Employee–robot relationship typology: the role of robot agency, communion and competence Open Access

Human–robot teams are revolutionising organizational workflows and service delivery systems. In recent years, the adoption of service robots in tourism and hospitality has transformed how organizations operate and deliver services (Liu et al., 2025b). This trend is expected to accelerate as the industry increasingly relies on AI technologies to enhance efficiency (Bakir et al., 2025). However, the challenge lies in integrating both into a cohesive and harmonious workforce (You and Robert, 2022).

Employee–robot collaboration is crucial for enhancing customer experience and service delivery (Mele et al., 2024). The integration of robots facilitates service providers delivering a more personalized and extraordinary experience for customers while lowering businesses’ labour costs (Ma and Ye, 2022) and has the potential to substantially reshape job designs and workflows, influencing the dynamics of employee interactions with both customers and robots (Horpynich et al., 2024). Nevertheless, the adoption of service robots has often been implemented without full consideration of their potential disruption of well-established working relationships (Ma and Ye, 2022).

Much of the existing research on human–robot interaction focuses on customers’ perspectives or use of robots (Shum et al., 2024). Early psychological impacts research shows job security threats in low-skilled workers from robot implementation (Chao and Kozlowski, 1986) and undesired outcomes such as withdrawal behaviours at work (Pan et al., 2024). Despite researchers highlighting the importance of relationship norms and robot traits in human–robot interaction (Liu and Wang, 2025), limited studies focus on how robots’ social presence influences workplace relations.

This lack of attention to the employee–robot relationships is significant. Companies may unintentionally create undesired workplace dynamics among employees. Because robots’ presence in the workplace may benefit employees’ overall service performance and psychological well-being (Sehgal et al., 2025), exploring the employee–robot relationship and its triggers in this emerging interaction context is both timely and necessary. This study explores the employee–robot relationship and the foundational dimensions of the relationship dynamics in an attempt to answer two research questions:

Despite the need to develop social skills for robots (Dautenhahn, 2007), previous studies have investigated the human-collaborative robots (Cobots) interaction in manufacturing, with a focus on improving safety, proximity and motion (Lasota and Shah, 2015) to enhance task execution (Makrini et al., 2019). Less attention has been paid to its relational and social dynamics in service contexts; hence, there is a need for studies on robot-specific criteria, such as capabilities and relationship quality (Fischer, 2012), and with a focus on the interaction between employees and robots to prepare organizations for potential challenges (Erebak and Turgut, 2018).

This gap highlights the need to integrate perspectives from organizational psychology and human–object assemblage to understand how employees interpret robots in the workplace. This paper addresses this gap by conceptualizing and introducing a typology of employee–robot relationships, using social cognition theory and confirming their triggers through assemblage theory. Drawing on qualitative interviews and a scenario-based experiment, this study explains relationship types as the result of how people perceive their coworkers and develop judgments and stereotypes, and consistent with Novak and Hoffman (2018), it integrates the circumplex model and assemblage theory as a practical way to represent human–object relationships in dynamic assemblages. Both studies’ procedures were reviewed and approved by the university’s Ethics Review Board.

Coworkers affect work environments and, consequently, influence employees’ attitudes at work (Avci, 2017). A harmonious relationship between employees encourages positive job outcomes. Conversely, when the workplace is marked by conflict among employees, it undermines performance (Abugre and Acquaah, 2022). Furthermore, the level of exchange quality between coworkers defines their relationship quality (Love and Dustin, 2013). An optimal work environment entails both physical and emotional safety (Anitha, 2014). Therefore, coworker support is an essential part of the relationship. This encompasses both emotional support, in which coworkers listen to and empathize with one another, and instrumental support, which involves assisting with task completion (Singh et al., 2019).

The existence of robots disrupts not only concepts but also norms of human relations (Friedman, 2025). In human–robot partnerships, people treat robots as machine-human hybrids who are alive enough to be anthropomorphized but also machine enough to be treated as not-quite human (Bankins and Formosa, 2019). Hence, human–robot relations could be approached through a combination of a human–human and a human–object lens.

