The Impact of Artificial Intelligence (AI) on Customer Relationship Management: A Qualitative Study

Ever since the commercialization of the Internet in the '90s, technology has been evolving faster than ever with the advent of cloud computing, social media, ubiquitous mobile devices, Internet of things (IoT), blockchain, and more. A staggering number of three billion internet users, five billion mobile users, and six billion devices are now connected through this massive global network of networks, facilitating customer information exchange and interaction never before seen in history. Driven by recent technological advances in computing power, big data, high-speed internet connection and easier access to models built with advanced algorithms, Artificial Intelligence (AI) is the next wave of innovation, which has already come into widespread awareness in the consumer world with the emergence of virtual assistants and chatbots (e.g., Amazon's Alexa, Apple's Siri, Google's Assistant), image recognition (e.g., Facebook Photos, Google ImageNet), personalized recommendations (e.g., Netflix, Amazon) and autonomous driving (e.g., Tesla, Google Waymo). This qualitative research study intends to learn about the impact of AI on customer relationship management (CRM), specifically in the area of customer service of problem resolution. Most prior research focuses on the AI technologies leveraged in CRM systems, such as machine learning, natural language processing, voice recognition, chatbots, data analytics, and cloud infrastructure. Few extant studies have used a qualitative research methodology to gather data from industry experts to truly understand the impact of AI technologies on customer relationship management, especially in the area of customer service and problem resolution. This study aims to fill this research gap. This research contributes to the literature on AI in the context of CRM and is of value to both academics and practitioners as it provides a detailed analysis and documentation of the impact of AI on the customer service domain

also better interact with the interviewees through an emergent design.In addition, the qualitative research provides an opportunity for us to examine existing theoretical and conceptual foundations in the AI-driven CRM context.The central questions of the study are designed to analyze the general, industry, and firmspecific impact of AI on customer service and problem resolution.
1) What are the most remarkable features of AI used in the area of customer service and problem resolution? 2) How did these AI features come about and what factors drove the implementation of these AI features?3) What were the key factors that made the AI features impactful in the area of customer service and problem resolution?What were the key barriers that had to be overcome?4) What is the nature and pace of change in customer service and problem resolution that these AI-driven CRMs have produced?Will this pace of change slow down, remain the same, or accelerate over the next decade?5) What role has the specific respondent's firm played in the area and how has it assisted other firms?What key accomplishments and challenges have been encountered?6) What are the specific steps or programs the firm has taken (or helped other firms) to support customer service and problem resolution using AI technologies?Also, the main threats posed by AI are job losses as a result (Siau & Wang, 2018; Siau & Yang, 2017) and the skills of these professionals may also shift from a series of hard-selling or simple customer service to soft skills such as relationship building and the emotional connectivity (Singh et al., 2019).Therefore, this study also aims to learn the impact of AI on customer service job loss and change of job roles.1) What is the firm-specific or industry-level impact of AI on service job loss and change of job roles of customer service personnel?2) How would AI change the job roles and required skills of customer service professionals?The qualitative depth interview employed by this study is based on McCracken (1988) four-phase implementtation framework, which calls for a careful review of "analytical categories and relationships" and "cultural categories and relationships".This framework (Fig. 1) has been adapted by (Chakravarti & Crabbe, 2019) to fit a variety of ethnographic methods.

Analytic Categories and Concepts
The analytic categories establish the domain of investigation & organize the knowledge in the domain.
In the specific domain of  Table 1 explains the analytic categories and concepts of AI technologies most relevant to the customer service and problem resolution context.Some of these key technical terms appeared frequently in the answers from various informants during the interviews.AI can be further classified into Weak and Strong AI.Weak AI, also known as narrow artificial intelligence, focuses on specific narrow tasks such as customer service and problem resolution.The strong AI, or general AI, where machines acquire cognitive and self-aware capabilities similar to human intelligence, will take years to appear in CRM systems.The term AI is widely used in the CRM space.AIdriven CRM refers to CRM systems with various levels of AI capabilities for handling specific tasks.Machine Learning (ML) ML is the core driver of AI and involves computers learning from data with minimal programming.ML programs the desired outcome and feeds data to train the machine to achieve the outcome on its own (Salesforce.com, 2017).
All AI-driven CRMs are powered by machine learning, which analyzes and learns from massive amounts of customer data.

