Are Customers Ready for Intelligent Automation Service Without Humans? The Role of Customer Engagement and Emotional Intelligence

Abstract

Intelligent automation is increasingly deployed in service settings, yet its effects on customer loyalty remain unclear. This study examines how different degrees of intelligent automation; human-only service, partial automation with human support, and full automation without human involvement, shape customer loyalty, and investigates the mediating role of customer engagement and the moderating role of emotional intelligence. Using a field experiment in the hospitality context, the findings reveal a non-uniform pattern: full intelligent automation reduces customer loyalty relative to human-only service, whereas partial automation mitigates this negative effect and performs more favorably than full automation. These effects operate through selected dimensions of customer engagement, while other dimensions show limited or null effects. Emotional intelligence further conditions customer responses, highlighting important boundary conditions. Overall, the results demonstrate that the impact of intelligent automation on customer loyalty is contingent rather than uniformly positive.

Introduction

Intelligent automation involves the integration of artificial intelligence (AI), machine learning, and advanced robotics into service processes to enable autonomous, adaptive, and context-aware decision-making (Castillo et al., 2021). In this paper, intelligent automation is defined as AI-enabled autonomous service execution, in which digital systems independently carry out service tasks and manage interaction flows without continuous human involvement at the point of service. The AI-based self-check-in system automates decision execution and customer interaction sequences through predefined algorithms and interface logic, rather than through continuous machine learning or real-time model adaptation. This conceptualization is distinct from fully self-learning adaptive AI systems, which dynamically update decision rules and personalize responses through ongoing learning from accumulated data. While such systems represent an important frontier in service automation (M.-H. Huang & Rust, 2021), they are not the focus of the present study. Instead, we examine a widely deployed and managerially relevant form of intelligent automation that reshapes service encounters by reallocating execution and coordination tasks traditionally performed by frontline employees to AI-enabled systems (M.-H. Huang & Rust, 2021; Larivière et al., 2017). Following Castillo et al. (2021), we draw on this stream of research to frame how AI-enabled service interactions alter customer experiences and engagement, without implying that the system under study possesses advanced self-learning capabilities.

Intelligent automation signifies a paradigm shift in service delivery, harnessing advanced technologies including AI, big data analytics, machine learning, and robotics (Yao et al., 2022; Zhang et al., 2023). These technologies possess a remarkable capacity to acquire knowledge, adapt to dynamic circumstances, and improve their performance over time, leading to the automation of tasks previously carried out by human agents (Castillo et al., 2021). A growing body of research suggests that intelligent automation is expected to outperform human abilities in various domains over the next 10 years, with projections indicating that approximately 47% of jobs in the United States could be replaced by 2033 (Frey & Osborne, 2017; Grace et al., 2018).

A notable trend is the increasing role of intelligent automation in consumer-facing services, traditionally managed by frontline employees like sales assistants, hotel receptionists, travel agents, and financial advisors. AI and robotic systems are transforming these service encounters (Castillo et al., 2021). For example, some hotels use AI-based self-service technologies, such as kiosks for check-in/check-out, like Hilton’s virtual assistant Connie. In 2019, automated customer service agents or robots attracted the most global investment in AI, totalling USD 4.5 billion (Castillo et al., 2021).

These innovations are transitioning service interfaces from human-centered to a more autonomous space primarily governed by technology (Larivière et al., 2017). AI tools offer customers cost and time savings and a sense of contribution to the service creation process (Ho & Ko, 2008). The convenience and accessibility facilitated by intelligent automation systems have been proven to foster positive customer experiences (M. Nguyen et al., 2023). Given that AI technologies are available round-the-clock, self-service kiosks and robotic customer support have the capability to effectively address customer needs and inquiries at all times, surmounting the constraints of traditional service hours (Jiang & Wen, 2020; Xu et al., 2023). For instance, hotel front desk automation can consistently and warmly greet visitors and provide information without fatigue, leading to a superior service experience to their customers, leading to heightened satisfaction and loyalty (Castillo et al., 2021).

However, other research has shown that some customers are resistant to intelligent automation due to fear of highly intelligent machines and lack of control over autonomous devices (e.g. Gillath et al., 2021). AI technologies, which necessitate user input, can increase service complexity and potential failures (Hilton & Hughes, 2013; D. Nguyen, 2018). In such instances if customers invest more time and effort into self-service tasks, they are more likely to experience frustration and disappointment if the service is unsatisfactory (Grönroos & Voima, 2013; Harrison & Waite, 2015). This suggests that it may be preferable to use AI to assist rather than replace employees, serving as an extension of the workforce rather than eliminating human roles. This trade-off between positive and negative influences and the usage of AI as an assistant has not been studied. Further, in the intelligent automation literature, the primary emphasis has been placed on its acceptance rather than its impact on firm performance. The principal aim of this research is to fill this gap and investigate the impact of implementing intelligent automation services on customer loyalty. Specifically, we compare services delivered by human employees to those provided by AI, both with and without the presence of a human employee.

Furthermore, the exploration of the mechanisms underlying how intelligent automation can foster customer engagement with firms, and ultimately foster customer loyalty has received little attention. This study therefore aims to examine the mediating role of customer engagement in the relationship between intelligent automation and customer loyalty. By comprehending the degree to which customer engagement mediates the impact of intelligent automation on loyalty, valuable insights can be extracted into the underlying mechanisms that drive customer loyalty within the context of automation. Furthermore, this research aims to explore how emotional intelligence moderates the relationship between intelligent automation and customer loyalty. By investigating how individual variations in emotional intelligence influence customers’ reactions to intelligent automation, the study intends to provide a deeper comprehension of how emotional intelligence shapes customer perceptions, attitudes, and loyalty in the domain of automated service interactions.

Finally, we focus on the moderating role of emotional intelligence, the ability to perceive, use, understand, and manage emotions in oneself and others (Mayer & Salovey, 1997), which is anticipated to play an important role in the adoption of AI technologies.

Specifically, customers possessing higher emotional intelligence may have better control over their emotions and be more open to the benefits of intelligent automation. On the other hand, those with lower emotional intelligence might require additional assistance and personalized interactions to navigate and fully utilize the technology. Previous research has also shown that emotional intelligence can influence consumer food choices and product decisions as well as how consumers cope with service failures (Tsarenko & Strizhakova, 2013). Therefore, understanding how emotional intelligence influences reactions to technology-mediated interactions is crucial for enhancing service experiences and customer satisfaction (Prentice & Nguyen, 2020). This can subsequently result in enhanced service outcomes and increased customer loyalty.

Overall, this paper introduces an innovative framework for intelligent automation services, contributing an original perspective to the literature on service management. This addresses a significant knowledge gap by merging two distinct areas of study, intelligent automation and service strategy, in novel ways. The framework aims to elucidate customer preferences for intelligent automation, ultimately enhancing customer loyalty. As intelligent automation serves as the point of contact between firms and customers, this study examines customer engagement with firms during interactions involving intelligent automation, alongside different dimensions of customer emotional intelligence. Accordingly, this study conceptualizes intelligent automation as a consumer-facing service interface that shapes customer–firm interactions, rather than as a mechanism of workforce substitution. Our focus is on how automation alters the structure and quality of service encounters, thereby influencing customers’ service experiences, engagement, and loyalty. Although intelligent automation may have implications for employee roles and labor structures, these considerations lie beyond the scope of the present research. Instead, we examine intelligent automation as a key point of contact between customers and service organizations, emphasizing its experiential and relational consequences for customers.