Social cognition theory incorporates cognitive processes through which individuals interpret emotions, behaviors and thoughts in social situations (Carpenter et al., 2024), enabling employees to make sense of their social surroundings (Lynch and Rodell, 2018). Although the theory originated in human–human interaction, it applies to human–robot contexts, as robots are placed somewhere between social partners and tools (Henschel et al., 2020). Additionally, the social cognition theory posits that people form impressions based on what they know and observe about others (Goffman, 1959) and mentally sort information into schemas that shape their judgments of others, and by doing so, they can anticipate others’ traits and behaviors through inferences, attributions and stereotype formation (Karl et al., 2023).

Coworker relationships are an important social context factor (Verbos et al., 2014). In personal relationships, people interpret others’ actions by attributing them to the situation, the person’s own traits or a mix of both (Blanchard-Fields and Cooper, 2003). In this view, employees’ relationships with robots can be investigated through employees’ social cognition and perceptions. Furthermore, humans’ perceptions of robots are influenced not only by robots’ characteristics but also by individuals’ cultural backgrounds, including their beliefs, values and social norms (Kou and Zhang, 2024), such as higher acceptance of robots in eastern versus western cultures (Li et al., 2022). Nevertheless, this study focuses on perceptions of robots’ traits and situational attributions.

Human–robot research has revealed various forms of relationships, and it has been established that different relationship styles can coexist. For example, coopetition (simultaneous collaboration and competition) can occur, as can complementarity and collaboration (Tan et al., 2024).

Collaborative relationships entail interdependence between employees and robots toward shared goals, joint workflows and joint decision-making authority (Le et al., 2022). Employees with relatively higher trust in robots actively and comfortably engage in collaboration (He et al., 2023). Additionally, because humans and robots have capabilities and limitations, researchers suggest that humans and robots should collaborate to maximize the unique strengths of each (Wang et al., 2024). On the other hand, competition can occur because of specific dynamics such as social comparison to robots’ performance (Jin et al., 2025), job insecurity when employees compete to secure their job (Khaliq et al., 2021), higher robot intelligence (Jin et al., 2025) and threats to employees’ status or role at work (Yang et al., 2024).

Supplementary fit exists when two parties share at least one common characteristic, such as values, culture, goals or personality (Ryu, 2014). It particularly occurs when organizations hire employees with skills that are replicable and already exist in their workforce (Cable and Edwards, 2004). Alternatively, when employees view robots as tools, removing them does not affect service. The connection between equipment functionality and hardship reduction suggests that employees develop a positive attitude toward using the equipment to ease physically demanding tasks (Na et al., 2022). In this sense, robots supplement employees in various contexts, such as automated baristas and restaurant delivery bots (Wirtz et al., 2023).

Complementary fit refers to situations when the weakness or needs of one party are offset by the other party’s strength (Cable and Edwards, 2004). The dissimilarity adds value by combining the other person’s value (Mo et al., 2024). Moreover, according to the complementarity theory, employees prefer working with others who complement them (Zhang et al., 2023b). Complementary strengths between humans and robots highlight their respective capabilities (Tan et al., 2024); therefore, it is anticipated that robots will significantly complement the service workforce in the coming years, and human–robot teams will serve customers in most service encounters (Xiao and Kumar, 2019).

On this basis, the following proposition is made:

While social cognition is used to explain employee–robot relationships at a micro-psychological level, the assemblage theory is used in this study from a macro-relational human–object perspective. In other words, this study conceptualizes employee–robot relationships not as static trait-like evaluations but as fluid emergent configurations produced by human perceptions, robot features and behaviours and their interaction.

Assemblage theory posits two types of relations: Relations of interiority are identity-defining, such that actors derive their meaning from the relationship itself (e.g. father–daughter relationships). By contrast, relations of exteriority connect heterogeneous actors without redefining their identities (DeLanda, 2016), as in employee–robot interactions, where both actors remain distinct while temporarily aligned through co-functioning.

Assemblage theory underscores the role of objects as actors, using their capacity to act agentically, in which objects can express themselves without hindrance within the assemblage (Huff et al., 2020). These assemblages are diverse compositions or collections of entities, defined by their emergent properties and capacities and develop from ongoing, complex and interconnected interactions among their parts, which consist of humans, materials and meanings (Huff et al., 2020).