Deep Learning (DL) and Neural Network
DL is a subset of machine learning and uses complex algorithms that mimic the brain's neural network to learn a domain with little or no The reason for the recent surge in the use of AI in CRM is mainly due to relatively recent advances in deep learning, which creates outputs in much more sophisticated ways using multiple layers of interconnected neurons (called human supervision (Salesforce.com, 2017).

Natural Language Processing (NLP)
NLP uses machine learning techniques to find patterns within large data sets in order to recognize natural language (Salesforce.com, 2017;Mezgebe, 2020).
NLP is widely used in customer service applications such as chatbots and virtual assistants.It is also capable of doing sentiment analysis to understand how customers feel about products or services.Big Data is the raw fuel of AI, which includes large amounts of structured or unstructured information that provide the inputs for surfacing patterns and making predictions (Salesforce.com, 2017).
The recent explosion of high volume, high velocity, and high variety of customer data provide the ideal inputs for CRM machine learning models, which make these systems smarter and smarter.

Cultural Categories and Concepts
The cultural categories call for a detailed inventory of the key features of the researcher's experience with the focal topic (Chakravarti & Crabbe, 2019, p. 75).This also allows recognition and admittance of matches and mismatches with other cultural categories which may merge from the respondent's subjective experience (McCracken, 1988).As a former practitioner in the management and IT consulting industry, my personal experience tells us that it is crucially important to assess the cultural categories from the industry and functional area perspectives.In the domain of AIdriven CRMs, the following are the cultural categories identified based on the major stakeholders in the CRM ecosystem (the functional cultures) for this specific study.The depth interview questions and probes are designed to analyze the relationships among these entities.

Case Study
This refers to a research method, which is used to describe an entity that forms a single unit such as a person, an organization, or an institution.

Chosen Method
This qualitative study aims to conduct the personal interviews with four industry experts to gain insights into the impact of AI technologies on customer service and problem resolution.Therefore, the proposed research method for this study is an interactive depth interview which is a commonly used ethnographic data collection technique.Moreover, following Brinkmann, (2016) interview guidelines, this study has chosen the semi-structured interview method among the spectrum of different types of the interviews (i.e., structured, unstructured, & semi-structured).The semi-structured interview is defined as "an interview with the purpose of descriptions of life world of the interviewee in order to interpret the meaning of described phenomena" (Brinkmann & Kvale, 2015, p. 6).Compared with the structured and unstructured interviews, the semistructured interview has the following characteristics (Cohen & Crabtree, 2006), which fit the requirements of this study well.
1) The interviewer and respondents engage in a formal interview.
2) The interviewer designs and uses a formal interview protocol, which is a list of questions and topics that need to be covered during the conversation, usually in a particular order.
3) The interviewer follows the protocol but is able to follow topical trajectories in the conversation that may stray from the guide when he or she feels this is appropriate.

Interview protocol Preliminary Draft
The preliminary interview protocol was developed based on the central research questions of this study, a careful literature review of the use of AI in the space of customer service and problem resolution, and an assessment of the professional backgrounds of the respondents so that the scope of topics are appropriate for the interactive depth interview.In addition, elaboration probes are developed to the facilitate data collection according to the proposed analytic and cultural categories in section 3. The preliminary interview protocol was initially tested with a senior product manager with Coupa Software and responsible for delivering a roadmap for the search and shopping module within the larger Coupa platform suite.Based on the initial interview observations and the expert feedback, we made the following adjustments to the preliminary interview protocol.
1) For Q1, the time frame was specified to be "the last 10 years" of the AI implementation in CRMs to avoid any confusion about earlier AI features used in legacy systems; the scope of implementtation was specified to be the general application space for various businesses to avoid potential confusion with firm-specific implementation.2) For Q2, the question was made more concise by removing "to the best of your knowledge", which was a given condition.3) For Q5, the question was modified to distinguish between the firm-specific versus industry-level impacts on jobs and change of job roles.4) The time checks were also adjusted accordingly based on the initial test observations.