We report the results of a field experiment that examines the practical impacts of intelligent automation on customer loyalty, moving beyond theoretical concepts. This approach enhances the empirical strength and real-world relevance of the research. The findings offer valuable insights for managers on when and where to implement automation or human service personnel, based on both theory and practical application. By addressing the effective integration of intelligent automation in service organizations, this research contributes significantly to academic literature and practical management. The results provide managers with reliable models to make informed decisions amid the growing interest in intelligent automation in business.

Literature Review

Intelligent Automation Services

Automation refers to the process by which a machine agent, often a computer, takes on a task that was previously performed by a human (Parasuraman & Riley, 1997). In the context of service personnel, Davenport and Kirby (2016) have specifically employed the term ‘automation’ to denote the utilization of intelligent technologies as a replacement for human employees. They assert that what distinguishes this contemporary transition from ‘conventional’ automation of repetitive manual tasks is the incorporation of AI to automate service-related functions.

The evolution of automation has taken a significant stride with the emergence of AI, ushering in a new era characterized by intelligent computer technology (Davenport & Kirby, 2016). This contemporary variant of automation goes beyond conventional approaches by integrating advanced learning capabilities, adaptability, and continual enhancement through data-driven experiences (Brynjolfsson & McAfee, 2014). While AI might not yet match human intelligence, it is showcasing remarkable cognitive abilities that hold the potential to replace humans in specific service-related tasks (Davenport & Kirby, 2016; Song & Kim 2022). This transformative shift in automation signifies a profound change in labor dynamics, emphasizing the importance of acknowledging and harnessing the unprecedented potential of AI in shaping the future of work. In this paper, we define intelligent automation as the application of technologies, particularly AI, to substitute human cognitive abilities, specifically those related to learning and problem-solving, for the completion of tasks that were previously carried out by humans (Castillo et al., 2021).

The implementation of intelligent automation services is often perceived as a straightforward cost-benefit evaluation (Ivanov et al., 2018). However, a crucial question remains: does intelligent automation align with corporate strategy and contribute to enhanced performance? In the intelligent automation literature, the primary focus has been on the acceptance or preference for intelligent automation applications (e.g. the acceptance of robot shopping assistant (Bertacchini et al., 2017), preference for robot staff, (Kim et al., 2021; Ng et al., 2020), or the perception of humanoid service robot appearance, (Choi et al., 2021)) rather than measuring the effectiveness of intelligent automation services (see Table 1). Only a handful of studies examine intelligent automation from a service aspect and its impact on firm outcomes. Based on our literature review, so far, only B. Huang and Philp (2021) have studied the service aspect of intelligent automation, but they focused on its ability to provide personalized recommendations. Little attention has been paid to the service aspect of intelligent automation and its impact on firm performance.

Table 1.

Literature Review on Intelligent Automation.

Authors	Method	Intelligent automation application	Aspect of intelligent automation application		Intelligent automation
			Acceptance/appearance	Service	Full	Absent	Partial	Mediator/moderator	Sample	Findings
This study	Field experiment	AI-based self check in	✓	✓	✓	✓	✓	Mediator: customer engagement Moderator: emotional intelligence	Hotel guests	The first time partial intelligent automation (the combination between intelligence automation and humans support) has been explored. Different dimensions of customer engagement mediate in different ways in the relationship between intelligent automation and customer loyalty. Customers’ emotional intelligence levels moderate their response to intelligent automation, influencing their preferences for human interaction or automated service.
Bertacchini et al. (2017)	Field experiment	robotic shopping assistant	✓		✓	✓			Consumers	Researchers study the intelligent automation ability to interact with consumers in a real-life setting. Findings indicate that consumer-intelligent automation interactions are useful and funny.
Choi et al. (2021)	Survey experiment	humanoid service robot	✓		✓	✓		Mediator: warmthModerator: human intervention	Amazon Mechanical Turk	The study found that humanoid robots are expected to exhibit more warmth than non-humanoid robots, leading to increased dissatisfaction when a process failure occurs. However, humanoid robots can independently recover from service failures through sincere apologies and explanations, while human intervention can help mitigate dissatisfaction caused by non-humanoid robots’ inadequate recovery.
Fuentes-Moraleda et al. (2020)	Content analysis	hotel reception robots, robot room service, robot luggage storing							Hotel guests	The study indicates that the majority of consumer-human service robot interactions were functional (e.g. information and room service, baggage handling), whereas the social-emotional interactions were less in number but produced the ‘wow’ effect when they occurred.
Kim et al. (2021)	Survey experiment	robot-staff in hotels	✓		✓	✓		Mediator: concerns on safety and social distancingModerator: subjective perceived threat	Amazon Mechanical Turk	The results of this study differed from previous research conducted prior to the COVID-19 pandemic. The significant moderating role of perceived threat in consumers’ preference for robot-staffed hotels indicated that the preference for robots was influenced by the global health crisis.
Kuo et al. (2017)	Expert panel and semi-structured interview							No		Human-like service robot interactions are significantly attractive in this industry.
B. Huang and Philp (2021)	Survey experiment	AI recommendation system personalization		✓	✓	✓		Mediator: Self-AI connectionModerator: Personalized AI	Amazon Mechanical Turk	Despite the fact that there is no difference in the failure, firm blame, or customer dissatisfaction with the failure, consumers are less likely to spread bad word of mouth after a service failure caused by an AI recommendation system than a human employee. This effect is generated by consumers’ connection with the AI that predicts their future preferences based on their past behavior.
Ivanov and Webster (2024)	Survey	AI’s involvement in decision-making	✓						General managers and department heads	Hotel managers tend to prefer delegating decisions to AI when those decisions involve low emotional intelligence or minimal interaction with customers or staff.
Niemelä et al. (2017)	Qualitative	Expectation about social service robot in a shopping mall							Consumers	Consumers and retailers perceive that the usage of an HSR, ‘Pepper’, is novel and positive as it can create a ‘pleasant atmosphere’ and ‘friendly service’ continuously.
Pinillos et al. (2016)	Field experiment	Humanoid service robot	✓		✓				Hotel guests	The research focuses on observing how hotel guests request information. Findings show that consumers interacted the most with ‘touchscreens’ when requesting information. Consumers were reluctant to interact with the humanoid service robot.
Sabelli and Kanda (2016)	Qualitative	Humanoid robot	✓						Consumers	The research focuses on consumers’ reactions among visitors when interacting with social robots in a shopping center. Findings showed feelings of ‘surprise’ (positive) when consumers interacted with a social robot.
Song and Kim (2022)	Survey, interview	Fashion robot advisor	✓		✓	✓			Students	The research demonstrates that fashion retail HSRs with cognitive and social intelligence and human appearance have significantly higher adoption rate with consumers with high innovativeness and technological self-efficacy.
Tuomi et al. (2021)	Semi-structured interviews								Senior executives of robotized hospitality businesses	The present study analyses the specific roles of intelligent automation in hospitality service encounters. Key findings indicate that human-robot collaborations in hospitality settings would positively affect consumers’ intelligent automation acceptance.
Yao et al. (2022)	Field experimentSurvey experiment	An intelligent dental robot or a room service robot.AI service robots as personal fitness trainers and the casual dining restaurant waitperson	✓		✓	✓		Mediator: risk aversion	Dental clinic or hotel guests	In credence-based service contexts, participants in the higher social classes were more inclined to choose robot services than those in the lower social classes.More notably, participants’ preference for robot services was mediated by risk aversion, which also moderated the interaction impact of social class and service type.
Zhang et al. (2023)	Survey	Emerging technologies	✓					Mediator: guest satisfaction	Hotel guests	Examine how hotel guests’ preferences for technology affect their behavioral intentions within the hotel setting as well as at the broader tourism destination.

