Artificial intelligence in services: current trends, benefits and challenges

ABSTRACT

The automation of services taking advantage of the significant opportunities offered by artificial intelligence and other Industry 4.0 technologies is receiving increasing attention both from academics and practitioners. Interest in the subject has been boosted significantly by the healthcare crisis generated by COVID-19 and the need to maintain social distancing while continuing to provide efficient services. The purpose of this brief paper is threefold: (i) to introduce and summarize the current state of automated forms of interaction in services; (ii) to provide an overview of the six papers published in this special issue; and (iii) to describe the possibilities for future research that emerged at the AIRSI2019 (Artificial Intelligence and Robotics in Service Interactions) Conference. The AIRSI2019 conference was the precursor to this special issue and provided an excellent opportunity to explore with leading international researchers the extraordinary development possibilities presented in this research context.

摘要

利用人工智能和其他工业4.0技术提供的重要机会进行的服务自动化正受到学术界和从业人员越来越多的关注。由COVID-19引发的医疗危机以及在继续提供高效服务的同时保持社会距离的需要极大地推动了人们对这一主题的兴趣。这篇短文的目的有三点。(i)介绍和总结服务中自动交互形式的现状；(ii)概述本特刊发表的六篇论文；(iii)描述AIRSI2019（服务交互中的人工智能和机器人技术）会议上出现的未来研究的可能性。AIRSI2019会议是本特刊的前身，为我们提供了一个极好的机会，与国际领先的研究人员一起探索这一研究背景下所呈现的非凡发展可能性。

KEYWORDS:

• Artificial Intelligence
• Service Interaction
• Robot
• Virtual assistant
• Agenda for future research
• Consumer behaviour

关键字。:

• 人工智能
• 服务互动
• 机器人
• 虚拟助理
• 未来研究议程
• 消费者行为

Introduction

As with previous technology-based revolutions (e.g. industrial, digital), the implementation of artificial intelligence in business may radically transform the marketplace (Bock et al., 2020), and have an important impact on economies and employment. This form of automation will not only replace manual jobs, but also those involving analytical, intuitive and empathetic skills (Huang & Rust, 2018). In a further and more challenging step, automation has been used recently to interact directly and physically with customers in frontline services, which is shaking up service delivery and customer-firm relationships in various sectors such as banking (e.g. Belanche et al., 2019; Flavián et al., in press; Jung et al., 2018), hospitality and tourism (e.g. Akdim et al., 2021; Belanche et al., 2020a; Byrd et al., 2021; Romero & Lado, 2021), service delivery (e.g. Ivanov & Webster, 2021) and healthcare (e.g. Foster, 2018; Wirtz et al., 2021).

The use of automated forms of interaction in services is an innovation that can affect customer choices (e.g. Van Doorn et al., 2017) as well as customer experience, service quality and productivity (Wirtz et al., 2021). Therefore, as Bock et al. (2020, p. 317) suggested, ‘the potential for disruption by AI is particularly high in services’. However, although automation has clear benefits in some contexts (e.g. product transportation), the results of its use in social settings and in replacing human interactions (such as in services) are as yet less obvious. For instance, as Mende et al. (2019) noted, it is unclear whether AI services (i.e. humanoid robots) will create positive or negative consequences for consumers and companies. Similarly, Puntoni et al. (2021) acknowledged not only the benefits that AI can provide to consumers but also the costs consumers can experience when interacting with AI.

In this respect, recent studies have focused on one mainstream: the analysis of the determinants of customer adoption of AI in services. Specifically, in the context of service robots, it has been found that design characteristics (e.g. anthropomorphism), customer characteristics (e.g. age, gender) and service encounter characteristics (e.g. degree of user involvement in the decisions adopted by the system) are determinants of customer acceptance, satisfaction, intentions and behaviors (e.g. Belanche et al., 2020a, in press; Flavián et al., in press; Mende et al., 2019). These findings are consistent with the framework proposed for service robots by Belanche et al. (2020b). This model can be generalized to the context of AI (see Figure 1). In particular, we can argue that consumer acceptance, satisfaction, intentions and behavior will be conditioned by: (1) the AI system characteristics (e.g. AI-based system design features; notification or omission of notification that the user is interacting with an AI-based system, etc.); (2) the consumer characteristics (e.g. consumer's predisposition towards new technologies, age, etc.) and (3) and the features of the service offered (e.g. how much information is provided to the user and how it is provided, to what extent the user should be involved in the decisions taken, etc.). For a more exhaustive analysis and detailed understanding of the potential of this model, see Belanche et al. (2020b).

Figure 1 .#Three-part framework for AI services (adapted from Belanche et al., 2020b).

However, the main contributions to this emerging field are mainly theoretical (e.g. Bock et al., 2020; Huang & Rust, 2018, 2021; Puntoni et al., 2021; Van Doorn et al., 2017); consequently, there is a need to confront experts’ predictions with evidence obtained from the use of automation in frontline services. Accordingly, this special issue addresses the increasing interest in implementing automated forms of interaction in services (e.g. Paluch et al., 2020) with the aim of advancing the current understanding of the use of artificial intelligence in the context of services industries, focusing not only on its positive side but also on the potential barriers to its adoption and its potential drawbacks.

