From automation to employee loyalty: understanding the balance between robots and workforce stability

The hospitality industry is facing unprecedented challenges. China’s hospitality sector is experiencing considerable workforce instability, with post-COVID turnover at 58% (Xu et al., 2023). This issue is further complicated by the swift integration of automation and inconsistent labor policies. Amid this turmoil, advancements in artificial intelligence are driving the integration of robots into service delivery, promising operational efficiency and innovative guest experiences (Park et al., 2023). By 2023, service robots are expected to comprise nearly 25% of the global hospitality workforce, and AI and automation are anticipated to replace one-third of jobs by 2025 (Xu et al., 2023, 2024). Managers operationalize robot usage as the extent to which hotels deploy automated systems (e.g. AI-powered concierge robots, automated room-cleaning devices) to replace or augment human tasks. For instance, Japan’s Henn-na Hotel uses robots for 40% of its front-desk operations, whereas US hotels lag behind at 15% due to labor union resistance (Reis et al., 2020; Willcocks, 2020). These differences may suggest that labor market conditions, technology adoption and technological infrastructure influence automation trends in different regions/countries.

Robots are increasingly deployed in customer reception, room cleaning and food and beverage service to optimize processes and reduce costs (Koo et al., 2023). However, the ability of robots to perform tasks independently may threaten people’s sense of job security, but the extent of this impact will vary depending on an individual’s job role. Haldorai et al. (2024) suggested that frontline employees experience higher levels of job insecurity due to direct exposure to automation. Conversely, back-office employees tend to perceive robots as effective in improving their productivity.

Effective robot implementation typically involves collaboration with human workers, requiring a nuanced understanding of employee perceptions and emotions (Li et al., 2024). Hotel managers must address these insecurities proactively to prevent increased turnover. Ideally, the introduction of robots should create a symbiotic cycle where organizations invest in employee skill development, thereby enhancing workforce efficiency, expertise and job satisfaction.

Some studies suggest that robot integration can increase employee turnover due to job displacement (Fulmore et al., 2023; Zhang et al., 2023; Brougham and Haar, 2020), while others indicate that this impact can be mitigated through organizational support (Başer et al., 2025). Conservation of resources (COR) theory suggests that employees evaluate the loss of job security against the support they receive. However, little research has examined this in high-automation settings, such as China. While some findings suggest that robots may increase turnover intentions due to concerns about job security (Zhang et al., 2023), others indicate that perceived organizational support and employee development can foster positive attitudes toward human–robot collaboration (Başer et al., 2025). Ekuma (2024) confirmed that enhanced training and development opportunities can mitigate the negative effects of automation integration, and that training programs that focus on human–machine collaboration significantly increase employee job security. Stacho et al. (2024) implemented a human–robot collaboration curriculum (e.g. troubleshooting robot errors, optimizing task allocation between staff and robots), which reduced turnover by 20%. This highlights the importance of role-specific training, rather than generic upskilling. Therefore, understanding the reciprocal relationship between robot usage and employee outcomes is crucial for balancing technological benefits with workforce sustainability.

This study examines the reciprocal relationship using the COR theory, investigating how the introduction of robots affects job insecurity, perceived organizational support, job satisfaction and turnover intention in the hospitality sector. COR theory suggests that individuals strive to retain and build resources, such as job security and support, to manage work demands. The cycle of reciprocity between automation integration and the employee depends not only on the employee’s own ability to adapt, but also on management strategies that optimize human–robot collaboration (Khoa et al., 2023; Leocádio et al., 2024). Hotel managers play a crucial role in maintaining this balance by assigning tasks effectively, redesigning roles, providing structured training and making ongoing operational adjustments. For example, robots perform repetitive tasks, while employees focus on providing personalized and emotionally valuable interactions to guests (Khoa et al., 2023), aiming to improve service quality and work engagement. In addition, the effectiveness of structured training for various roles, including retraining programs, cross-training and technology adaptation workshops, can be beneficial in reducing employee turnover (Pinnington et al., 2024).

This study aims to fill the research gap on the interplay of robot use on employees’ job outcomes and turnover intentions by exploring the impact of robot use on job insecurity, examining the link between job insecurity and turnover intentions, assessing the mediating role of job satisfaction and perceived organizational support and examining the moderating role of skill development opportunities. This research used partial least squares structural equation modeling (PLS-SEM) and fuzzy set qualitative comparative analysis (fsQCA) to comprehensively address these objectives. PLS-SEM examines complex theoretical models, while fsQCA analyzes multiple causal pathways, providing nuanced insights into factors influencing turnover intention. By integrating these methods, the study provides a robust understanding of how robot use affects employee retention, offering hotel management practical guidance to enhance workforce sustainability amid rapid technological advancements.