The assemblage theory is suitable to explain human–object relations, wherein both consumers and objects have some experience and ability to express agentic and/or communal roles in their interactions as part of an assemblage (Novak and Hoffman, 2018). The theory proposes an ontology in which relational meanings, uses and encounters emerge dynamically in process from the totality of multiple interacting components (Canniford and Shankar, 2012). On this basis, the following proposition is made:

Given the research questions’ exploratory nature, the first study uses a qualitative approach, using semi-structured interviews to investigate employees’ perceptions of working with robots, the types of employee–robot relationships and the factors that influence these relationships. The findings are used to develop a framework, which is tested quantitatively through a scenario-based experiment in Study 2.

Employees across different restaurant functional roles have been recruited to maintain heterogeneity with a purposeful sampling. To receive in-depth knowledge, a 3-month working experience with robots was required. A total of 12 interviews were conducted, and interviewing ended when responses became similar to those in prior interviews, suggesting theoretical saturation without yielding any new insights. The interviews lasted, on average, 45 min. Table 1 summarizes the interviewees’ demographic profile.

The researcher contacted restaurants directly to recruit participants and conducted 10 online and two in-person interviews in English. Participants were originally from China and living in Australia at the time of the interviews. Interviewees received an information sheet and consent to participate, after which a transcript was sent for any necessary changes. A thank-you gift was offered to participants as a gesture of gratitude for their time. The interview questions were based on theories related to the proposed relationships, including interdependence, resource scarcity, resource conservation, expectancy and social exchange. Question samples were:

The interviews were transcribed and analyzed using NVivo and Microsoft Excel, using a framework analysis. This technique is suitable for providing practical findings from managing qualitative data through a dynamic approach (Parkinson et al., 2016). The method follows a structuring and interpretative approach, facilitating researchers to relate data to existing theories or phenomena (Boström et al., 2025). The analysis involved five steps:

1. data familiarization;

2. identifying a thematic framework;

3. indexing;

4. charting; and

5. mapping and interpretation (Goldsmith, 2021).

To mitigate researcher bias, two researchers/coders independently produced initial codes by listening to the interview recordings and reading transcripts. Afterward, the coders collectively reviewed the initial codes and categorized them to create our framework (Lee et al., 2022).

The four proposed relationship types were found. In collaboration, employees expressed trust in relying on robots, perceiving them as helpful most of the time. They heavily rely on robots, which are not yet human replacements, necessitating their mutual reliance. For instance, robots are limited to processing basic tasks, whereas employees handle complex work situations. Furthermore, they expressed their interconnected goals, evident in robots delivering food while they work on the following order, thereby sharing a common goal of serving the customer.

Competition highlighted social comparison: employees compared their physical abilities – such as getting tired – to the robot’s, which can work longer hours. They also considered whether a robot, being more efficient, would outperform them at work. Employees expressed concerns that the robot’s intelligence may threaten their job security, potentially eliminating the need for human workers. There were expectations and anxiety that robots would take over human jobs in the future. Furthermore, they expressed that if robots can make decisions, they will be direct competition to employees.

In supplementary relationships, employees view robots as restaurant staff, currently limited in capabilities. However, they will be able to fully support them once they are upgraded with capabilities similar to those of employees. Additionally, robots were seen as a supporting tool that could assist employees by taking over basic, repetitive tasks.

In complementary relationships, robots are viewed as tireless agents capable of executing multiple orders and traveling long distances and offer a means to circumvent human interactions and emotional labour. Employees experienced added value from robots in saving their energy and enabling them to multitask, additionally in saving costs, and in providing a preferable option for introvert employees who prefer interacting with robots rather than humans, or when employees need to avoid facing emotional exhaustion in dealing with negative customer behaviour through mediation of their robot co-workers in these interactions.

Two dimensions were repeated across participants’ answers, leading to the expression of different emotions or conflicts: the robots’ agency and communion.

A failure in the robot’s decision-making and judgment, and its mismatch with employees’ expectations, frustrated some employees or led them to feel that robots lowered efficiency. Employees reported instances in which robots slowed or stopped without reason, got stuck performing tasks, moved slowly or took action on their own. Robots were seen as inflexible in communication, with only pre-set functions. Furthermore, employees preferred robots to be more assertive in notifying before conflicting with others.

Employees needed robots to learn and adapt to evolving service situations and to engage in interactive, two-way communication. Employees preferred robots to give and receive feedback. On the other hand, empathy and emotional connection were important to some, while indifferent to others; however, employees expressed the perception that they see them as more than mechanical machines when robots express empathy. Table 2 presents the interview codes identified in the analysis.