When AI technologies replace even the most highly skilled human service providers (De Keyser et al., 2019), they can transform organizational capital assets previously controlled by the employer into human capital that is now under the employee’s control. On the other hand, intelligent automation services can lead to customer dissatisfaction if deployed incorrectly or for inappropriate reasons. An intricate phone tree, an online ‘bot’ incapable of handling specific situations, or the perception of being treated as part of an undifferentiated mass can all result in significant costs to the company, far exceeding the expenses associated with setting up an intelligent automated system.

Therefore, there is a need to explore the impact of the services offered by intelligent automation on firm performance, including customer loyalty. Decisions about applying intelligent automation services are undoubtedly complex, but they will benefit from effective frameworks. Essentially, automating a service involves transferring the knowledge of how to execute it from a human to a machine capable of performing the same action, whether it be cognitive or physical. Manufacturing has developed a pattern for turning ordinary physical operations into machines but not for turning sophisticated knowledge into machinery.

Intelligent Automation Services and Customer Loyalty

As service industries increasingly adopt intelligent automation, the implications for customer loyalty have become a central concern. While intelligent automation services such as service robots, chatbots, and AI-powered kiosks, promise operational efficiency, consistency, and scalability, they may also undermine the emotional and relational components of service interactions that are critical for fostering customer loyalty. By contrast, human-only service encounters rely on interpersonal warmth, empathy, and social bonding, which are known to be central to loyalty formation (M.-H. Huang & Rust, 2021). Human service employees play a vital role in building strong emotional connections with customers by offering empathy, understanding, and personalized attention (Wirtz et al., 2013). These interactions foster a sense of relational closeness, which strengthens affective trust and promotes long-term loyalty (Ball et al., 2006; Rajaobelina et al., 2021). Relative to such human-only interactions, full intelligent automation replaces relational value with task efficiency, potentially weakening the experiential foundations of loyalty. In contrast, intelligent automation services tend to deliver standardized interactions that, while efficient, often lack the warmth and adaptability associated with human touch (M.-H. Huang & Rust, 2021). Even though recent studies have shown that AI can provide personalized recommendations and enhance empathy (B. Huang & Philp, 2021; Kim & Hur, 2024) customers tend to prefer human service providers over those offering intelligent automation services (M.-H. Huang & Rust, 2021; Puntoni et al., 2021).

When customers interact with intelligent automation services, they might experience a reduced sense of connection with the firm compared to interactions with human employees providing the same service. Thus, full intelligent automation may erode relational and emotional engagement benefits that are naturally provided by human-only service settings. Human employees have the capacity to offer personalized services, providing tailored recommendations and forming authentic human connections with customers (Wirtz et al., 2013). In contrast, intelligent automation systems are programmed to deliver standardized service, lacking the personal touch that customers appreciate (M.-H. Huang & Rust, 2021). This difference in personalized interaction can result in a diminished sense of loyalty toward the firm (Ball et al., 2006).

Interacting with service robots can evoke feelings of uneasiness and threaten customers’ sense of human identity (Mende et al., 2019). Customers may experience a lack of trust and confidence in the capabilities of intelligent automation, particularly when it comes to complex or sensitive tasks (Tuomi et al., 2021). This can result in a reduced willingness to engage with intelligent automation services (Tuomi et al., 2021) and, leading to a decline in customer loyalty. Furthermore, the limited cognitive flexibility of intelligent automation systems may overlook the uniqueness and individual distinctions of customers (Longoni et al., 2019). Customers value being recognized and treated as individuals with specific needs and preferences. When intelligent automation fails to provide this level of personalized attention, customers may feel neglected and undervalued, thus reducing their loyalty toward the firm (Ball et al., 2006; Rajaobelina et al., 2021; Shanahan et al., 2019).

Social comparison theory suggests that individuals tend to differentiate their thoughts, abilities, and qualities from inanimate objects like machines and computers (Longoni et al., 2019). Customers perceive human employees as capable of adapting to unique circumstances and offering customized solutions, while intelligent automation is often perceived as rigid and lacking the ability to cater to individual needs (Longoni et al., 2019). This perception of inflexibility can contribute to a decline in customer loyalty as customers seek a more personalized and tailored experience (Tuomi et al., 2021). The implementation of intelligent automation may undermine the customer-firm relationship by diminishing the sense of connection, eroding trust, and neglecting the unique needs of individual customers. When customers feel their unique preferences are ignored or misunderstood, they may experience a sense of relational neglect, which negatively affects loyalty (Shanahan et al., 2019; Rajaobelina et al., 2021). Therefore, we hypothesize:

• H1a: Full intelligent automation services (vs. absent) will result in lower customer loyalty.

Partial intelligent automation services involves a unique blend of intelligent automation and human involvement. In this service model, tasks are primarily carried out by intelligent automation systems, but humans are present to offer support and assistance whenever required. Relative to full automation, partial automation retains the efficiency benefits of intelligent systems while preserving human support for relational, emotional, and adaptive requirements of the service encounter. The presence of humans in partial intelligent automation services will positively influence customer loyalty (Prentice & Nguyen, 2020). When customers encounter a service environment where intelligent automation is augmented by human support, they experience a sense of reassurance (Mende et al., 2019). The availability of human assistance provides a safety net, alleviating concerns related to potential errors or uncertainties that may arise during the service process (Mende et al., 2019). Moreover, the human element in partial intelligent automation services allows for personalized interactions and tailored assistance. Humans possess the ability to understand unique customer needs, empathize with their concerns, and provide customized solutions that go beyond the capabilities of intelligent automation systems (Verhoef et al., 2010). Such hybrid configurations support relational and personalization benefits that are absent in full automation, resulting in a more engaging and satisfying customer experience, increased loyalty, and positive word-of-mouth recommendations (Van Esch & Stewart Black, 2021; Verhoef et al., 2010). By combining the efficiency and precision of intelligent automation with the presence of humans, organizations can strike a balance that addresses the limitations of full automation while leveraging its benefits (Mende et al., 2019). Accordingly, partial intelligent automation is expected to outperform full automation in customer loyalty outcomes, as it offers both operational and relational value rather than prioritizing efficiency at the expense of human interaction. This balance, provided by partial intelligent automation services, will result in higher customer loyalty compared to scenarios where customers solely interact with full intelligent automation systems.

• H1b: Partial intelligent automation services will result in higher customer loyalty compared to full intelligent automation.

Partial automation represents an efficiency-oriented configuration with human backup that manages uncertainty and risk, whereas absent automation represents a relational human-centered configuration that emphasizes interpersonal connection and experiential value. Given these divergent value trade-offs, existing theory does not support a clear directional prediction, nor a dominance relationship between the two.

