The mental health implications of artificial intelligence adoption: the crucial role of self-efficacy

AI adoption and burnout

We propose that AI adoption in an organization will increase employee burnout. Adopting AI is a complicated and multi-stage process that begins with raising awareness and conducting exploratory research and continues through large-scale implementation and ongoing refinements (Kaplan and Haenlein, 2020). A comprehensive assessment of an organization’s preparedness is required, which involves looking at, for example, data availability, human capital, leadership support, technical infrastructure, and the culture of the company (Berente et al. 2021; Chen et al. 2023; Magistretti et al. 2019; Makridis and Mishra, 2022; Uren and Edwards, 2023). The proper prioritization of AI use cases is required for organizations to achieve their strategic goals (Davenport and Ronanki, 2018). In addition, the ethical, social, and governance concerns raised by AI technologies must be addressed in order to facilitate their responsible and transparent deployment (Chen et al. 2023; Dwivedi et al. 2019). In order to fully utilize AI and keep employees engaged and well-cared-for, organizations must adjust their strategies, provide reskilling and upskilling opportunities, and cultivate a culture of lifelong learning (Magistretti et al. 2019; Makridis and Mishra, 2022; Uren and Edwards, 2023).

Prolonged interpersonal stresses in one’s job can lead to burnout, a mental condition defined by emotional weariness, depersonalization, and a diminished sense of personal success (Maslach and Leiter, 2016; Maslach et al. 2001). Depersonalization is characterized by a negative, detached, and unempathetic attitude toward one’s work and colleagues, while emotional exhaustion involves feelings of being emotionally drained and overextended (Maslach et al. 2001; Shirom and Melamed, 2006). Reduced self-confidence causes unfavorable assessments of one’s abilities and, ultimately, discontent with one’s life (Maslach et al. 2001; Leiter and Maslach, 2016). Burnout sets in when a person’s emotional and physical resources are exhausted because their work environment is unsuitable (Leiter and Maslach, 2003). Research has shown that it can harm both individuals’ and organizations’ well-being. Some of these negative outcomes include problems with mental and physical health, cognitive functioning, job satisfaction, organizational commitment, turnover intentions, productivity, absenteeism, healthcare costs, and service quality (Melamed et al. 2006; Salvagioni et al. 2017; Alarcon, 2011; Lee and Ashforth, 1996; Bakker et al. 2014; Taris, 2006). Hassard et al. (2018) found that healthcare costs, decreased productivity, and staff turnover contribute significantly to burnout’s substantial financial consequences.

We argue that the adoption of AI in the workplace will increase employee burnout based on the JD-R model, the conservation of resources (COR) theory, and the technostress model (Tarafdar et al. 2007).

First, the JD-R model states that job demands and job resources are the two main categories of job characteristics and that they interact to influence employee well-being outcomes, including burnout (Bakker and Demerouti, 2017). Job demands require mental and physical exertion from the worker, which might negatively affect their health (Demerouti, Bakker, Nachreiner, and Schaufeli, 2001). AI adoption can be seen as a job demand because it brings about new problems and complexities that require employees to acquire new skills, adjust to new work processes, and collaborate with AI systems (Makridis and Mishra, 2022; Pereira et al. 2023; Uren and Edwards,2023). Furthermore, workers may feel pressured to work faster and more efficiently if AI systems are implemented in the workplace (Berente et al. 2021; Chen et al. 2023), and stress and fatigue, two hallmarks of burnout, might worsen as a result of the increased workload (Maslach et al. 2001). Employees may also face additional psychological expenses due to job demands, such as the need to constantly learn and improve their skills to remain relevant in their professions (Dwivedi et al. 2021; Zirar et al. 2023).

Second, the COR theory (Hobfoll, 1989) supports the connection between AI adoption and employee burnout. Resources comprise the things, traits, circumstances, and energy that an individual values; according to this theory, people work hard to acquire, keep, and safeguard their resources (Hobfoll, 2001). According to Karasek (1979) and Ryan and Deci (2000), AI can undermine employees’ resources, including their confidence, independence, and job stability. Employees may suffer from burnout if they feel emotionally and mentally drained from worrying about losing these resources as a result of AI implementation (Hobfoll, 2001).