The second study aimed to verify the first study’s findings quantitatively, thereby confirming the existence of the four relationship types and testing the dimensions as influencers of these relationships. Hence, a scenario-based experiment was conducted.

In the human–object relationship, object agency refers to the robot’s ability to affect the assemblage (Novak and Hoffman, 2018). It does not imply that objects act with human-like intentions but rather that their capacities emerge through interaction with other entities (Schneider-Kamp et al., 2024). Moreover, it is relational, wherein objects do not have intrinsic agency, but it is gained by interacting with other entities such as humans, spaces or other objects; additionally, it can be constrained or empowered (Epp and Price, 2009). On the other hand, in the context of social cognition, agency is the ability to influence one’s life through beliefs about one’s own self-efficacy and self-regulation (Bandura, 1989).

Objects, such as robots, are not passive. However, they actively form the interactions within a sociotechnical system, contributing to a collective action that is not an independent actor but rather as mediators, enablers or constraints (Gilbert and Laporte, 2021). Based on the interview findings and the above discussion, the following hypotheses are proposed:

Competence refers to employees’ perceived ability to perform an expected service action and is represented by their intelligence, skill and efficacy (Calleja et al., 2023). It is one of two dimensions that influence human social cognition (Fiske et al., 2006) and is considered a context-based construct in which robots are seen as core competent when their abilities match the nature of the task (Hoang and Tran, 2022).

Competence is one of the main predictors that enhances trust development in interactions (Christoforakos et al., 2021). A high level of robot service competence reduces employees’ perceived risk of cooperation. In contrast, lower competence evokes resistance to collaboration (Kim, 2023). Moreover, from a social-cognition perspective, there is a connection between how humans perceive robots’ agency and their perceived competence and warmth (Calleja et al., 2023). Based on interview findings and the above discussion, the following hypotheses are proposed:

Communion is a dual-facet construct; at a macro level, in a human–object context, through assemblage theory, it refers to the robot’s ability to be affected by the assemblage (Novak and Hoffman, 2018), and at the micro level, in social science, specifically social cognition, communion refers to social warmth and prosociality, such as kindness, empathy and social connectedness (Bettinsoli and Formanowicz, 2022); in other words, it is the perception of an object’s warmth and affection that contributes to social relatedness and connection (Huprich et al., 2015).

Social perception in humans relies on two dimensions: perceived warmth and perceived competence (Fiske et al., 2006). Studies in human–robot interaction show that human traits affect their expectations of robots. For example, congruity theory suggests that customers with a communal relationship orientation expect their robot-social partner to comprehend their needs and be responsive and interactive (Chang and Kim, 2021). Furthermore, studies have shown that warmth moderates perception dynamics (Yang et al., 2024).

Based on the interview findings and the above discussion, the following hypotheses are proposed:

Drawing on the social cognitive theory, self-efficacy – the belief in one’s own abilities – is a central mechanism of a person’s agency and can be enhanced through mastery experiences (Bandura, 1989), and according to the assemblage theory, the ongoing agentic interaction between objects and humans shapes relational dimensions, whereas similar values in agency and communion lead to partnership (Hoffman and Novak, 2017).

Studies on human–collaborative robot interactions indicate that employee self-efficacy, operationalized as robot-use self-efficacy, reduces perceived robot threat from anthropomorphism, which in turn reduces negative attitudes toward robots (Liao et al., 2024). Furthermore, studies show that robot agency increases self-identity threat (Yu et al., 2025); therefore, the following hypothesis is proposed:

We examined participants’ reactions to service robots in a restaurant context. The experimental design was a 2 (agency: high assertive-dominant vs low unassured-submissive) x 2 (communion: high warm-agreeable vs low coldhearted) between-subjects design; a total of 347 valid responses were collected through Prolific from participants residing in Australia, the USA, the UK and Spain. On average, the survey took 11 min, and participants were rewarded at a rate of £6.16 per hour. Respondents’ demographic information is summarized in Table 3.

Following Chang and Kim (2021), we first developed four scenarios. The qualitative interviews informed scenario descriptions. In each scenario, robot agency and communion were manipulated. Participants were randomly assigned to one of the four scenarios and asked to read it as if they were the employees in that scenario. Consequently, participants were asked to answer agency and communion manipulation-check questions and to report their perceptions of robot competence and four relationship types.