Final Interview Protocol
The final interview protocol (Appendix A) is comprised of six questions in a particular order to facilitate the flow of the interview and also standardize the data to be collected from the respondents.Elaboration probes are developed in case topical trajectories in the conversation stray from the inter-view guide.Standard introduction and closing state-ments are also used to ensure the formality and consistency of the interview.Following Brinkmann's, (2016) guidelines, the final interview questions are designed with a purpose and refined to invite respondents to give descriptions of their experiences of and insights (lifeworld) into the AI-driven CRM space and seek to obtain data for interpretation of meaning.The key strength of this formal interview protocol is that it has adhered to the guidelines and best practices of the semi-structured interview.This allows the interviewer to be prepared and appear competent during the interview while giving the informants the freedom to express their views on their own terms (Cohen & Crabtree, 2006).With a standardized set of questions, it can provide reliable, consistent, and comparable qualitative data from respondents across different functional domains and various professional levels.In the meantime, the interactive elaboration probes offer flexibility in discovering analytic and cultural categories and an opportunity for identifying new ways of seeing and understanding the topic at hand.Besides the strengths mentioned above, there are also several weaknesses of this interview protocol.
First, the open-ended questions are difficult to analyze due to the varied responses from respondents.Second, the flexibility of using elaboration probes in the interview may reduce reliability, as respondents may not receive the same probe questions and it would be hard to compare answers.Third, in order to get enough data from respondents during the one-hour-long interview, the questions are designed to include the subparts, which makes it cumbersome for respondents to remember to answer all parts of a specific question.

Informant Selection Process/Criteria
In order to obtain reliable and useful data for this qualitative study, the informants of this study are selected based on the following criteria.
1) Who has relevant and the useful information?Informants should have extensive experience and knowledge in the AI-driven CRM space.2) Who is interested and available?Informants should be interested in the research topic and be available for a one-hour-long interview.3) Who is willing to provide reliable information?
Informants should have a good professional reputation and be willing to share their honest answers to the interview questions.
Based on these three key criteria, we selected four industry professionals for the final interview from my own professional network, which represents a good mix of professional levels (2 executive levels and 2 middle management levels), cultural domains (3 from the U.S. and 1 from China), functional sectors (3 from consulting firms, 1 from an e-commerce company).All four respondents have agreed to share their names in the study.

Nature of Material Collected
The materials collected from the respondents are mostly non-numerical, descriptive data and therefore, not appropriate for the quantitative analysis using advanced statistical tools.Most of the descriptive data are fairly current, with only a small number of the references to historical contexts.None of the materials collected are deemed classified, private use only, or confidential.Most of the materials collected from the respondents are experience-based while some of them are knowledge-based derived from other industry sources (e.g., Gartner, CIO Daily).Some materials are business-oriented in nature, especially during the discussions on the demand-side of the AI features.Some materials, on the other hand, are technical in the nature with many acronyms used.These technical acronyms are manually noted in the full interview transcripts.

Findings General Interpretative Approach
Qualitative data analysis entails certain distinct activeties.The first and most important one is ongoing discovery, which is about identifying themes and developing concepts and propositions (Taylor et al., 2015).Following Brinkmann, (2016) guidelines on semi-structured interviews, this study has adopted a general interpretative approach and conducted a broach thematic analysis on the interview transcripts.After reading and rereading data collected from the respondents, we kept track of the hunches, interpretations, and ideas by using Microsoft OneNote, which allowed us to write down ideas anytime and anywhere.We looked for broad themes and major topics that emerged from the interview transcripts and developed a coding scheme (Fig. 4) to analyze the data.One key observation that has emerged from the interviews is the dynamic relationship between the demand-side (CRM system customers/clients) & the supply-side (technology providers) of the implementation of AI features.Based on the coding template suggested by (Taylor et al., 2015, p. 184), I modeled my coding scheme to highlight the cultural and analytic categories that frequently appeared in the respondents' accounts of their experiences (Fig. 4).
The coding scheme is applied using a red font to some sections in the transcripts to demonstrate the process.
Fig. 4: Analysis: Coding Scheme for the AI-Driven CRM Space.

Emergent Themes Cultural Categories
The cultural categories emerged when all four respondents were talking about the key stakeholders in the AI-driven CRM space, especially the AI leaders, CRM vendors, CRM implementers, & CRM customers/ clients.One emergent theme is that CRM customers/ clients want AI features in the CRM system in order to achieve their strategic goals, such as the reducing operational costs and improving customer satisfaction.