The Mediating Effect of Customer Engagement

Customer engagement is a multifaceted construct characterized by cognitive, emotional, behavioral, sensory, and social reactions of customers actively connecting with a company (Lemon & Verhoef, 2016; Verhoef et al., 2010). Customer engagement has been identified as a critical factor in driving customer loyalty (Wirtz et al., 2013). Engaged customers are more likely to develop positive attitudes, engage in co-creation activities, provide word-of-mouth recommendations, exhibit loyal buying behavior, and acquire knowledge-based behaviors (Lemon & Verhoef, 2016). Moreover, customer engagement fosters a sense of connection, trust, and loyalty between customers and the service organization (Brodie et al., 2011; Hoyer et al., 2010; Nambisan & Nambisan, 2008).

Recent work in AI has focused on personalization and AI emotions and empathy with the objective of enhancing customer engagement (N. T. K. Chi & Hoang Vu 2023; M.-H. Huang & Philp 2021; Kim & Hur, 2024; Teepapal, 2025). This research presents mixed findings on the role of AI-driven empathy and personalization in shaping customer trust and engagement. N. T. K. Chi and Hoang Vu (2023) found that when AI is perceived as capable of responding with empathy, it enhances customer trust. Similarly, Kim and Hur (2024) demonstrated that AI chatbots with personalized interactions and anthropomorphic features are perceived as more competent and warmer, which fosters empathy and increases users’ willingness to engage with the chatbot. However, not all studies report positive outcomes. Park et al. (2023) found that chatbot empathy and identity disclosure did not significantly increase users’ willingness to donate, suggesting that the effect of AI-driven empathy may be context-dependent. Likewise, Teepapal (2025), in a study of AI-enabled personalization in social media marketing, reported a positive effect on trust but no significant impact on customer engagement. These findings highlight that while AI’s empathetic and personalized capabilities can enhance trust, their influence on deeper forms of engagement may be limited or vary across contexts.

Intelligent automation services, while efficient and capable of performing programmed tasks, lacks the personalization and emotional connection that human employees can provide (Erdoğmuş & Cicek, 2012; Yao et al., 2022). Human employees can offer personalized services, such as personalized recommendations and tailored interactions, which can result in longer and more engaging interactions with customers (Vivek et al., 2012). Interactive and memorable experiences delivered by human employees have been shown to increase customer engagement with a service company (Vivek et al., 2012). Customers can be more engaged by human employees since they maintain lengthier engagement (Erdoğmuş & Cicek, 2012). In contrast, intelligent automation services are limited to standardized tasks and may not fulfill the diverse and evolving needs of customers. Intelligent automation is designed to perform standardized tasks based on programmed instructions, lacking the flexibility and personalized touch that human employees offer (Tuomi et al., 2021). Customers may feel less engaged with the firm when interacting with intelligent automation services. Lacking the personalization and emotional engagement provided by human employees will result in decreased customer engagement. The absence of personalized interactions and the limitations of intelligent automation may hinder the cognitive, emotional, behavioral, sensory, and social reactions (Song & Kim, 2022) that define customer engagement (Lemon & Verhoef, 2016). Lower levels of customer engagement are expected to translate into reduced customer loyalty. Thus, the implementation of full intelligent automation services, which is expected to decrease customer engagement by removing the personalized touch and interactive experiences provided by human employees, will lead to lower levels of customer loyalty. Therefore:

• H2: Customer engagement will mediate the effect of intelligent automation services, such that full intelligent automation services (vs. absent) will decrease customer engagement, resulting in lower customer loyalty.

The Moderating Effect of Emotional Intelligence

Emotional intelligence involves the capacity to receive, react to, and manipulate emotional information, even without fully comprehending it (Mayer & Salovey, 1997). It encompasses the ability to understand and manage emotions, regardless of whether one accurately perceives or fully experiences them (Mayer & Salovey, 1997). EI can be divided into four distinct components: emotional perception, emotion assimilation, emotional understanding, and emotion management (Mayer & Salovey, 1997).

Emotional perception refers to the ability to accurately identify and express emotions, as well as to differentiate between facial expressions and actual feelings (Mayer & Salovey, 1997). The second component, emotional assimilation, involves using emotions as a basis for reasoning and decision-making, recognizing their relevance in guiding important cognitive processes (Mayer & Salovey, 1997). Emotional understanding is another integral part of emotional intelligence, encompassing the capacity to identify and categorize emotions both linguistically and experientially (Mayer & Salovey, 1997). The final component, emotion management, pertains to the ability to effectively regulate and respond to both positive and negative emotions (Mayer & Salovey, 1997).

EI has the potential to significantly contribute to the prediction of customer acceptance of intelligent automation services (Mayer & Salovey, 1997). In a service encounter, customers with higher levels of EI can effectively connect with both employees and intelligent automation on an emotional level, thereby enhancing their overall service experience (Mayer & Salovey, 1997). This emotional connection subsequently influences customer engagement with the service organization and their loyalty, as EI enables individuals to understand and empathize with others.

While customer engagement explains how intelligent automation influences customer loyalty, it does not fully account for why customers differ in their responses to automated service encounters. Prior research suggests that individual differences in emotional capabilities shape how customers interpret, cope with, and adapt to service environments in which emotional and relational cues are constrained (Kidwell et al., 2008). Such differences are particularly salient in technology-mediated service contexts, where interactions are standardized and emotionally less expressive (M.-H. Huang & Rust, 2021; Puntoni et al., 2021). Accordingly, beyond its indirect effects through engagement, intelligent automation is likely to interact with customers’ emotional intelligence to influence customer loyalty.

Emotional intelligence reflects individuals’ ability to perceive, assimilate, understand, and manage emotions (Mayer & Salovey, 1997). Customers high in emotional perception and emotion assimilation are more sensitive to affective cues and rely more heavily on emotional information when evaluating service encounters (Kidwell et al., 2008; Prentice & Nguyen, 2020). Because intelligent automation typically provides limited emotional feedback and interpersonal responsiveness, such customers may experience greater emotional discomfort in fully automated service encounters, resulting in lower loyalty.

By contrast, customers high in emotional understanding and emotion management possess stronger abilities to cognitively interpret emotional situations and regulate affective responses (Tsarenko & Strizhakova, 2013). These capabilities enable them to adapt more effectively to emotionally constrained service environments, reducing frustration and allowing greater emphasis on instrumental benefits such as efficiency and convenience (M.-H. Huang & Rust, 2021; Longoni et al., 2019). As a result, emotional understanding and emotion management are expected to attenuate the negative effects of intelligent automation on customer loyalty.

Within the dimensions of EI, customers’ responses to intelligent automation vary, ultimately influencing their loyalty. Customers with lower levels of emotional perception and assimilation may understand their own and others’ emotions but struggle to regulate and manage them effectively (Prentice & Nguyen, 2020). Such customers are more inclined to seek interaction with human employees. If the intelligent automation system fails to deliver satisfactory service, they may struggle to cope with their emotions and attribute blame to the firm for its standardized service, ultimately diminishing their loyalty (Salovey & Mayer, 1990).

In contrast, customers with higher emotional understanding and management possess the ability to recognize and appropriately manage emotions (Prentice & Nguyen, 2020). Emotional intelligence is also a predictor of emotional resilience in the face of stress or unpleasant situations (Lee et al., 2017). Since emotional understanding and management involve regulating emotions to enhance cognition (Mayer & Salovey, 1997), customers with higher emotional understanding and management have greater control over their psychological state. They are more adapt at managing stressful or unpleasant events, investing effort to regulate their own emotions, and minimizing negative affective states. These customers are also skilled at alleviating emotional tension in others, which does not significantly impact their loyalty intentions toward the firm (Salovey & Mayer, 1990).