We used a seven-point Likert scale and used items from the revised interpersonal adjective scales (Wiggins et al., 1988), which align with Novak and Hoffman (2018), to capture agency and communion dynamics in relationships. The questions were about participants’ perceptions of the robot “ALEXBOT.” Two items measured assertive-dominant, three measured unassured-submissive, five items measured warm-agreeable and cold-hearted was measured by three items.

All the other variables were measured on seven-point Likert scales as well. The items were adopted from previous studies. Collaboration was measured using five items from Zhang et al. (2023a). The competition scale, adopted from Gordils et al. (2023), consisted of three items. Supplementary and complementary relationships were adopted and measured by Piasentin and Chapman (2006), with five and four items, respectively. Perceived robot competence was measured on a four-item scale adopted from Hu et al. (2020). The robot-use self-efficacy scale, adapted from Liao et al. (2024), consisted of four items. The control variables in this study were the demographic variables of employees (gender, age and education level), similar to Liu et al. (2025a). The scenarios used in the experiment and the full list of measurement items are provided in the Supplementary Material.

Manipulation checks confirmed the effectiveness of both experimental manipulations through one-way ANOVA. Participants who read the assured–dominant scenario reported significantly higher perceived robot agency than those in the unassured–submissive condition [M = 5.94 vs M = 2.76, F(1, 345) = 634.32, p < 0.001]. Similarly, participants exposed to the warm–agreeable scenario reported significantly higher perceived robot communion than those in the cold-hearted condition [M = 4.70 vs M = 3.63, F(1, 353) = 49.11, p < 0.001].

We used linear regression analyses to test H1 and H2. To examine mediation, moderated mediation and moderation, we used Hayes’ SPSS PROCESS macro with 5,000 bootstrap samples and 95% bias-corrected confidence intervals, applying Models 4, 14 and 5, respectively. Gender, age and education were included as demographic control variables in all analyses (Lian et al., 2025). The substantive pattern of results remained unchanged after including control variables. Table 4 presents the results for the proposed relationships, while Table 5 reports the mediation analysis results.

The results reveal that robot agency has a positive and significant effect on the four relationship types and on perceived robot competence, thus supporting H1a–H1d and H2. In other words, assertive robots lead to stronger perceptions of competence and more pronounced employee–robot relationships.

The results indicate a significant positive indirect effect of robot agency on collaborative and supplementary relationships via perceived robot competence. The direct effect of robot agency on them was not significant, indicating full mediation, thus supporting H3a and H3c.

Additionally, the indirect effect of robot agency on the complementary relationship was significant and positive, and the direct effect was significant and negative, indicating partial mediation. These results support H3d and suggest that while higher robot agency directly suppresses complementary relationships, it indirectly enhances complementarity by increasing employees’ perceptions of robot competence.

Although robot agency had a significant positive direct effect on competition, the indirect effect was not significant. Because the confidence interval includes zero, perceived competence does not mediate the relationship between agency and competition. Therefore, H3b is not supported. This indicates that competitive relationships are driven primarily by robot agency rather than by competence-based appraisals.

Overall, these findings demonstrate that assertive robot behavior functions as a competence cue that facilitates collaborative, supplementary and complementary relationships. In contrast, competition is triggered directly by agency cues independent of perceived competence.

The moderated mediation index was statistically significant for collaborative (Index = 0.334, 95% BootCI [0.009, 0.649]) and supplementary relationships (Index = 0.287, 95% BootCI [0.048, 0.555]), indicating that the indirect effect of robot agency on these relationship types via perceived robot competence varied as a function of robot communion. In contrast, the indices of moderated mediation for competitive and complementary relationships were not statistically significant.

To further probe the moderation effects, a simple slope analysis was conducted. The positive effect of perceived robot competence on collaborative and supplementary relationships was stronger under high robot communion.

Self-efficacy significantly moderated the relationship between robot agency and competitive relationships only. Specifically, the interaction between robot agency and self-efficacy was negative and significant, indicating that higher levels of employee self-efficacy weaken the positive association between robot agency and perceived competition. Simple slope analyses further show that the effect of robot agency on competition is strongest at low self-efficacy and becomes weaker and non-significant at high self-efficacy levels.