COR theory posits that individuals strive to acquire and protect resources essential to their well-being, experiencing stress when these resources are threatened or lost (Hobfoll and Freedy, 2017). Widely applied to workplace stress and employee resilience, COR emphasizes that resources such as job security and organizational support are crucial for maintaining employee well-being and optimal performance. When employees perceive an inability to safeguard these resources, turnover intentions may increase (Anasori et al., 2021; Raper et al., 2024; Xiaoxin et al., 2025).

In this study, hotel robots are found to play a dual role. For some employees, the improvements in work efficiency; for others, they raise concerns about job security and signal a potential loss of valued resources (Gonzalez-Jimenez and Costa Pinto, 2024; Tojib et al., 2022). The relationship between employee outcomes and robot use is dynamic. The initial use of automation raises concerns about barriers to work, prompting firms to invoke labor development programs. As employees master human–robot collaboration fits, firms may adjust their automation strategies (Li et al., 2024; Tan et al., 2025). This iteration underscores the developmental nature of integrating labor and technology, thereby promoting a cycle of resource replenishment that aligns with COR’s conservation-restoration principles (Guo et al., 2023; Koo et al., 2023). This study advances COR theory by introducing resource trade-off cycles. In this cycle, automation can decrease job security for workers because machines take over certain tasks. At the same time, automation increases efficiency, streamlining processes and boosting productivity. This update to COR’s conservation principle shows how employees weigh gains against losses. When their resources change, such as when job stability falls but efficiency rises, employees have to rethink what matters most in their work environment. Unlike prior work (Zhang et al., 2023), researchers model this duality through moderated mediation, revealing how training transforms depletion into resilience. technological context (see Figure 1).

The integration of artificial intelligence and robots in hospitality fundamentally transforms service operations, impacting hotels, employees and guests. Existing research highlights both the benefits and challenges of robotics in hospitality (Rampersad, 2020; Shao et al., 2022), with job insecurity emerging as a key concern for employees. The perceived threat of robot replacement heightens job insecurity, leading to stress, reduced motivation and impacts on cognitive and emotional well-being (De Cuyper et al., 2020; Langerak et al., 2022). This insecurity can result in dissatisfaction, anxiety and increased turnover intentions (Nemteanu et al., 2021). Initially, employees may be curious about working alongside robots, but this can shift to insecurity if they feel their roles are at risk (Na et al., 2023). Adekiya (2024) noted that job insecurity primarily arises from fears of involuntary unemployment, which a lack of organizational support can exacerbate (Wang et al., 2020a, 2020b, 2020c). However, organizational support and skill development opportunities can mitigate these negative effects by providing employees with resources and security, thereby reducing job insecurity and further highlighting the importance of a supportive work environment for job stability. According to COR theory, Yam et al. (2023) found that fears of automation increase employee stress and anxiety, prompting them to seek additional training and internal transfers to protect their resources. Based on these insights, researchers propose the following hypothesis:

Turnover intention, influenced by factors such as job stress, satisfaction and perceived organizational support, refers to an employee’s intent to leave an organization or role in the near future (Yan et al., 2021a, 2021b; Yu et al., 2021). It is a significant predictor of actual turnover, strongly impacting employees’ decisions to exit (Wong and Cheng, 2020). Job insecurity, particularly during technological transitions like robot integration, is a key predictor of turnover intention (Brougham and Haar, 2020). When employees perceive a threat to their employment, their intentions to turnover increase. According to COR theory, perceived resource loss due to job insecurity motivates employees to seek new opportunities or disengage from their current roles. Since this turnover often results from perceived instability, researchers propose the following hypothesis:

The integration of robots in hotels has significantly enhanced service efficiency in tasks such as room service and customer interactions (Garcia-Haro et al., 2020; Zeng et al., 2020; Gupta et al., 2022; Lin and Mattila, 2021). However, as robots take on more responsibilities, employees may experience increased job insecurity, leading to anxiety and reduced job satisfaction. Job satisfaction, defined as employees’ positive perception of their roles and work environment (Khatun et al., 2022; Taj et al., 2020), is crucial in reducing turnover intentions. Research shows that job satisfaction is influenced by both organizational and personal factors (Díaz-Carrión et al., 2020). Employees who perceive strong training opportunities and a supportive environment tend to report lower job insecurity and turnover intentions (Díaz-Carrión et al., 2020; Viseu et al., 2020). From the COR theory perspective, perceived resource losses, such as increased workloads or insufficient support, heighten stress and diminish job satisfaction, thereby driving turnover intention. Thus, we propose that job satisfaction acts as a vital mediator between job insecurity and turnover intention. Accordingly, researchers propose the following hypotheses:

Perceived organizational support refers to the beliefs and subjective feelings of employees that the organization values their well-being and recognizes their contributions (Aldabbas et al., 2023; Wang et al., 2020a, 2020b, 2020c). It typically involves holistic organizational care, such as job resources and psychological support, which affects employees’ sense of belonging and loyalty (Chen et al., 2023; Li et al., 2020; Takaya and Ramli, 2020). When employees feel supported, they are more likely to perceive their importance within the organization, leading to increased effort and loyalty (Asghar et al., 2021; Eisenberger et al., 2020). Research shows a positive correlation between high levels of perceived organizational support (POS) and reduced turnover intentions, fostering a sense of belonging (Asghar et al., 2021). In addition, POS can alleviate negative emotions and stress during challenging work conditions, enhancing psychological well-being and further reducing turnover intentions (Akgunduz et al., 2023; Tetteh et al., 2020). From a COR theory perspective, POS serves as a vital resource by providing emotional, psychological and material support, motivating employees to mitigate job insecurity and lower turnover intentions (Bardoel and Drago, 2021; Eisenberger et al., 2020). Lu et al. (2024) found that providing appropriate training and psychological support in automated environments effectively reduces employee stress and enhances loyalty. Thus, researchers propose the following hypotheses:

Training and development opportunities are crucial for enhancing employee skills and competencies, as well as fostering organizational identification (Piwowar-Sulej, 2021; Rampa and Agogué, 2021). It focuses on enhancing employees’ skills and promoting continuous professional growth, typically through training programs, on-the-job learning or promotional opportunities (Tan et al., 2023). Practical training ensures employees maintain the knowledge and skills needed for high service quality and a positive organizational image (Nguyen and Malik, 2022). Research highlights the importance of equipping frontline staff with job-specific and behavioral skills to enhance service delivery and foster long-term career growth (Kamna and Ilkhanizadeh, 2022; Lan et al., 2021). Organizations investing in career development often see reduced turnover intentions (Fulmore et al., 2023).

Training can buffer the adverse effects of job insecurity on turnover intentions (Lin and Huang, 2021). Employees who perceive ample training are more likely to demonstrate loyalty and commitment, thereby reducing the likelihood of turnover (Sun et al., 2025; Memon et al., 2021). However, some argue that increased training may boost employees’ confidence in pursuing external opportunities, potentially increasing turnover intentions (Salleh et al., 2020; Vizano et al., 2021). In addition, high-potential employees may seek better prospects after gaining new skills and dissatisfaction with misaligned training can expedite turnover decisions (Deller, 2023). According to COR theory, training and development enable employees to build knowledge and skills, thereby reducing stress; however, they may also lead to negative outcomes. Consequently, researchers propose the following hypotheses:

Guangdong, as one of the first provinces in China to adopt automation in the hospitality industry, provides a unique and representative context for exploring the impact of robot integration. Its high level of internationalization offers valuable insights for countries/regions with similar economic and cultural backgrounds. Therefore, this study was conducted in Guangdong Province, China, focusing on frontline employees at 20 hotels who have integrated robots into their hotel operations. Eligible participants were at least 18 years old and had experience using robots in their work tasks. Data collection was facilitated by the HR managers of each hotel, who distributed questionnaires via a link to an electronic survey hosted on Wenjuanxing (Link to wjxLink to the cited website) through internal WeChat employee group chats. This approach ensured that employees could complete the survey privately and anonymously. Data collection took place in August 2024.

The questionnaire measured six constructs: usage of robots (Zhang et al., 2023), job insecurity, perceived organizational support (Akgunduz and Sanli, 2017), job satisfaction (Wang et al., 2020a, 2020b, 2020c), training and development opportunities and turnover intention (Yavas et al., 2003). In addition, it included demographic information and a screening question. All items were rated on a seven-point Likert scale from 1 (strongly disagree) to 7 (strongly agree).

Prior to the formal study, a pilot test was conducted with 20 hotel employees who interacted with robots, leading to revisions for improved clarity and comprehension. Scale validation involved calculating correlations and Cronbach’s alpha coefficients. The questionnaire was also reviewed by two hospitality researchers and two HR managers to ensure relevance and content validity. A direct-translation method was used to maintain consistency and accuracy for Chinese respondents (Wang and Wang, 2022).