Emergent Themes Analytic Categories
The analytic categories establish the domain of investigation and organize the knowledge in the domain.In the specific domain of AI-driven CRM systems, a plethora of AI technologies are identified and defined in Fig.The respondents have also converged on the several themes regarding the impact of AI on customer service and problem resolution, which include 1) Dramatically reducing the number of service personnel answering simple & routine customer inquiries; 2) Understanding each customer individually and providing personalized user experience to the achieve customer intimacy; 3) Offering fast & accurate answers and efficiently resolving problems; 4) Achieving higher customer satisfaction & the obtaining competitive advantages over competetors; 5) Ultimately realizing a satisfactory return on the investment (ROI) and maximizing profits; 6) Reducing the number of low-level customer service personnel and transitioning them to more value-added roles, e.g., virtual agent managers and trainers.The impact is more about role changes, rather than simple job loss.

Integrated Interpretation
The descriptive data collected from the respondents support the relationships of the cultural and analytic categories proposed in earlier sections.That is, a plethora of AI technologies are becoming increasingly accessible through AI leaders, and are leveraged by CRM vendors and implementers to embed or integrate into CRM systems as useful AI features, which in turn impact customer service and problem resolution.

DISCUSSIONS: Summary of Findings
This qualitative research study intends to learn about the impact of AI on customer relationship management (CRM), specifically in the area of customer service of problem resolution.The following are the key findings based on the qualitative analysis of the data collected through semi-structured interviews with four respondents who have the extensive industry experiences in this space.
1) Overall, AI adoption by businesses for various purposes has continued to gain momentum in recent years.With a focus on the customer service and problem resolution domain, the pace of adoption will accelerate in the next few years.
2) The most impactful AI features used in CRM systems include 1) virtual agents, chatbots and agent assist systems, which rely on voice/ speech recognition and natural language processing (NLP); 2) Next Best Action (NBA) systems, which leverage predictive data analytics; and 3) personalization technology, which is powered by advanced machine learning (ML) & the big customer data.The key factors that made these AI features impactful in the area of customer service and problem resolution include 1) cost savings from reducing the number of low-level service personnel; 2) faster and more accurate answers to customer inquiries; 3) personalized customer experience; & 4) overall improved quality of the customer service & customer satisfaction.
3) The factors that drive the implementation of these AI features are mainly two folds -1) from the demand-side perspective, an increasing number of companies want AI features in their CRM systems to the obtain both tangible and intangible benefits and achieve their strategic goals; 2) from the supply-side perspective, AI leaders, CRM vendors, implementers, especially, are making AI technologies and AI-driven CRMs more accessible to companies of various sizes.The key stake holders of AI-driven CRMs have formed an ecosystem, which is pushing the accelerated adoption of AI features in this space.4) AI has been replacing the jobs of low-level customer service personnel and some of the employees have transitioned to more valueadded roles, e.g., virtual agent managers and trainers.The impact is more about changing job roles, rather than simple job loss.5) All respondents stated that their firms have formal programs to embrace new technologies (or help other firms), including AI.There is also a formal process to evaluate the success of such programs.

Limitations & Future Research
This study has several limitations.First, the sample size of this study is very small and the informant selection process is not random, thereby limiting the generalizability of the findings.Second, due to the project time constraints, the four informants were selected from our own professional network.Further research could explore strategies to minimize biases linked to the informant selection, ensuring a more objective and comprehensive analysis.Two informants are from the same company and three informants are from the IT consulting industry, thereby increasing the chance of biased responses.A longer vetting process would help in this aspect.Moreover, for conducting an in-depth interview, the researcher's personal interviewing skills (e.g., proper use of elaboration probes) could affect the quality of data collected.The test interview helped us immensely before conducting the actual interviews.Lastly, the flexibility of the using elaboration probes in the interview may reduce the reliability, as the respondents did not receive the same probe questions, which made it difficult to compare answers.Future research could increase the sample size and explore strategies to minimize biases linked to informant selection, ensuring a more objective & comprehensive analysis.In addition, further research could delve deeper into the transition of job roles resulting from the AI implementation.Analyze how employees are adapting to new roles such as virtual agent managers and trainers, and assess the factors influencing a smooth transition.