Despite no significance on other relationships, it has a positive direction on collaborative and supplementary relationships. Consistent with social cognitive theory, self-efficacy primarily operates as a protective psychological resource that mitigates threat-based competition.

This study examined how frontline service-industry employees form different relationship types with service robots, depending on robot agency, communion and competence. The findings suggest that competence is the primary driver across positive relationship types, while agency and communion play more conditional roles.

The study’s results indicate that the agency exerted a positive influence across both constructive and adverse relationship types. Despite the agency positively shaping perceptions of competence (e.g. service robots were seen as more capable) and collaborative, supplementary and complementary relationships, its direct effect on the competitive relationship was positive and significant. Robots that were assured dominant were often perceived as more competitive. In the service industry, employees value teamwork and task coordination, which makes highly dominant robots be perceived as rivals rather than allies. This suggests that embedding agency cues in robots must be carefully designed to ensure that robots signal initiative without compromising human autonomy.

Across the three positive relationship types – collaborative, supplementary and complementary – perceived robot competence consistently functioned as a proxy for agency in shaping these relationships. Results indicated that the higher the competence, the stronger the relationships with the robot when it was perceived as efficient, capable and skilful. For service workers operating under high pressure, competence serves as an important element. This reinforces prior research that emphasizes technical performance as a requirement for service robots’ social acceptance and as a fundamental dimension of social cognition, as in Liao et al. (2023).

Contrary to expectations, robots’ communion (warmth, agreeableness) had weak or adverse direct effects on relationship outcomes. Employees did not value warmth unless it was paired with competence. When robots greeted employees warmly and offered encouraging phrases, these gestures alone were insufficient if the robot failed to perform efficiently. This implies that warmth without competence may even trigger cognitive dissonance as the robot’s friendliness conflicts with its poor performance.

Although communion was weak on its own, its combination with competence was consistently significant across two positive relationship types. Competent robots that displayed warmth evoked stronger perceptions of collaborative and supplementary evaluations. This finding highlights the conditional effect of communion: it enhances relationship strength and quality only when paired with competence.

While employee robot-use self-efficacy was weak across positive relationships, its role in reducing competitive relationships was significant, aligning with assemblage theory, which posits the exchange of agentic roles between humans and objects in shaping relational dynamics. This finding aligns with past research that highlighted its role in moderating perceived threat from robot intelligence on negative attitudes toward cobots (Liao et al., 2024).

The combination of robots’ agency, communion and competence revealed how employees categorize their relationships with service robots.

Collaborative: when robots were competent and warm, combining high agency that manifested through elevated perceived competence and high communion. Employees viewed it as an opportunity to work collaboratively and make joint decisions. Similarly, in Study 1, collaboration was based on shared goals and interdependence.

Competitive relationships reflected a direct expression of agency rather than an agency–competence transformation. In this context, robots were perceived as highly agentic but low in communion, positioning their agency as rivalrous rather than supportive. Employees with lower robot-use self-efficacy were more likely to construe such agency as threatening, as highly autonomous robots may seem to undermine employees’ autonomy or compete with them for task control. In contrast, higher levels of robot-use self-efficacy attenuated competitive perceptions. This pattern is consistent with the qualitative findings, where competition emerged through social comparison with robots and perceptions of threat stemming from robots’ intelligence and autonomous capabilities.

Supplementary: when robots were high in communion and competent. Service industry employees work not only on tasks but also on aspects of warmth, such as greeting guests. When robots can replicate this, they become more capable, similar to human employees, and thus become supplementary. In other words, and as seen in interviews, supplementary comes from similarities either in task – treating robots as a tool – or in personality.

Complementary: centered on competence, which manifested in robot agency. Offering assistance in performing tasks efficiently without replicating or competing with humans’ autonomy and emotional capability in interacting with guests, for example. Consistent with Study 1 findings, a robot can add value by offering different characteristics or abilities to humans. Figures 2 and 3 illustrate employee–robot relationship types across levels of agency and communion (Figure 2) and competence and communion (Figure 3), whereas Figure 4 presents the relative effects of these factors.

The proposed relationship typology adds conceptual novelty to the field of organizational behavior, compared with previous frameworks such as coworker relations, human–robot teaming, coopetition and person–environment fit. While previous models emphasize either human–human relational qualities, task coordination initiatives, competitive and collaborative tensions, or compatibility between individuals and their environments, our typology centres on the structural configuration of agency and communion in employee–robot interaction and conceptualizes this relational structure as an integrative framework that bridges relationship models of human–human through social cognition and human–object through assemblage.