From the initial responses, 36 incomplete surveys were excluded, resulting in 343 valid responses. G*Power analysis confirmed that this sample size exceeded the minimum requirement of 146 (effect size f² = 0.15, α = 0.05, power = 0.95), ensuring adequacy for analysis. The demographic breakdown of respondents was 56.3% female and 43.2% aged 18–24 years. Educational backgrounds were primarily high school or below (37.6%) and associate degrees (38.5%). Most participants held service positions (49.3%), with 28% having a tenure of 6 months to 1 year, 30.6% working in hotels with fewer than 50 rooms and 19.5% employed in boutique hotels (see Table 1).

Li and Zhang (2023) indicated that demographic characteristics (such as gender, age or job occupation) can lead to differences in employee behavior intentions. Related studies have pointed out that employee turnover intentions are significantly associated with tenure, hotel size and hotel type (Ampofo and Karatepe, 2022; Başer et al., 2025; Li et al., 2021). Therefore, in this analysis, this study controls for four variables: demographic characteristics (such as job occupation), tenure, hotel size and hotel type.

Before hypothesis testing, a Harman one-way test was conducted to assess common method bias (CMB). The cumulative variance explained was 28.5%, below the 50% threshold, indicating that CMB is not a concern in this study. PLS-SEM identifies multiple pathways by examining linear causality between variables (Hair et al., 2023), whereas fsQCA emphasizes causal complexity (Lazzari et al., 2022) and can identify multiple pathways by revealing the common effects of different combinations of variables on outcomes. The combination of the two analytical approaches tests hypotheses and explores potential alternative pathways, thereby enhancing the robustness, theoretical depth and timeliness of the study’s conclusions. Therefore, data analysis in this study was performed using SmartPLS 4.0 PLS-SEM and fsQCA 3.0 software.

The measurement model’s reliability, convergent validity and discriminant validity were rigorously assessed. Convergent validity was evaluated using Cronbach’s alpha, factor loadings, composite reliability (CR) and average variance extracted (AVE). Both Cronbach’s alpha and CR exceeded the recommended threshold of 0.7 and AVE values surpassed 0.5, indicating satisfactory construct reliability. Factor loadings ranged from 0.790 to 0.916, all of which exceeded the 0.7 threshold (see Table 2), confirming convergent validity (Hair et al., 2024).

To ensure that no multicollinearity existed before testing hypotheses, Variance Inflation Factor (VIF) values were examined for the variables of robot usage, job insecurity, perceived organizational support, job satisfaction, training and development opportunities and turnover intention. VIF values ranged from 1.718 to 3.808, well below the 5 thresholds, indicating no multicollinearity issues (Hair et al., 2023). In addition, the Heterotrait-Monotrait ratio (HTMT) was calculated to assess discriminant validity, with all values below 0.85, confirming adequate discriminant validity. Detailed HTMT results are presented in Table 3.

The structural model evaluation results are summarized in Table 4. The analysis revealed that robot usage has a positive effect on job insecurity (β = 0.328, p < 0.001), supporting H1. Job insecurity, in turn, has a positive impact on turnover intention (β = 0.315, p < 0.001), confirming H2. In addition, job satisfaction negatively influences turnover intention (β = −0.277, p < 0.001), supporting H3. Job satisfaction also significantly mediates the relationship between job insecurity and turnover intention (β = 0.095, p < 0.001), supporting H4. Perceived organizational support negatively correlates with turnover intention (β = −0.312, p < 0.001), confirming H5. In addition, it significantly mediates the relationship between job insecurity and turnover intention (β = 0.101, p < 0.001), supporting H6.

Training and development opportunities significantly moderated the relationship between job insecurity and turnover intention (β = −0.153, p < 0.05) and between perceived organizational support and turnover intention (β = −0.101, p < 0.05), supporting H7b and H7c. However, training and development opportunities did not significantly moderate the relationship between job satisfaction and turnover intention (β = −0.082, n.s.), thereby rejecting H7a.

The model explained 41.8% of the variance in turnover intention (R² = 0.418), surpassing the 0.30 threshold and demonstrating good explanatory power. Model fit was confirmed with a standardized root mean square residual (SRMR) of 0.067, below the acceptable cutoff of 0.08. Furthermore, the positive Q² value for turnover intention (Q² = 0.036) validated the model’s predictive capability.

After model estimation, fsQCA was used to further analyze the data and identify factors associated with high turnover intention. The previous PLS-SEM findings clarified the relationships among variables and provided empirical evidence on how robot usage affects employee turnover intention. Recognizing that multiple factors influence turnover intention, with no single element dictating it entirely (Lazzari et al., 2022; Yan et al., 2021a, 2021b), researchers leveraged the PLS-SEM results to apply the fsQCA approach. This allowed us to explore effective strategies for mitigating turnover intention from a configurational perspective and identify optimal solutions (see Figure 2).