The articles in this special issue

Broadly, this special issue, which contains six papers, offers unique perspectives into the potential drivers (e.g. AI characteristics [anthropomorphism], customer expectations [hedonic, utilitarian] and emotions, social factors and service situations) and barriers (e.g. data breaches, service failures and failed interactions, ethical issues) of implementing artificial intelligence in services. Focusing on different service contexts (travel and hospitality, financial investments, food delivery), the studies provide insights into AI services in general as well as into specific technologies such as chatbots, service robots and virtual agents.

Focusing on potential drivers, the first paper, by Vitezić and Perić (2021), examines the willingness to accept AI devices by Generation Z. Specifically, these authors found that hedonic motivation (over other factors such as anthropomorphism, effort expectancy, performance expectancy and social influence) had the greatest effect on Gen Z members’ emotions and, subsequently, their willingness to use AI devices in hospitality. In addition, they found that the frequency of smartphone usage exerts a moderating role on the link between the perceived effort of using AI and emotions. Theoretically, the study contributes in several ways to the previous literature by its focus on Gen Z and its examination of the roles of a wide set of potential drivers of willingness to use AI devices and the moderating effect of the frequency of smartphone use. Practically, the study offers guidance to AI designers and business managers for the design and implementation of AI devices in a hospitality context.

In the second paper, Huang and Kao (2021) draw attention to the concept of social distancing, which has been widely used to limit the spread of the COVID-19 virus. Specifically, this research analyzes the influence of customers’ evaluations of social distancing on the use of chatbot services, as well as what factors influence these evaluations. The research, which combines an experimental design and structural equation modeling, found that attitude toward social distancing is especially important for determining the perceived usefulness of chatbots when they are being used for utilitarian. In hedonic service situations, the subjective norm of social distancing becomes more important. Finally, the customer’s fear of being infected influences his/her attitude both toward social distancing and the subjective norms of social distancing. Theoretically, the paper follows the theory of reasoned action (Ajzen & Fishbein, 1975) to extend our knowledge of the use of chatbots by examining the current context of social distancing driven by the COVID-19 pandemic, as well as the nature of the service provided. Practically, the study offers governments guidance as to how customers’ evaluations of social distancing are formed and suggests different strategies to increase chatbot use based on the hedonic/utilitarian nature of the service.

Castillo et al. (2021) analyze several barriers that may limit the use of chatbots, while noting that AI-powered interactions can fail. This may cause anger, confusion and customer dissatisfaction, leading to a process of value co-destruction. Following an exploratory approach based on in-depth interviews, the authors identified five antecedents of failed interactions (authenticity issues, cognition challenges, affective issues, functionality issues and integration conflicts) and found that these problems are largely attributed to the service providers. As a result, this paper contributes theoretically by offering a better understanding of value co-destruction in AI-powered service settings, attributions of resource loss and subsequent customer coping strategies. Practically, service managers may benefit from this paper by its insights into how to avoid value co-destruction in AI service settings.

Similarly, Huang and Philp (2021) focus on AI service failures by analyzing some potential negative consequences such as consumers’ propensity to share negative word-of-mouth. First, through three experiments carried out in different service contexts, these authors confirmed that consumers are less likely to share negative word-of-mouth about a service failure caused by an AI recommendation system than they are to a service failure caused by a human employee, even when the failures are identical and with no difference in firm blame and dissatisfaction with the failure. A second study suggests that this effect may be explained by consumers’ perceived connection with the AI. The paper offers interesting implications for both theory (e.g. a better understanding of consumer-AI interactions and the negative consequences of service failure when the service is provided by an AI system or a human employee) and practice (e.g. how to implement AI effectively in company’s service offerings).

Chen and Jai (2021) highlight a critical current concern with AI services, that is, how to protect consumers’ personal data. Focusing on hotel data breaches, the authors analyzed whether customers belonging to loyalty programs would have different perceptions regarding the severity, vulnerability, trust depletion and crisis response strategies of the organizations to deal with data breaches than would customers outside the program. They found that there were no significant differences in perceived vulnerability and severity between the two groups, but that loyalty program members demonstrated a greater loss of trust toward the organization. Theoretically, this study adds to previous works by providing a better understanding of the role of customer relationships and service recovery after a data breach. Practically, the study offers guidance on customer relationship management and how to respond after a data breach crisis.

The special issue closes with an interesting paper by Professor Russell W. Belk that highlights the need for thoughtful reflection on the role that artificial intelligence and robots might play in various controversial contexts. Belk (2021) first highlights the main ethical issues that have emerged as service robotics and AI services have developed. Specifically, Professor Belk raises five issues: (1) ubiquitous surveillance; (2) social engineering; (3) military robots; (4) sex robots and (5) transhumanism. According to Professor Belk, it is imperative to research and address these five issues now, as they will raise additional concerns in the future and will have serious consequences as these technologies develop further. The paper bridges this gap by addressing some of the ethical dilemmas of robotics and AI and offers key implications for public policy and the application of these technologies in services.