The findings of this study indicate that robots are perceived as social actors who affect and are affected within the workplace ecosystem, unlike the technology acceptance model, which focuses on utilitarian aspects (e.g. perceived usefulness and ease of use) (Lin et al., 2025), and therefore expanding employee–robot literature from a relational perspective.

This study extends human–robot interaction and social cognition theory from individual impression formation to relational schema construction. It is one of the early attempts to investigate the competence and communion dimensions in the human–robot relationship context, showing that employees use human-like cognitive and affective processes to evaluate and relate to service robots. Additionally, the findings explain how employees form social relationships with nonhuman agents. Furthermore, while warmth in human interaction is considered a primary dimension and is judged before competence (Fiske et al., 2006), our finding of communion does not have a strong impact on its own and suggests that people treat robots differently than they do humans, where competence is seen as a primary dimension for positive relationships.

Despite evolving research on individual aspects such as autonomy, competence and empathy in human–robot interaction, it remains unclear which of these aspects is more essential and how their interactions affect employees. Therefore, the findings of this study broaden the scope of the human–robot interaction and dynamics literature by revealing the employee–robot relationship dynamics.

Experience shows that robots currently face frequent failures operating in unstructured environments (Honig and Oron-Gilad, 2018). The findings of this study inform robotics developers on the necessity of robots to be competent in task execution with minimal chance of failures and the importance of embedding robot agency cues that signal initiative without compromising human autonomy; moreover, the necessity of raising robot communion traits, such as warmth or empathy cues, with the rise in their agency and competence.

Managers can ensure successful task allocation between employees and robots to enhance competence and consider the degree of robot autonomy and warmth in job design. Additionally, managers can facilitate employee–robot relationships (e.g. a kitchen may benefit from a complementary relationship, while the front office may benefit from a supplementary or collaborative relationship). Furthermore, managers can create training strategies and upskill employees to mitigate competition and enhance collaboration.

As robots’ presence affects workplace relations and organizational behavior, this study is important for human resources managers because it enhances their knowledge of employee–robot dynamics. Therefore, human resources managers can develop employee–robot relationship assessment tools to monitor and enhance workplace relations, such as surveys to measure relationship type and quality and robots’ competence in the workplace.

Knowledge about robot–coworker relationship dynamics equips employees to manage their interactions, minimize relational ambiguity and work more productively. By enhancing self-efficacy through training and experiences, employees can avoid competitive relationships with robots and use collaboration.

Finally, while tourism and hospitality organizations adopt robots to enhance operational processes and optimize costs, the study findings inform regulators’ and policymakers’ consideration of human–robot interaction ethical design guidelines and the incorporation of behavioral design and relational dynamics, especially regarding robots’ agency and communion, to foster collaboration and avoid competition with employees.

Despite the rigorous nature of this study, several limitations exist. The study collected data across the hospitality and service industries; therefore, the findings may not be generalized to other non-service industry organizations. Additionally, this study was built on the robot’s inherent traits and on the ongoing assemblage between employees and robots, i.e. they both impact each other; future research is encouraged to incorporate human employees’ agency, communion and competence into its approach.

While interviews collected data from Chinese employees living in Australia, the quantitative study participants were from Western countries; therefore, generalizability to Eastern and other cultures may be limited. Cross-cultural research across diverse regions is encouraged. Furthermore, our survey measured employees’ perceptions; longitudinal studies are encouraged. Future studies can investigate the impact of employee–robot relationships on employees’ behaviors, attitudes and organizational outcomes.

Although nearly half of the respondents were from the retail industry, a short-term robustness test comparing descriptive results for the hospitality/tourism subsample (n = 123) showed mean construct scores highly consistent with the full sample, indicating no concerns regarding applicability to hospitality and tourism contexts.

With the ongoing debate over power and authenticity dynamics, future research can investigate effective ways to integrate agency while maintaining workplace relations and power dynamics and communion while preserving robots’ authenticity perceptions. The research can further explore the human–robot relationships and develop frameworks to enhance workplace relations and organizational behavior, such as developing models for employee engagement and satisfaction.

The supplementary material for this article can be found online.