Data calibration is essential for conducting fsQCA (Fainshmidt et al., 2020). This process converts raw data into fuzzy sets with affiliation scores ranging from 0.00 to 1.00, accurate to two decimal places (Ragin, 2007). Scores of 0.00 and 1.00 denote complete non-membership and full membership, respectively, while 0.50 indicates maximum ambiguity and was adjusted to 0.499 based on prior studies (Drăgan et al., 2023). Calibration was performed using fsQCA 3.0 software, setting the 95th percentile, mean and 5th percentile for each construct as full membership, crossover point and full non-membership, respectively, following the direct calibration method recommended in the literature (Kang and Shao, 2023). This range of thresholds is widely used in fsQCA studies to ensure the effective inclusion of high-intensity cases and to maintain the scientific validity and comparability of the analysis.

After calibrating the data into fuzzy sets, researchers conducted a necessity analysis focusing on the dependent variable, turnover intention (TI), to determine if any of the five antecedent conditions, usage of robots (UOR), job insecurity (JI), POS, job satisfaction (JS) and training and development opportunities (TDO), were necessary predictors of TI. The adherence of cases to these conditions was measured by “consistency” scores, ranging from 0 to 1 (Liang et al., 2020). A condition was deemed necessary if its consistency score exceeded 0.90 (Sukhov et al., 2023). As shown in Table 5, the consistency scores for the five antecedents ranged from 0.425 to 0.764, all of which were below the 0.90 threshold, indicating that none were necessary for TI. Therefore, these five antecedent conditions can be combined for further analysis of their effects on TI.

Following the necessity analysis, researchers conducted a sufficiency analysis to identify combinations of conditions that predict turnover intention. According to Pappas and Woodside (2021), researchers first generated a truth table of all logical condition combinations, calculated as 2^k (where k is the number of conditions). This study sets the consistency threshold at 0.60, a value referenced from existing empirical research in the field of organizational management (Hossain et al., 2024; Liang et al., 2020; Wang et al., 2023). In management and organizational contexts, causal relationships often exhibit complexity and heterogeneity. A consistency standard of 0.6 helps identify causal paths that hold under specific combinations of conditions but are difficult to detect in the overall sample. This approach not only ensures the logical relevance of the analysis but also allows us to reveal the diverse and authentic configuration patterns in organizational practices more comprehensively. The fsQCA results produced three solutions: complex, economical and intermediate. Researchers choose the intermediate solution because the causal conditions of the complex solution may have redundant factors that affect the interpretability of the results. The parsimonious solution retains only the core conditions, potentially overlooking some peripheral conditions that are critical in a particular context. Therefore, the intermediate solution strikes a balance between the two, eliminating unnecessary conditions while retaining theoretically plausible causal pathways, thereby making the conclusion more explanatory (Kang and Shao, 2023). As shown in Table 6, six configurations were identified that lead to high turnover intention, with an overall solution coverage of 0.799 and consistency of 0.780, indicating strong explanatory power and consistency.

The first configuration (T1) revealed that high job insecurity, combined with low perceived organizational support, leads to elevated turnover intention (consistency = 0.804, raw coverage = 0.593). The second configuration (T2) indicated that the absence of perceived organizational support, job satisfaction and training and development opportunities contributes to turnover intention (consistency = 0.891, raw coverage = 0.352). The third configuration (T3) revealed that high robot usage alongside low perceived organizational support and low job satisfaction is associated with increased turnover intention (consistency = 0.836, raw coverage = 0.439). The fourth configuration (T4) demonstrated that high job satisfaction, when combined with high robot usage and job insecurity, results in high turnover intention (consistency = 0.802, raw coverage = 0.448). The fifth configuration (T5) revealed that high job satisfaction, in the presence of high job insecurity and high training and development opportunities, is associated with high turnover intention (consistency = 0.804, raw coverage = 0.442). Finally, the sixth configuration (T6) indicated that high robot usage, high job insecurity and high training and development opportunities contribute to increased turnover intention (consistency = 0.791, raw coverage = 0.406).

This study, grounded in COR theory, explores the complex relationship between robot usage and employee turnover intentions, highlighting the mediating roles of job insecurity, job satisfaction and perceived organizational support, as well as the moderating effects of training and development opportunities. By using both PLS-SEM and fsQCA methodologies, researchers examined the multifaceted interplay of these variables.