Third, the technostress model (Tarafdar et al. 2007) further clarifies the association between technology adoption and employee well-being. Technostress is the stress experienced by individuals due to the use of information and communication technologies (ICTs) in the workplace (Tarafdar et al. 2019). The adoption of AI technologies can be viewed as a specific instance of ICT implementation; thus, the technostress framework can be applied to understand AI’s impact on employee burnout. Tarafdar et al. (2007) identified five key technostress creators: techno-overload, techno-invasion, techno-complexity, techno-insecurity, and techno-uncertainty. These technostress creators can manifest during AI adoption, leading to increased stress and burnout among employees (Khedhaouria and Cucchi, 2019). Specifically, techno-overload occurs when AI technologies increase workload and accelerate work pace, compelling employees to work more intensively (Borle et al. 2021). Techno-invasion refers to the erosion of the boundaries between one’s work and personal life due to constant connectivity and availability expectations associated with AI systems (Tarafdar et al. 2007). Techno-complexity arises when employees find AI technologies difficult to understand and use, causing feelings of inadequacy and frustration (Tarafdar et al. 2011). Techno-insecurity is the fear of job loss or displacement due to AI, while techno-uncertainty denotes the continual changes and upgrades in AI technologies that require employees to adapt and acquire new skills regularly (Khedhaouria and Cucchi, 2019). In light of these arguments, the following hypothesis is proposed.

Hypothesis 1: AI adoption in an organization will increase burnout.

AI adoption and job stress

This paper suggests that AI adoption in an organization will increase job stress among employees. Job stress is a complex concept that encompasses the mental, emotional, and physical responses that workers display when confronted with challenges at work that are too difficult for them to handle (Lazarus and Folkman, 1984). When people face substantial pressure, difficulties, and demands at work, it can lead to negative bodily reactions as well as psychological strain (Ganster and Rosen, 2013).

We propose that the adoption of AI in an organization will increase job stress among employees by drawing on the JD-C) model (Karasek, 1979) and the person–environment (P–E) fit theory (Edwards et al. 1998).

First, according to the JDC model, job stress is caused by the interaction between job demands and job control. Workload, time constraints, and role conflicts are all examples of psychological stressors that employees face on the job (Karasek and Theorell, 1990). Conversely, job control is defined as the degree to which workers can exercise agency and competence in their work (Karasek, 1979). High job demands, coupled with a lack of control over those demands, increase job-related stress, according to the JD-C model (Häusser et al. 2010). One interpretation of AI’s widespread use in an organization is that it will increase the demands imposed on human workers. When AI technologies are implemented, personnel are typically required to learn new skills, adjust to different work processes, and work with AI systems (Pereira et al. 2023; Zirar et al. 2023). As workers try to adapt to the new demands imposed by AI, they may find themselves under stricter time and effort constraints (Bankins et al. 2024; Pereira et al. 2023). Role ambiguity and conflict may also arise from AI systems’ complexity and opaqueness as workers try to make sense of their evolving duties in relation to AI (Bankins et al. 2024; Pereira et al. 2023).

At the same time, AI adoption could make workers feel that they have less control over their jobs. As AI systems take over tasks and decision-making roles previously performed by humans, employees might experience a reduction in autonomy and decision-making power (Budhwar et al. 2022). The opaque nature of many AI algorithms can further erode employees’ sense of control, as they may find it challenging to understand and influence the outcomes generated by these systems (Shrestha et al. 2021). Furthermore, when AI is implemented in the workplace, individuals may experience heightened job stress in line with the JD-C model’s description of higher job expectations and lower job control.

Secondly, the P–E fit theory supports the idea that AI adoption can increase stress in the workplace. In this view, stress develops whenever an individual’s skills, needs, and values, for example, do not align with what is required or provided by their workplace (Edwards et al. 1998). The widespread use of AI can lead to a skills gap between workers’ present abilities and the abilities required by the technology (Makarius et al. 2020). A skills gap could emerge, adding to employee stress, as AI systems are always evolving and requiring continuous learning and adaptation (Nam, 2019). Another issue with AI adoption is that it can make workers feel less empowered than they actually are (Pereira et al. 2023). Workers may feel powerless because their desire for control conflicts with the reality of their work environment, especially as AI takes over decision-making and limits employees’ discretion. According to the P–E fit theory, this misalignment can intensify job stress (Edwards et al. 1998).