Concluding remarks and emerging themes for further research

The articles included in this special issue offer important insights to service managers by identifying some of the key drivers of, and barriers to, customer acceptance, depending on the service situation, that can help them to successfully implement AI services. On the one hand, effectively managing AI characteristics, customer expectations and emotions, and social factors may increase customers’ willingness to interact with AI and use AI services. On the other, a better understanding of the main barriers to AI services implementation may reduce potential negative consequences such as value co-destruction, trust depletion and negative word-of-mouth.

It is certain that the future evolution of the service sector will depend on the appropriate exploitation of the possibilities offered by artificial intelligence and other Industry 4.0 technologies. The challenge we face consists of taking advantage of the opportunities offered by these technologies, not only in terms of the organizations providing the services but also in terms of the end users, who will enjoy them and must be willing to use them. The guest editors of this special issue hope that these results will also inspire and motivate future studies so that we can continue moving forward in this emerging field of research.

Finally, again with the objective of stimulating further research in the field, we would like to present a list of emerging research themes, various future research questions that new studies should address, and propose specific topics that should be analyzed in each case (see Table 1). These emerging themes were thoroughly addressed, discussed and organized in a comprehensive manner by a think tank developed within the framework of the AIRSI 2019 Conference.

Table 1. Some issues to be addressed in future research on the application of AI to services.

Research themes	Research questions	Possible research topics
AI Design Challenges	How should affect/emotions be integrated into AI services?	✓ Capacity to identify emotions ✓ How to design robot personality ✓ Appropriate level of robot proactivity ✓ How to adapt to customers ✓ Investigate possible differences due to individual characteristics or service context
	Which additional elements should be incorporated into AI to provide added value?	✓ Analyze the optimal level of anthropomorphism ✓ Investigate other physical capabilities ✓ Appropriate level of empowerment ✓ Investigate possible differences due to individual characteristics or service context
AI and Employment	Should firms replace employees with AI initiatives?	✓ In which services may humans still be needed (e.g. babysitting)? ✓ Analyze the consequences of possible government regulations, law
	Do customers prefer employee-AI collaboration rather than substitution?	✓ Identify differences due to the service task ✓ Further understanding of the strengths and limitations of both systems ✓ Identify customer segments. Should firms give the opportunity to customers to choose?
Customer Experience with AI	At which points of the customer journey should customers interact with AI?	✓ Understand customer preference for AI services to a greater extent: Why do customers prefer AI services? Which customers? ✓ Analyze customer responses to AI services in different stages (e.g. post-sale, service recovery …). When do they prefer interacting with AI services?
	How can AI initiatives create value for customers?	✓ Provide a deeper understanding of the key factors (e.g. customer-based, service-related, technology-related …) that influence value creation through the use of AI services ✓ Analyze the value of AI services for existing and new customers: Is AI more effective for retaining existing customers or attracting new customers? ✓ Analyze AI initiatives in both B2C and B2B contexts ✓ How to deal with privacy concerns arising from the use of AI to analyze customer data
Ethics	Who is responsible for AI failure?	✓ Analyze the responsibility attributed to company (owner), developer, government, robot itself, society, customers …  ✓ Evaluate the perceived stability of the failure ✓ Identify possible differences depending on service task
	What are the possible harmful consequences of implementing AI services?	✓ Identify not only physical but also potential psychological harm (e.g. feelings of loneliness, stupidity) ✓ Analyze potential inequalities (e.g. loss of jobs, robot owner profits) ✓ Investigate the consequences of potential dehumanization ✓ Identify other unexpected consequences

The AIRSI 2019 Conference, which was held at the University of Zaragoza in December 2019, offered an opportunity for participants to present their work and discuss emerging themes in artificial intelligence and robotics in service interactions. Table 1 presents some challenges we need to address to properly integrate artificial intelligence into service design, maintain an appropriate balance between artificial intelligence systems and human employees, understand how the user experience may be affected, and examine the various related ethical issues. A closer examination of these aspects may serve to deepen our understanding of AI services and to offer more precise implications for practice. In addition, these topics may be relevant for a wide range of specific applications such as chatbots, robots delivering room service in hotels, new drone or robotized delivery options, financial robo-advisors, AI-based big data analyses and sex robots.

Last, but not least, we would like to thank Professor Levent Altinay, Editor-in-Chief of The Service Industries Journal, for the opportunity to host this special issue in the journal, which is enjoying growing recognition and scientific impact. Similarly, we would like to express our appreciation to the keynote speakers who presented at AIRSI2019, and all the participants for sharing their ideas and knowledge, which have been the precursors to this special issue. Special thanks go to Professors Roland Rust (University of Maryland, US), Ming-Hui Huang (National Taiwan University, Taiwan), Jeroen Schepers (Eindhoven University of Technology, The Netherlands), Linda Hollebeek (Montpellier Business School, France) and Russell W. Belk (York University, Canada). Finally, we would like to thank the anonymous reviewers for providing constructive as well as timely feedback during the review process, which was highly appreciated by the authors and the special issue guest editors.