Our PLS-SEM analysis reveals that robot usage significantly increases job insecurity (H1), aligning with existing literature (Koo et al., 2023) and suggesting that automation heightens concerns about job stability, especially in roles susceptible to automation. This heightened anxiety likely drives employees to seek more stable employment alternatives, indicating that organizations must address job insecurity to effectively mitigate turnover intentions (Adekiya, 2024; Na et al., 2023; Nemteanu et al., 2021). Furthermore, job insecurity emerges as a critical predictor of turnover intentions (H2), where psychological stress prompts employees to explore other opportunities (Yu et al., 2021). Organizations can reduce turnover rates by implementing strategies that alleviate job insecurity while leveraging productivity gains (Lu et al., 2024; Wong and Cheng, 2020).

In addition, our analysis reveals significant negative relationships between job satisfaction and perceived organizational support, as well as between these factors and turnover intention (H3, H5). While previous studies have emphasized job autonomy, our findings highlight the importance of job satisfaction and perceived organizational support in meeting employees’ psychological needs. Organizations prioritizing these factors can enhance talent retention and satisfaction (Khatun et al., 2022; Yu et al., 2021). Moreover, both job satisfaction and perceived organizational support mediate the impact of job insecurity on turnover intention (H4, H6). High levels of satisfaction and perceived support can mitigate the effects of job insecurity on turnover intentions, consistent with findings from Addai et al. (2022).

Our findings also confirm the moderating role of training and development opportunities (H7b, H7c), indicating that such opportunities can lessen the negative effects of job insecurity on turnover intentions and enhance perceived organizational support (Brougham and Haar, 2020; Chen et al., 2023; Yao et al., 2024). This investment not only develops professional competencies but also fosters a sense of belonging among employees (Fulmore et al., 2023; Tan et al., 2023). However, training and development opportunities do not significantly moderate the relationship between job satisfaction and turnover intention (H7a). The rejection of H7a aligns with emerging evidence on technology-driven skill markets (Ekuma, 2024). This may reflect a dual role of training: while it enhances loyalty by signaling organizational investment (Fulmore et al., 2023), it also equips employees with skills that increase their external marketability. For example, a front-desk employee trained in robot maintenance may leverage this skill to seek higher-paying tech roles, thereby offsetting the need for additional satisfaction (Salleh et al., 2020). COR theory explains this paradox by examining how people assess their resources. Training provides employees with new skills, enabling them to assess the resources they already have. When employees perceive that their organization’s support for initiatives such as career growth is lacking, they may seek better opportunities elsewhere (Hobfoll and Freedy, 2017).

In addition, this study controlled for variables such as job occupation, tenure, hotel size and hotel type. Although the control variables did not show statistical significance, they still have the potential to explain the differences between robot exposure and employee responses. For example, employees in frontline service departments (such as housekeeping and food and beverage) are more likely to interact directly with robots. At the same time, back-office positions or mid-to-senior-level roles have relatively lower interaction frequencies. Such interaction differences may influence employees’ perceptions of technological replacement, training needs and career development, thereby indirectly affecting their turnover intentions and work attitudes.

The fsQCA results identify six configurations leading to high turnover intention, notably highlighting that low perceived organizational support exacerbates turnover intentions among employees facing job insecurity (Solution 1). Other configurations demonstrate that the interplay between job satisfaction, job insecurity and training opportunities significantly influences turnover intention, thereby validating our hypotheses (H1, H5, H7c). While PLS-SEM identified linear relationships (e.g. robot usage → job insecurity), fsQCA revealed non-linear pathways. For instance, high robot usage alone (PLS-SEM: β = 0.328) does not guarantee turnover unless combined with low POS (T3: consistency = 0.836). This aligns with COR theory’s emphasis on resource bundles rather than isolated factors. This comprehensive analysis enhances our understanding of the factors driving turnover intentions.

Notably, the fsQCA results (T5–T6) indicate that high training opportunities and high turnover intentions can coexist, a phenomenon that contradicts the conventional wisdom in traditional human resource management, which holds that “training helps retain employees.” However, from the perspective of COR theory, on the one hand, training provides employees with important resources to enhance their professional capabilities and strengthen their personal competitiveness; on the other hand, training may also enhance employees’ awareness of their own market value, prompting them to reassess the resource environment provided by their current job. When employees perceive that there is limited space for growth within the organization or that there are more attractive resource returns in the external environment, they may choose to leave the organization to further accumulate and optimize their resources. It can thus be seen that training is not only a resource investment by the enterprise in its employees, but may also become a trigger for employees’ intention to leave the organization. This dual mechanism is highly consistent with the logic of proactive behavior emphasized by COR theory, whereby individuals take proactive actions to obtain, maintain and enhance resources.