Together, these arguments suggest that emotional intelligence functions as a boundary condition that moderates customers’ responses to intelligent automation, rather than as a general predictor of loyalty. Accordingly, we hypothesize:

• H3a: Emotional perception negatively moderates the effect of full intelligent automation on customer loyalty, such that the negative effect is stronger for customers with higher emotional perception.

• H3b: Emotion assimilation negatively moderates the effect of full intelligent automation on customer loyalty, such that the negative effect is stronger for customers with higher emotion assimilation.

• H3c: Emotional understanding positively moderates the effect of full intelligent automation on customer loyalty, such that the negative effect is weaker for customers with higher emotional understanding.

• H3d: Emotion management positively moderates the effect of full intelligent automation on customer loyalty, such that the negative effect is weaker for customers with higher emotion management.

Figure 1 shows the conceptual framework of this paper.

Figure 1.

Conceptual model.

Method

Method Research Context

The field experiment was conducted with the cooperation of four hotels in Vietnam. Vietnam has welcomed a growing number of international visitors, 12.6 million in 2023 alone. However, hotels in Vietnam face acute staff shortages and rising wage pressures while still needing to maintain high service standards for a diverse and increasingly international guest mix. As a result, hotel management views intelligent automation as a dual strategy to contain costs and enhance service quality. Our study captures this real-world tension.

A list of hotels from the website hotel84 (2019) was used to approach all hotels in Vietnam. A phone call was made to check which hotels use intelligent automation for their AI-based self-service system. AI-based self-service system is a technology-driven solution that allows customers to check themselves in without the need for human assistance. It utilizes artificial intelligence algorithms and machine learning to automate the check-in process, enabling customers to input their information, verify their identity, and obtain necessary documentation or access to services independently. This self-service option enhances efficiency, reduces waiting times, and provides a seamless and convenient experience for customers. About twenty hotels were identified to use AI-based self-service systems and were invited to participate in a research study. Four hotels agreed to support the data collection. AI-based self-service system was chosen as the focus of this research project as it was a major alternative to having self-service without employee interaction, which made it an ideal context to compare human versus intelligent automation services.

Research Design

Three distinct check-in service modes were employed: full intelligent automation services, partial intelligent automation services, and absent intelligent automation services (see Appendix 1). The ‘absent’ intelligent automation services mode refers to situations in which customers are required to interact with hotel receptionists for all check-in steps. The ‘full’ intelligent automation services entails a scenario in which customers receive a notification indicating the presence of AI-based self-service systems near the reception area, prompting them to perform their check-in steps there. In this scenario, no staff members are present around the check-in area for assistance. The ‘partial’ intelligent automation services, serving as the control condition, involves a combination of AI and human assistance. AI-based self-service systems are positioned near the hotel entrance with a notification instructing customers to proceed with self-check-in using the AI-based systems. However, hotel staff are stationed nearby to provide assistance if needed. Each condition was implemented for 1 month across all four hotels simultaneously, with a planned 2-week break between each condition.

This study adopts a field experimental design in which experimental manipulation is introduced through naturally occurring variation in service delivery conditions rather than individual-level random assignment. Such designs are common in marketing and service research, where random assignment at the individual level is often impractical or would compromise ecological validity (List & Rasul, 2011; Popkowski Leszczyc & Rothkopf, 2010). By varying service configurations across time while holding the service context constant, this approach allows causal inference under real-world conditions while preserving the authenticity of customer experiences. To mitigate potential time- or context-related confounds, we implemented robustness checks with hotel and time fixed effects, which yield consistent results. In this field experiment, we adopted a between-subjects, time-based design across four hotels that had the infrastructure to implement AI-based self-service systems. The three experimental conditions (absent automation, partial automation, and full automation) were implemented sequentially across all four hotels, rather than concurrently. Each condition was applied uniformly across all sites for a 1-month period, followed by a 2-week break to allow for a reset in staff and guest expectations, recalibration of equipment, and preparation for the next phase. During these breaks, no experimental automation was deployed; instead, hotels returned to their standard service operations. This interval served two purposes. First, it provided for the necessary time to recalibrate equipment, brief staff on the next condition, and ensure consistency in deployment across all four hotel sites. Second, it also prevented guests from being exposed to multiple experimental conditions. As such, we did not exert influence over the check-in decision-making process; rather, we merely exposed hotel guests to different levels of automation under real-world service conditions. This approach is consistent with standard field experimental methods, where manipulation occurs through natural variation in service delivery rather than participant instruction or awareness (e.g. Popkowski Leszczyc & Rothkopf, 2010).

While our design did not allow for counterbalancing conditions across time, we carefully monitored potential time-related confounds. The number of guests remained relatively consistent across the three data collection phases, and no guests were repeated across conditions. Additionally, there were no major public holidays or special events that might have skewed customer expectations or satisfaction. Notably, one key strength of our approach is the elimination of self-selection bias: guests were unaware of alternative service configurations and could not choose among conditions, thus ensuring that each participant experienced only one version of the check-in process. This enhances the internal validity of the study while maintaining ecological relevance.

A paper-based survey was used to collect customer feedback on their experience with the check-in service. Due to hotel policies prohibiting digital survey distribution, this method was the only available option. Following the check-in procedure, hotel staff provided each guest with a sealed envelope containing the survey and an information sheet, inviting them to complete the survey at their convenience during their stay. Guests were instructed to either leave the completed survey in their rooms or return it at check-out. Specifically, guests could place the completed survey in a secure drop-box located near the reception desk or hand it directly to a designated staff member. Hotel staff facilitated the distribution and collection process in a way that ensured guest privacy and minimized disruption to their stay.

Throughout the span of 3 months, approximately 2,400 questionnaires were distributed. Out of these, 905 completed questionnaires were returned. However, after a careful review, 433 questionnaires were excluded from the data analysis due to incomplete responses. As a result, a total of 478 questionnaires were considered for the subsequent data analysis.

Questionnaire

The questionnaire comprised 25 items related to customer engagement, 16 items concerning emotional intelligence, and four items targeting customer loyalty. All items were evaluated using a 7-point scale. Additionally, two manipulation check questions were incorporated to inquire about participants’ awareness of the check-in service provided by the hotel and their check-in date. Furthermore, several demographic questions were positioned at the conclusion of the survey.

Measures

Customer engagement was measured using the 25 items, 7-point semantic differential scale employed by So et al. (2016). Emotional intelligence was estimated using 16 items adapted from Law et al. (2004). Customer loyalty was measured using four items which were adapted from Zeithaml et al. (1996).

Participant Demographics

Men and women constituted 61.5% and 38.1%, respectively, of the sample. The majority of participants fell into the 26 to 35 age group (38.5%), followed by the 36 to 45 age group (27.4%), while only 20.5% were in the 18- to 25-year-old category. Participants exhibited a high level of education, with 73.2% possessing a bachelor’s degree or higher. Nearly 80% of respondents reported an income ranging between $20,000 and $79,000, while fewer than 8% had incomes below $20,000 or exceeding $100,000. In terms of marital status, 41.2% identified as single, while 58.6% were married. Notably, the data indicated that nearly 60% of respondents undertook travel for business purposes.

Results

For the manipulation check, participants were asked about their awareness of the check-in options: human staff only (absent intelligent automation), AI-based self-service only (full intelligent automation), and AI-based self-service with staff around (partial intelligent automation). Participants were also asked about their check-in date to compare this with the time of check-in mode conditions. All participants were aware of their check-in service modes correctly, as the date of check-in matched with the duration of check-in mode conditions, resulting in 146, 143, and 189 participants in three conditions of absent, full, and partial intelligent automation, respectively.