CONCLUSION:
In Respondent 1 is a veteran with diverse IT industry experiences & helps customers make strategic decisions regarding their IT infrastructure, digital platforms such as web content management, portals, search systems, marketing platforms, and marketing systems.His focus is on the actual technology used in a customer experience environment, which is very relevant to the research topic of this study.It would be invaluable to get his insight into the industry-level and strategic-level dynamics of the AI-driven CRM space.

Respondent 1
My role at proficient is called chief strategist for customer experience platforms.So I deal on the technology side of anything that has to do with a customer touch point, so whether that's websites, a portal, or customer service in the background.Any of those places where a customer is going to interact with a company, I help companies with strategy around the technology to enable those interactions.We do a lot with AI across a wide range of areas from personali-zation, to chatbot, to targeting customer service agent, all those things.I've been with proficient for 16 years and I've been in this industry for over 30 years now.

INTERVIEW QUESTIONS
Q1. Think about the implementation of artificial intelligence (AI) features in CRM systems by businesses in the last ten years.What comes to mind as among the most remarkable features of AI used in the area of the customer service and problem resolution?What makes these AI features stand out in your mind?Elaboration Probes 1) At least three AI features implemented in customer service and problem resolution (e.g., chatbots) 2) Focus on the impact on customer service and problem resolution 3) Nature of impact 4) Note terms that the respondent uses to describe the scale and locus of impact.5) If needed, help respondents using an unrelated AI example (e.g., AI-driven fraud detection systems).

Q2. How did these AI features come about and what factors drove the implementation of these AI features?
Elaboration Probes 1) The three AI features identified in Q1.
2) Origins: Needs for operational excellence and customer intimacy 1.New development in AI technologies and big data 3) Implementation Drivers: 1. AI technology development (e.g., machine learning, deep learning, natural language processing, neural networks) 2. Industry actions and trends 3. Firm actions on the supply-side (e.g., R&D, operations, excellence) 4. Firm actions on the demand-side (e.g., customer satisfaction, service excellence) 5. Other Background General Literature Review of AI in the CRM Space Recent technological advances in Artificial Intelligence (AI) technologies, especially in the fields of machine learning, deep learning, neural networks and big data (Moreno & Redondo, 2016; Zhang et al., 2018), ubiquitous mobile computing (He et al., 2019) have fueled the growth of the next-generation digital platforms (Khalid et al., 2019; Rai et al., 2019; Zhang et al., 2019), which have progressively achieved the human (sometimes super-human) level of performance in various areas including autonomous driving, medical diagnosis (e.g., cancer screening), robots/ drones, chatbots, virtual assistants, language translation, governance monitoring (e.g., copycats, content violation), complex game playing and recommendation systems.AI features embedded in customer relation-ship management (CRM) platforms also create new possibilities for the customer experience, with rich insights into customer needs (Kumar Deb et al., 2018).These innovations have been driven by a manifold increase in processing power, lower-cost hardware, and the exploding creation and availability of customer data (Gantz et al., 2017; Hossain et al., 2022).
Taylor et al. (2015) recommend identifying themes by thoroughly exploring the data & propose the following steps to maintain focus.
Q3. What, in your opinion, were the key factors that made the AI features impactful in the area of customer service and problem resolution?What do now.I think we need to worry about what's going to happen in the next ten to tweeny years.I mean, what happens if my job is placed by robots?It's possible that robots can analyze the data and come up with a better answer.So I want to say that, at least everyone who work in the internet industry should worry about what AI can do in the future.Researcher Thank you again for your time and your insights.We truly appreciate your participation & willingness to share your views with us.Record response: [ X ] Yes -feel free to thank me by name [ ] No -do not identify me.Citation: Wang JF. (2023).The impact of artificial intelligence (AI) on customer relationship management: a qualitative study, Int.J. Manag.Account.5(5), 74-90.https://doi.org/10.34104/ijma.023.0074090

Table 1 :
Analytic Categories and Concepts.