Overall, our research indicates that robot integration increases job insecurity, supporting COR theory, which posits that employees view job security as a valuable resource and respond to its threat by seeking to preserve it. Job satisfaction acts as a mediator for turnover intentions, helping employees manage feelings of insecurity (Díaz-Carrión et al., 2020; Khatun et al., 2022), while perceived organizational support enhances commitment and mitigates job-related insecurities (Eisenberger et al., 2020; Aldabbas et al., 2023). Training opportunities have a mixed effect; they improve skills and boost organizational commitment but may also provide employees with resources to pursue external opportunities if more appealing prospects arise (Fulmore et al., 2023; Memon et al., 2021).

This study advances the literature on robot usage and employee turnover intention through three key contributions. First, it extends the COR theory by exploring the circular reciprocity between organizations and employees, demonstrating how robot usage increases job insecurity, which in turn affects turnover intentions through job satisfaction and perceived organizational support (Zhang et al., 2023).

Second, it develops COR literature by showing that job satisfaction, perceived organizational support and training opportunities mitigate turnover intentions, enhancing employee motivation and organizational commitment. We extended the COR theory by identifying cycles in resources that result from automation. Specifically, technological resources, such as robots, can create dynamic shifts between loss and gain. When companies invest in employee training for using these technologies, it might unintentionally lead to higher turnover rates. This occurs when employees assess their job resources and opportunities based on external environments, favoring what they could find elsewhere rather than what is available within their current organization. These findings not only highlight the value of training as an investment by organizations to enhance employee skills and productivity but also challenge the traditional human resource management view of training as a one-way tool for talent retention. They reveal that while training improves employee capabilities, it may also trigger a reassessment of resources and lead to increased turnover intentions. This study deepens researchers’ understanding of the cognitive and behavioral mechanisms of employee resources and further expands the applicability and explanatory power of COR theory in organizational contexts.

Finally, the study identifies multiple configurations leading to high turnover intention by integrating PLS-SEM and fsQCA. In addition, fsQCA emphasizes causal complexity and equifinality, suggesting that the effect of robot usage on employee turnover intention is not linear, but depends on the interaction of job insecurity, support, satisfaction and training. This finding enriches COR theory, which is based on the principle of reciprocal symbiosis between organizations and employees. It also suggests that the success of technological change is not only dependent on the technology itself (Yan et al., 2024; Yan et al., 2025), but is also influenced by a combination of the accumulation of resources and the work environment to which employees are adapted, providing a nuanced understanding of the factors that influence turnover.

In today’s rapidly evolving hospitality landscape, the increasing prevalence of robotics and artificial intelligence presents hotel managers with critical opportunities and challenges. This study highlights that effective robot integration not only enhances operational efficiency but also influences employee perceptions and turnover intentions. By leveraging these insights, managers can create environments that reduce turnover and improve organizational health.

To integrate robots successfully, managers must address the psychological impacts of automation, recognizing that employees may perceive robots as a threat. Path 1 (T1) of fsQCA reveals high levels of job insecurity, accompanied by low levels of perceived organizational support. Managers can prioritize weekly one-on-one meetings to address employee concerns and allocate budgets for mental health programs. Investing in comprehensive training programs can mitigate these concerns, fostering trust and helping staff perceive robots as partners rather than competitors. Transparent communication is essential; informing employees about how robotics will affect their roles and the workplace, coupled with regular feedback sessions and open forums, empowers staff and fosters an inclusive culture. The use of robotics may lead to employees’ dependence on automation, weakening their autonomous decision-making abilities and career paths.

Moreover, tailored training and development initiatives are crucial. Specialized programs that prepare employees for effective human–robot interaction and practical automation applications enhance confidence and skills, strengthening emotional attachment and promoting engagement and productivity. Path 3 of fsQCA (T3) shows that in situations where high robot usage coexists with low job satisfaction, employees are more likely to develop a tendency to leave, suggesting that high automation may lead to issues such as job monotony and a lack of professional fulfillment. In response, managers can try introducing job rotation mechanisms (such as rotating between robot supervision and customer service positions) to reduce job monotony and enhance employee engagement and work commitment. Meanwhile, although the moderating effect of training, development and opportunities on job satisfaction and turnover intention, as hypothesized in H7a, was not significantly supported, this does not mean that organizations should abandon efforts to improve satisfaction. This study recommends implementing resource-conserving strategies for job satisfaction training. This includes combining technical training with retention incentives, such as tuition reimbursement tied to a two-year commitment. It is important to align skill development with opportunities for internal career advancement. In addition, incorporating “future-proofing” workshops can help alleviate anxiety around technological changes. On the contrary, the results suggest that relying solely on emotional motivation or traditional training methods may be insufficient when addressing the adaptive challenges posed by technological change. Therefore, when formulating training strategies, companies should place greater emphasis on achieving resource-oriented goals, such as enhancing employees’ technical adaptability, career development awareness and sense of resource control, so that employees are not only satisfied with their current work situation but also have the confidence to cope with future changes. For small and medium-sized budget hotels with limited resources, low-cost methods such as online courses, internal knowledge sharing and collaborative learning can be used to enhance the coverage and effectiveness of training, ensuring that every employee receives the necessary support and growth opportunities. Only by effectively linking the practical needs of employee development with organizational change goals can training and management strategies truly play a role in stabilizing human resources and reducing turnover risks in the context of technological transformation.

