophisticated personalization algorithms. Amazon is not only the world's largest online retailer by market cap it is also one of the top five biggest tech companies in the world possessing an enormous amount of data about its users (Lekkas, 2020). This creates a lot of data points that are used to make each visit to the website a unique user experience every time. Smith & Linden (2017) describes the user experience “It's as

if you walked into a store and the shelves started rearranging themselves, with what you might want moving to the front, and what you're unlikely to be interested in shuffling further away”. In

a catalog with millions of items, the algorithm will provide real-time recommendations depending on interests and the context, including previous behavior amongst similar users (Smith & Linden, 2017). It is not the similarities between the users but the correlation between the items consumed by the users that are of importance

3.7 Gaps in Literature

Through this extensive review of looking at the patterns in our data and connecting the different themes, the next step was to further investigate the existing research within the field, which opened a gap for further investigation. Reviewing the main findings in the frame of reference, the literature illustrates that while there is substantial research on Artificial Intelligence in relation to marketing. Hence, with the explosive growth of AI, there is an ongoing hype and

opportunity amongst businesses. This alarming rate of development around the hype could lead

to potential risks. Whilst AI takes the world by storm the influence on marketing can present problems; organizations or marketers who are not sure how to mobilize this technology properly. This study intends to contribute to the existing literature by providing practical implications for marketers to recognize the different entanglements that come with these opportunities; the risks and threats.

(35)

4. Findings & Analysis

The following chapter contains the empirical findings of this study. All the information gathered from the interviews will be presented in this section. Each section begins with a description of the topic: Sections 4.2 and 4.3 present the potential of AI for marketing and sections 4.4 and 4.5 present the dark sides of AI. The chapter will outline 4 key themes and how this might create scenarios that could be beneficial or detrimental for marketers. To clarify, the quotes in this chapter are presented in italics. The end of the findings and analysis related to the theme availability impact on marketing are discussed with a visual presentation.

4.1 Conducting the Findings & Analysis

One may argue that the greatest blessings of this generation are the data people have collected over the past couple of decades. AI has been here for centuries, but the definitions and scope of what it is have been ever-changing. Leaving professionals in these fields to be unaware of the endless possibilities of this technology. The term for AI has continuously changed, however, the first use of this term, “To proceed on the basis of the conjecture that every aspect of learning or

any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it” (Dick, 2019). The findings of this research comply with the fact that AI is

constantly evolving, several participants claim that the technology will flourish as we keep working on it.

Throughout the time of collecting data, four different themes were uncovered, namely; AI as an

opportunity, as a hype, as a risk, and as a threat. All the participants were asked what AI meant

to them, in which almost everyone responded it can mean many things depending on the context. Participant 5 explained AI as the science of making things smart and can be defined as human intelligence - done by machines, not only robots but programs to automate tasks.

In addition to this, Participant 6 means that the development of technology has enabled the industry to explore new things, such as machine- and deep learning which are still new phenomena. Hence, many areas within AI are yet to be explored and discovered. As of right

(36)

now, predictions can be made, which could help risk management. Some participants claim that the next step is if one can enhance this technology. That is why Participant 5 focuses on

knowledge representation which is something that telecom companies are working very hard for.

This is for example addressed in the concept of the basic human knowledge gap:

“If I show a picture to a kid, my 5-year-old, show him a picture, ask for what is this, this is a cat, and okay what is this, this is the water and I ask him a question, will cat drink water? and he will say JA, the cat will drink water. Then we will see a picture of a cat drinking water, but what he has in his mind is a cat, it is an animal, water... All animals drink water, for instance, he does not

need to see the cat to drink water before coming to this conclusion. This is what we call human knowledge, common sense and that is where the AI needs to evolve to the next step.” (P5)

Although this example of the child was never taught the relationship between a cat and water, the child knew there was a correlation. Just as technology, this knowledge representation could be beneficial for marketing. This example correlates with Chen (2019) who explains reinforcement learning as something that can create a learning environment for the computer in order to teach the computer about which approach is best for different situations. This additional technology could be the next step to prove even more how AI can be seen as an opportunity. Nevertheless, the findings of the research comply with previous studies highlighting that implementing AI into marketing comes with both benefits as well as dark sides, there are always two sides to a coin.

(37)

Table 2: Overview of how the initial codes emerged to the four themes

4.2 AI as an Opportunity

To begin with, this theme incorporates answers related to those stated by all the participants when asked about their view on AI, and how it has impacted the efforts in marketing or business. All of the participants were aligned with the idea that AI provides a new opportunity for marketing as well as for various roles at Ericsson amongst other companies. However, when asking directly about AI and the potential dark sides, all respondents identified AI partly as a

threat, as it might be damaging in terms of brand reputation, have legal implications, narrow

Marketers are Watching You : An exploration of AI in relation to marketing, existential threats, and opportunities