Table 2 :
Rossman, 2014).It encompasses a variety of accepted methods and structures, including four major types which are the most commonly used.They are summarized in Table2.Qualitative Study Design Methods, Adapted from Astalin (2013, pp.120-122).
This refers to a type of qualitative research methodology that allows theory/theories to emerge from the data that is collected.It employs a systematic yet flexible process to collect data, code the data, make connections, and see what theory/theories are generated or are built from the data.
Respondent 1 is the Chief Strategist of the Customer Experience Platforms at Perficient, Inc.He has more than 35 years of the strategic technology advising experience.He and his team build great customer, partner, and employee experiences.His primary focus is on the actual technology used in the customer experience environment and takes into account his experience with digital platforms such as web content management, portals, search systems, the marketing platforms, and marketing systems.Looking to the future, he and his team are expanding their focus to understand how emerging technologies, such as the artificial intelligence (AI), relate to those customer experience platforms.It's quite a broad range of technology, but the common thread is they all connect with the customer.His primary role is to help his clients understand the technology that they need to have in place, why it's important, and how it integrates and affects all parts of their business.He has a B.S. in Computer Science from Purdue University and an M.B.A. from Cleveland State University (Adapted from respondent's Perficient profile and respondentsupplied information, 2019).Before joining Accenture, He was a senior consultant at Capgemini, where he focused on digital transformations using technologies such as NetSuite and Salesforce.He later shifted to talent and organizational effectiveness work with a heavy emphasis on mergers and integrations, talent learning and development, and change management.He has a B.S. in the Information Systems from Marietta College (Adapted from respondent-supplied information, 2019).
Respondent 1 Readiness Lead, responsible for monitoring and reporting on the progress/state of the overall transformation using various data analytics and metrics tracking tools.His current project is at a leading US utility company implementing new customer inter-action software across their regional call centers.
Taylor et al. (2015)-side factors that drive the implementation of AI features in CRM systems.From the supply-side perspective, one emergent theme is that CRM vendors (e.g., Salesforce, NetSuite, SAP) are leveraging the tools and platforms from AI leaders (e.g., Google, Microsoft, Amazon, IBM) that conduct fundamental research & development in AI technologies.These tools are increasingly accessible through PaaS (platform as a service).The CRM vendors either embed the AI capabilities in their own applications or integrate with the AI leaders' platforms (e.g., IBM Watson).According to respondent 1, the supply side is playing a predominant role in pushing the adoption of AI features in CRM systems.Another emergent theme from the supply side is that CRM implementers (e.g., Accenture, Deloitte & Touche, Perficient), usually partner with multiple CRM vendors and advise and facilitate the adoption of these AI-driven CRM systems based on the assessment of customers' needs.The availability of advanced AI technologies and the promotion and adoption campaigns by CRM vendors and implementers are key supply-side factors that drive the implementation of AI features in CRM systems.All four respondents predict that due to the dynamic growth of this ecosystem, the pace of change in customer service produced by AI-driven CRMs will accelerate over the next few years.However, she predicts that the acceleration will not continue all the way through a decade, though, because 80% of adoption will probably come in the next three to five years due to the current high demand for AI features in CRM systems.FollowingTaylor et al. (2015)guidelines, I constructed the following analytic diagram (Fig.5) to illustrate the dynamics of the key cultural categories.

about the implementation of artificial intelligence (AI) features in CRM systems by busi- nesses in the last ten years. What comes to mind as among the most remarkable features of AI used in the area of the customer service and problem resolution? What makes these AI features stand out in your mind?
Please take 3 minutes to talk about your background and current role in the organization, including how long you have worked here & the nature of your current professional responsibilities.
topic as they reflect your experiences.Although I will take notes during the interview, it will be very helpful if I have your permission to video record this conversation so that I accurately capture your opinions and input.To begin the interview, I would like to know a little bit more about you.

What, in your opinion, were the key factors that made the AI features impactful in the area of
Elaboration Probes None Thank you again for your time and your insights.We truly appreciate your participation and willingness to share your views with us.We would like to reaffirm our original assurance of confidentiality.Would you have any problems with us thanking you by name for your help in our finished report (without attributing any opinions)?Record response: [ ] Yes -feel free to thank me by name [ ] No -do not identify me.Finally, we will be very pleased to share a copy of our final report with you.The report should be completed by the end of this summer.Once again, thank you so much for your help.