All constructs reported Cronbach’s alpha higher than .7. The reliability of the measurement items was evaluated using composite reliability (CR) values. The results indicated that all constructs had CR values exceeding the threshold of 0.7 (Hair et al., 2022). Additionally, the factor loadings of all measurement items exceeded 0.7. For convergent validity, the average variance extracted (AVE) values for each construct exceeded the threshold of 0.50, demonstrating the convergent validity of the research model. The AVE values were higher than the squared correlations of the corresponding pair of constructs (Fornell & Larcker, 1981), showing discriminant validity. Harman’s single-factor test revealed that the factor explained below the 50% threshold (Podsakoff et al., 2012); therefore, common method bias was not an issue in this study.

Customer loyalty was the dependent variable. We conducted one-way ANOVA across 3 conditions: full vs partial vs absent intelligent automation services. The results revealed a significant main effect for service modes (F(2, 477) = 678.476, p < .001). In particular, participants in the absent intelligence automation service (M_{absent automation intelligent} = 6.47; SD = 0.19) reported higher levels of customer loyalty than those in the full intelligent automation condition (M_{full automation intelligent} = 5.63; SD = 0.27); thus, H1a is supported. Participants in the partial intelligence automation service (M_{partial automation intelligent} = 6.09; SD = 0.12)), reported higher levels of customer loyalty than those in the full automation intelligent (M_{full automation intelligent} = 5.63; SD = 0.27); thus, H1b is supported.

We estimated a generalized linear model with cluster-robust standard errors at the hotel level to account for non-independent observations within hotels. Results indicate a significant main effect of condition on customer loyalty (Wald χ²(2) = 657.020, p < .001), demonstrating that treatment differences persist even after adjusting for hotel-level clustering.

As customer engagement is a multiple dimension variable, an exploratory factor analysis of the measurement model was conducted to further verify the dimensional structure. Five factors were generated that were aligned with the five dimensions of customer engagement in previous studies: interaction, absorption, attention, enthusiasm, and identification. These five factors of customer engagement were used for mediation analysis.

To test H2, we conducted a mediation analysis using Hayes’ (2017) PROCESS Model 4 with 5,000 bootstrap resamples. This model was used to examine the indirect effect of intelligent automation (full vs absent) on customer loyalty via five factors of customer engagement. The intelligent automation as the independent variable was set multicategory with full and absent intelligent automation as indicators and partial intelligent automation as the reference group.

The results (see Table 2) indicated that the indirect effect was negatively significant for intelligent automation services in four dimensions: interaction (β = −.03, SE = 0.01, 95% CI [−0.06, −0.01]), absorption (β = −.02, SE = 0.01, 95% CI [−0.04, −0.00]), enthusiasm (β = −.02, SE = 0.01, 95% CI [−0.04, −0.01]), identification (β = −.02, SE = 0.01, 95% CI [−0.05, −0.01]). The indirect effect was positively significant for human-based service in two dimensions: absorption (β = .02, SE = 0.01, 95% CI [0.01, 0.04]) and attention (β = .04, SE = 0.01, 95% CI [0.03, 0.07]). These findings offered partial support for H2.

Table 2.

Indirect Effect Intelligent Automation Services Mode on Customer Loyalty via Customer Engagement.

Mediator (M)	Independent variable (X)	X→M	X→Y	Indirect	Total
Interaction	Full intelligence automation service	−.71*	−.43*	−.03*	−.46*
	Absent intelligence automation service	.01	.40*	.00	.38*
Absorption	Full intelligence automation service	−.55*	−.44*	−.02*	−.46*
	Absent intelligence automation service	.34*	.35*	.02*	.38*
Attention	Full intelligence automation service	−.84*	−.46*	−.00	−.46*
	Absent intelligence automation service	−.62*	.36	.04*	.38*
Enthusiasm	Full intelligence automation service	−.30*	−.45*	−.02*	−.46*
	Absent intelligence automation service	−.16	.38*	−.00	.38*
Identification	Full intelligence automation service	−.45*	−.44*	−.02*	−.46*
	Absent intelligence automation service	.26*	.38*	.00	.38*

Note. Dependent variable = customer loyalty (Y).

*

P < .05.

Emotional intelligence, being a multi-dimensional variable, underwent exploratory factor analysis encompassing all emotional intelligence items. This analysis yielded four factors, which corresponded to those established in the existing literature: emotional perception, emotion assimilation, emotional understanding, and emotion management.

To test H3, we conducted a regression analysis with the following formula:

$$\begin{array}{l}\mathrm{Y}=\alpha +\mathrm{\beta }1Fullintelligentautomationservices\\ +\beta 2Absentintelligentautomationservices\\ +\mathrm{\beta }3Emotionalperception\\ +\mathrm{\beta }4Emotionassimilation\\ +\mathrm{\beta }5Emotionalunderstanding\\ +\mathrm{\beta }6Emotionmanagement\\ +\mathrm{\beta }7(FullintelligentautomationservicesxEmotionalperception)\\ +\mathrm{\beta }8(AbsentintelligentautomationservicesxEmotionalperception)\\ +\mathrm{\beta }9(FullintelligentautomationservicesxEmotionassimilation)\\ +\mathrm{\beta }10(AbsentintelligentautomationservicesxEmotionassimilation)\\ +\mathrm{\beta }11(FullintelligentautomationservicesxEmotionalunderstanding)\\ +\mathrm{\beta }12(AbsentintelligentautomationservicesxEmotionalunderstanding)\\ +\mathrm{\beta }13(FullintelligentautomationservicesxEmotionmanagement)\\ +\mathrm{\beta }14(AbsentintelligentautomationservicesxEmotionmanagement)\end{array}$$

The results in Table 3 show that interactions between full intelligent automation services and emotion assimilation (β9 = −.05, p = .012), between full intelligent automation services and emotion management (β13 = .10, p < .001) and between absent intelligent automation services and emotion perception (β8 = .05, p = .077) were significant. Emotional perception (β7 = .02, p = .472) and emotion understanding (β11 = .02, p = .404) did not interact with full intelligent automation services; emotion assimilation (β10 = −.02, p = .436), emotional understanding (β12 = .04, p = .113) and emotion management (β14 = .01, p = .589) did not interact with absent intelligent automation services. Therefore, H3d was supported, H3a and H3b were partially supported, but H3c was not supported (see Figure 2).

Table 3.

Moderation Effect of Emotional Intelligence.

Variables	β	SE	Sig
Constant	6.09***	.01	.000
Full intelligent automation services	−.46***	.02	<.001
Absent intelligent automation services	.36***	.02	<.001
Emotional perception	−.02	.02	.232
Emotion assimilation	.05	.02	.002
Emotional understanding	−.01	.02	.551
Emotion management	−.01	.02	.362
Full intelligent automation services × emotional perception	.02	.02	.472
Absent intelligent automation services × emotion perception	.05 †	.03	.077
Full intelligent automation services × emotion assimilation	−.05*	.02	.012
Absent intelligent automation services × emotional assimilation	−.02	.03	.436
Full intelligent automation services × emotional understanding	.02	.02	.404
Absent intelligent automation services × emotional understanding	.04	.02	.113
Full intelligent automation services × emotion management	.10***	.02	<.001
Absent intelligent automation services × emotion management	.01	.02	.589
R 2	.77
Adjusted R2	.76

†

p < .1. *p < .05. ***p < .001, customer loyalty as dependent variable.