Ultimately, promoting technological integration in the hotel industry requires striking a balance between innovation and empathy. Managers should promote technological progress while fostering an organizational culture that prioritizes employee well-being, ensuring that employees feel safe, valued and engaged, thereby transforming potential challenges into opportunities for organizational growth and development. However, it is worth noting that this study is grounded in a specific cultural and institutional context, particularly concerning Chinese organizational culture and human resource practices. Therefore, although the study provides a series of management recommendations for human resource managers, their applicability should be carefully considered in light of the specific circumstances of different countries and regions.

While this study offers valuable insights, it has several limitations. First, data were collected only from Guangdong, China, which may not represent employee responses to robots and job insecurity in different cultural contexts. Future research should include diverse global settings to compare how cultural factors influence job satisfaction, organizational support, training opportunities and turnover intentions. Second, the rapid advancement of robotics may affect the long-term relevance of the findings. Longitudinal studies that track employee perceptions over time would better capture how job insecurity and turnover intentions evolve in response to technological integration. Third, while fsQCA identifies configurations, it does not capture dynamic interactions over time. Future studies should use longitudinal fsQCA to track how turnover pathways evolve with prolonged robot exposure. Forth, the applicability of these recommendations may be influenced by the type, size and cultural context of the hotel. Future studies should apply this dual-method approach to compare the impact of automation in collectivist (e.g. Japan) versus individualist (e.g. the USA) cultures. For example, fsQCA could identify whether high POS universally offsets job insecurity or is culturally contingent, advancing COR theory’s cross-cultural applicability. Future research could further expand the sample to include the perspectives of more experienced or managerial employees, thereby providing a more comprehensive understanding of the impact of robot use on job insecurity and turnover intentions among employees at different levels. Furthermore, this study did not thoroughly investigate the interaction mechanisms of various control variables. Given the potential systemic differences between job types and the degree of contact with robots, future research may consider using multi-group analysis to systematically compare the psychological responses and behavioral tendencies of different groups in the context of technology application, thereby further revealing the heterogeneous effects of robotics technology on outcome variables such as employee turnover intentions. Finally, some scholars have pointed out that the sensitivity of Harman’s single-factor test for assessing CMB is limited (Bozionelos and Simmering, 2022; Howard et al., 2024; Zhang et al., 2022). However, in the design of this study, methodological bias was mitigated through control strategies, including anonymous completion and item randomization. Therefore, the results of this test can still provide a basic reference for assessing CMB. Of course, future research could further integrate more advanced technical approaches (such as marker variable methods or latent variable factor analysis) to enhance the ability to identify potential biases and strengthen the robustness of empirical conclusions.

Corrigendum: It has come to the attention of the publisher that the article Xiaoxin L, Konar R, Ali F (2025), “From automation to employee loyalty: understanding the balance between robots and workforce stability”. Journal of Hospitality and Tourism Technology, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/JHTT-11-2024-0720 mistakenly reported a composite reliability value of 1.054 in Table 2; this value was generated by the statistical software output. The correct value should be 0.957, which is fully consistent with the already-reported Cronbach's alpha (0.953), AVE (0.762), and factor loadings. All factor loadings, AVE, alpha, HTMT, VIF and structural path estimates remain valid, and this error does not affect the article's overall findings, theoretical contributions, or conclusions. Further, the authors wish to document the details of the ethics approval and informed consent process employed for this research, which do not feature in the body of the manuscript. This research was approved by the Human Ethics Committee at Taylor's University (HEC 2024/119) on 26 May 2024, and it was conducted in accordance with recognized institutional and international ethical guidelines. Prior to commencing data collection, informed consent was obtained from all participants. Participants were informed of the study's aims, risks and benefits, and the right to refuse participation or withdraw from the study at any time; the authors confirm that the autonomy and confidentiality of the participants were maintained and respected throughout, and no personal or identifying data were collected. This corrigendum is issued to ensure clarity and transparency for readers. The authors sincerely apologize for any inconvenience caused.