Figure 2.

Moderation effect of EI.

Discussion

The outcomes of this study indicate that full intelligent automation led to lower customer loyalty compared to absent intelligent automation. This suggests that customers have a preference for increased interaction with employees, which in turn positively influences customer loyalty. Intelligent automation services tend to deliver uniform and standardized assistance, treating all customers uniformly. Conversely, interactions with employees are more inclined to enhance positive sentiments toward the hotels, thereby rendering the hotel touchpoint more memorable. Consequently, customers who receive services from employees are more prone to exhibit higher levels of customer loyalty than those who experience standard service provided by full intelligent automation.

This paper marks the initial exploration of partial intelligent automation. The study revealed that partial intelligent automation yields higher customer loyalty compared to full intelligent automation. The inclusion of human agents in partial intelligent automation services fosters a sense of psychological security for customers. The awareness that human assistance is readily available instills confidence and trust in the service. This psychological reassurance can significantly impact their perception of the service provider and consequently elevate their loyalty. Partial intelligent automation introduces greater adaptability and flexibility in meeting customer needs. While intelligent automation excels in efficiently executing predetermined tasks, it may encounter challenges in addressing unique or intricate situations that demand human judgment, empathy, and problem-solving capabilities. The presence of human support in partial intelligent automation guarantees that customers’ distinctive requirements and preferences receive adequate attention, thereby amplifying customer loyalty. We do not advance a directional hypothesis comparing partial intelligent automation with human-only service because these configurations involve qualitatively different value propositions rather than a monotonic change in automation intensity. Partial automation combines efficiency and standardization with human support, whereas human-only service emphasizes relational interaction and social presence, activating distinct and potentially offsetting mechanisms. As a result, their relative effects on customer loyalty are theoretically ambiguous ex ante (M.-H. Huang & Rust, 2021).

Historically, research has predominantly concentrated on the merits and difficulties linked to either full or absent intelligent automation (e.g. acceptance of robot service, Yao et al., 2022). The concept of partial intelligent automation, which harmonizes the strengths of automation and human interaction, has not been widely investigated. Consequently, this discovery augments comprehension regarding how varying levels of automation can impact customer loyalty, thereby introducing fresh insights into achieving an optimal equilibrium between automation and human engagement.

This paper’s significant contribution lies in its exploration of the mediating role of customer engagement across its five dimensions: interaction, absorption, attention, enthusiasm, and identification (So et al., 2016). The study’s findings also unveil that customer engagement mediates the influence of intelligent automation services on customer loyalty through two divergent pathways. Regarding full intelligent automation services, all dimensions of customer engagement, excluding attention, negatively and indirectly impact the relationship between full intelligent automation and customer loyalty. Conversely, absorption and attention amplify the indirect effect of absent intelligent automation services on customer loyalty. Particularly noteworthy is the mediating role of interaction, one dimension of customer engagement, in diminishing the impact of full intelligent automation on customer loyalty. This outcome suggests that full intelligent automation services discourages customers from engaging in conversations within the hotel’s community. Since full intelligent automation solely executes programmed tasks, it fails to immerse customers in the hotel’s communal interactions.

In a parallel vein, the impact of full intelligent automation on customers via enthusiasm was also mitigated. This phenomenon suggests that customers experience a decrease in their enthusiasm toward the hotel when they are served primarily by full intelligent automation. Intelligent automation fails to evoke a sense of eagerness and excitement, thereby dampening their loyalty. Put differently, while the standardized service facilitated by intelligent automation does alleviate the workload of frontline employees, it simultaneously erodes customer enthusiasm toward the hotel, consequently diminishing customer loyalty.

Enthusiasm materializes as an emotional connection and proactive engagement on the part of hotel guests. This affective involvement is conspicuously evident in customers’ enthusiasm. The study divulges that the emotional engagement fostered by an intelligent automation solution, specifically enthusiasm, serves as a predictor of customer loyalty. This insight underscores the notion that intelligent automation fails to ignite the excitement of customers about the hotel, resulting in reduced customer loyalty.

Absorption was identified as a negative mediator of the impact of full intelligent automation services on customer loyalty, while it was established as a positive mediator of the effect of employee interactions on customer loyalty. This signifies that when customers engage with full intelligent automation services, their ability to immerse themselves in the experience is limited. Conversely, interactions with employees are more captivating, often causing consumers to become fully engrossed in the conversation and making it challenging for them to disengage. These findings highlight the contrasting nature of service provided by full intelligent automation and human employees. Employees possess the capacity to offer memorable experiences through personalized offerings, heightening the engagement level and subsequently nurturing customer loyalty to the hotel. Initial interactions with hotels should prioritize immersive engagement through personalized services to enhance customer loyalty.

The mediating role of consumer engagement in this study is particularly significant when examined within the context of the Vietnamese service industry. In Vietnam, where service encounters are traditionally characterized by high levels of interpersonal warmth, attentiveness, and relational interaction, consumer engagement becomes a crucial psychological mechanism (Q. Nguyen et al., 2022) that bridges the shift from human-delivered to technology-assisted service. Our findings suggest that partial automation, where human staff remain available alongside automated systems, fosters higher levels of affective, cognitive, and behavioral engagement compared to full automation. This is likely because it preserves the culturally valued human touch while introducing the efficiency and novelty of technology. In high-context, collectivist cultures like Vietnam, customers often expect personalized, relationship-oriented service; when this is compromised, engagement and satisfaction may decline. Thus, consumer engagement not only mediates the impact of automation on customer outcomes but also reflects deeper cultural expectations within service delivery. Future research may explore how different forms of engagement evolve over time as Vietnamese consumers become more familiar with AI-driven service innovations.

The findings also revealed that the impact of full intelligent automation on customer loyalty was diminished by means of identification. Intelligent automation systems typically deliver a standardized service, treating all customers uniformly. Despite the convenience offered by intelligent automation systems, their applications may be restricted and constrained by their programmed nature. This can lead to a lack of flexibility and an inability to create a unique and personalized experience. As a result, customers may not perceive themselves as special and are unable to forge a genuine connection or bond with the hotel. Yoo and Arnold (2019) elucidated that customers are inclined to establish connections with entities they find relatable and wish to be treated distinctively.

Surprisingly, attention did not appear to convey the perception of service quality by full intelligent automation, whereas absent intelligent automation appeared to convey the perception of service quality to customer loyalty. This pattern suggests that merely directing cognitive resources toward fully automated service encounters may heighten awareness without eliciting the affective or evaluative responses needed to strengthen customer–firm relationships. In contrast, when intelligent automation is absent, attention may become a more meaningful cue for interpreting service quality, as customers rely more heavily on human employees and interpersonal interactions to form judgments.

Another noteworthy contribution of this paper lies in its elucidation of the preference for intelligent automation, facilitated by the exploration of the moderating role of EI. Customers’ EI can be harnessed to enhance their service experiences in personal interactions, ultimately fostering heightened loyalty. The findings of this study highlight that customers possessing high emotional perception lean toward absent intelligent automation, while those with elevated emotion assimilation exhibit a disinclination toward intelligent automation systems. Furthermore, customers endowed with high emotional perception show a proclivity for interacting with human employees, leading to an elevation in customer loyalty. This tendency can be attributed to customers’ adeptness in recognizing their emotions and their desire to communicate these feelings to employees. Human employees are often better equipped to handle diverse customer emotions, rendering interactions with them more conducive to crafting memorable service experiences and engendering higher intentions of customer loyalty. These results are aligned with the findings of Prentice and Nguyen (2020), who found that EI has a significant moderation effect on customer engagement.

Conversely, customers displaying high levels of emotion assimilation, characterized by heightened sensitivity to emotions conveyed through others’ behavior, tend to harbor an aversion to full intelligent automation. This sentiment reduction, in turn, impacts their customer loyalty. Such customers are less inclined to embrace services delivered by intelligent automation systems, which lack emotional components. Intelligent automation, being mechanized, does not resonate with customers possessing high emotion assimilation, as it lacks emotional depth.

Conversely, customers characterized by high emotion management exhibit a favorable disposition toward intelligent automation services. This predisposition can be attributed to their heightened emotional self-regulation skills, which allow them to appreciate the favorable attributes of intelligent automation, such as convenience, and acknowledge its limitations in delivering personalized services. These findings support those of N. T. K. Chi and Hoang Vu (2023), who found that consumers can trust in AI and perceive that AI can have empathy, indicating that they can engage with AI.

Moreover, the findings indicate that customers endowed with emotional understanding and management do not perceive a pronounced need for interactions within absent intelligent automation services. This subset of customers is self-sufficient in managing their emotions and therefore doesn’t necessitate assistance from employees to navigate emotional situations. Their adeptness in regulating their emotions enables them to independently manage their emotional experiences, mitigating the need for extensive communication about their emotions.

Implications

Theoretical Implications

This study contributes to theory by showing that the effects of intelligent automation on customer loyalty are contingent and mechanism-specific, rather than uniform. In particular, our findings identify three boundary conditions, service configuration, operative engagement mechanisms, and customer emotional intelligence, that jointly determine when and how intelligent automation shapes customer outcomes.

First, service configuration constitutes a structural boundary condition. The effects of intelligent automation depend on the degree of autonomous service execution. Full intelligent automation (AI-only service) and partial intelligent automation (AI supported by humans) produce systematically different loyalty outcomes, indicating that automation effects are not monotonic along a simple automation continuum. Instead, the availability of human support fundamentally conditions how customers experience and evaluate automated service encounters (Castillo et al., 2021).

Second, customer engagement operates as a selective mediating mechanism. Intelligent automation influences customer loyalty only through specific dimensions of engagement (interaction, absorption, enthusiasm, and identification) while other dimensions exhibit limited or null effects. This pattern reveals a mechanism boundary condition: automation activates relational and experiential engagement pathways but does not uniformly affect all facets of engagement. By unpacking engagement into its constituent dimensions, the study clarifies which psychological processes transmit automation effects to loyalty and which do not. It highlights the necessity to consider multiple dimensions of customer engagement (So et al., 2016) as intermediaries that can either impede or augment the link between intelligent automation and customer loyalty.

Third, emotional intelligence represents an individual-level boundary condition. Emotional intelligence does not function as a general predictor of customer loyalty; rather, specific emotional capabilities shape customers’ tolerance for automated service encounters. Emotional perception and emotion assimilation amplify negative reactions to full automation, whereas emotion management attenuates these effects, and emotional understanding shows limited moderating influence. These asymmetric moderation patterns delineate when emotional intelligence intensifies, buffers, or does not affect customer responses to intelligent automation. Taken together, these findings advance theory beyond main-effect models toward a contingent, mechanism-based perspective on intelligent automation, emphasizing that its impact depends on service design, operative engagement pathways, and customers’ emotional capabilities.

In summary, the effects of intelligent automation are bounded by (1) the degree of automation implemented, (2) the specific engagement mechanisms activated, and (3) customers’ emotional intelligence profiles, clarifying when intelligent automation enhances or undermines customer loyalty.

Practical Implications

The findings of this study hold substantial managerial implications for organizations to consider the application of intelligent automation within their service delivery processes, particularly in the realm of hospitality. First, managers should be cautious about adopting fully automated service configurations when customer loyalty is a key objective. Our results show that full intelligent automation reduces customer loyalty relative to human-only service, whereas partial automation with human support performs more favorably than full automation. This suggests that hybrid service designs are more effective than fully automated ones for sustaining customer relationships.

Second, managers should focus on the specific engagement mechanisms through which automation influences loyalty. The mediation results indicate that intelligent automation affects loyalty through selected dimensions of customer engagement (namely interaction, absorption, enthusiasm, and identification), while other dimensions show limited or null effects. Practically, this implies that automated services should be designed to promote meaningful interaction and customer identification, rather than assuming that automation will uniformly enhance engagement.

Third, managers should account for customer heterogeneity in emotional capabilities when deploying automation. The moderation analyses reveal that emotional intelligence conditions customer responses to automated service encounters: customers high in emotional perception and assimilation react more negatively to full automation, whereas those high in emotion management are less adversely affected. This suggests that automation strategies should be tailored to customer segments, with greater human involvement for emotionally sensitive customers. Overall, the findings indicate that the effectiveness of intelligent automation depends on aligning service configuration, engagement design, and customer emotional characteristics, rather than maximizing automation intensity.

Limitation and Directions for Future Research

It’s important to note that the preference for partial intelligent automation may not be universally applicable to all customer segments or industries. Some customers might prioritize the convenience and swiftness of fully automated services, particularly for routine and standardized tasks. Additionally, the effectiveness of partial intelligent automation in fostering customer loyalty could hinge on factors such as the caliber of human assistance, the seamless integration between automation and human agents, and the ease of transitioning between automated and human-supported interactions. Another limitation of this paper is that it does not capture alternative mediators, such as perceived efficiency and trust (O. H. Chi et al., 2023; Langer et al., 2023), or potential moderators, which may also influence the relationship between intelligent automation and customer loyalty. Furthermore, although this paper focuses specifically on the hotel check-in context, it remains uncertain whether these findings generalize to other service settings. Future research should study the effectiveness of partial or full intelligent automation in different service settings, like service failures and personal recommendations. Especially, under what conditions can intelligent automation truly enhance customer loyalty during failure recovery?

While this study employed a quantitative experimental design to establish causal relationships between levels of automation and customer responses, future research could benefit from adopting a mixed-method approach. Combining experiments with qualitative methods such as interviews, focus groups, or ethnographic observations would offer richer insights into the underlying motivations, cultural interpretations, and contextual factors shaping customer and managerial perceptions of intelligent automation. Such an approach could also enhance the practical relevance of the findings for hotel practitioners by uncovering implementation challenges and guest expectations in real-world settings.

The call for further research is warranted to examine the specific mechanisms and underlying factors that steer the observed relationship between partial intelligent automation and customer loyalty. By investing customer perceptions, expectations, and preferences vis-à-vis intelligent automation, organizations can fine-tune their strategies and engineer service encounters that not only amplify customer loyalty but also pave the way for enduring success. Although the paper specifically examines the check-in context, it remains uncertain whether these findings can be generalized across other service environments. Future research should explore how the benefits of partial or full intelligent automation in handling service failures vary by industry and service type, and under what conditions these automated solutions truly enhance customer loyalty during failure recovery. Additionally, further investigation into alternative mediators and moderators is warranted to provide a more comprehensive understanding of the factors at play.