Thirdly, the socio-technical systems theory (Trist and Bamforth, 1951) highlights the link between AI adoption and job stress. This theory emphasizes that an organization’s social and technological subsystems must be aligned to improve employee performance and well-being (Baxter and Sommerville, 2011). A new set of tools, methods of operation, and criteria for making decisions are created by the introduction of AI (Vrontis et al. 2022). Workers may find that their needs conflict with those of the AI-enhanced workplace if the social subsystem—which comprises elements like job responsibilities, abilities, and interpersonal relationships—is not properly adjusted to the new technical subsystem (Pereira et al. 2023; Zirar et al. 2023). Because of this misalignment, workers may experience more job stress as they try to adapt to their new positions, learn necessary skills, and deal with the social effects of AI adoption (Shrestha et al. 2021).

Hypothesis 2: AI adoption in an organization will increase job stress.

Job stress and employee burnout

We argue that job stress will increase burnout among employees based on the COR theory (Hobfoll, 1989), the transactional model of stress and coping (TMSC; Lazarus and Folkman, 1984), and the JD-R model (Demerouti et al. 2001).

To begin, according to the COR theory, people work hard to acquire, keep, defend, and cultivate resources that they value. Things, qualities, circumstances, or energies that an individual values or that help them acquire more important resources are examples of resources (Hobfoll et al. 2018). Individuals experience stress when they are at risk of losing resources, actually lose resources, or fail to acquire resources after making substantial investments (Hobfoll, 2001). Workplace stress can be viewed as a condition of diminished resources. According to Bakker and Demerouti (2017), workers experience high levels of workplace stress because they are being asked to do more with less. Emotional and physical tiredness, as well as mental stress, may be signs of resource depletion (Alarcon, 2011). Burnout can develop after a long period of work under stressful conditions and resource scarcity (Hobfoll et al. 2018).

Second, according to the TMSC, people experience stress when they feel like they lack the coping resources necessary to meet the demands of a scenario. This suggests that cognitive appraisal and coping mechanisms are integral parts of the transactional process through which an individual experiences stress (Lazarus, 1999). Workplace stress is more common among workers who believe their workload exceeds their ability to cope compared to those who do not hold this belief (Goh et al. 2015). The essential component of burnout is the physical, emotional, and mental depletion created by prolonged exposure to job stress (Maslach et al. 2001). According to the TMSC, burnout can occur when workers’ emotional and cognitive resources are depleted due to an inability to cope with job pressures (Guthier et al. 2020). This is in line with the COR theory’s supposition that depletion and loss of resources lead to burnout (Hobbow et al. 2018).

Third, the JD-R model shows how stress at work can lead to burnout. Job demands and job resources are the two main branches of job characteristics (Demerouti et al. 2001). Job demands comprise the social, organizational, and physically demanding factors of a job that cause one to incur physiological and psychological expenses over time (Bakker and Demerouti, 2007). High levels of emotional and mental strain, as well as time constraints, are examples of demands in the workplace (Bakker et al. 2014). Meanwhile, job resources are the physical, mental, social, and organizational parts of a job that help workers accomplish their goals, cope with stress, and develop themselves professionally and personally (Bakker and Demerouti, 2007). Social support, autonomy, and performance evaluation are examples of job resources (Van der Heijden et al. 2019). Worker strain and burnout are both predicted by the JD-R model, which states that workers are more likely to experience the former when job demands are high and job resources are low (Demerouti et al. 2001). Burnout is characterized by emotional, mental, and physical tiredness; it can be caused by job stress, which occurs when demands are high and resources are scarce (Maslach et al. 2001).

Hypothesis 3: Job stress will increase burnout.

The mediating role of job stress in the relationship between AI adoption and burnout

Examining how work stress mediates the association between AI adoption in an organization and burnout is one of the main goals of this study. By combining the JD-R model, the COR theory, and the TMSC, we expect job stress to function as a mediator in the AI adoption-burnout link.

First, employee well-being and organizational outcomes are affected by the interplay between job demands and job resources, which are classified according to the JD-R model (Bakker and Demerouti, 2017). AI adoption in an organization will create a need for more people to learn and adapt to new technology, more work to accomplish, and possibly more role ambiguity (Bankins et al. 2024; Pereira et al. 2023). The JD-R model postulates that burnout and other negative effects may occur when work demands exceed available resources.

Secondly, according to the COR theory, stress on the job mediates the link between AI adoption and burnout. According to this theory, people work hard to get what they want and keep what they have (Hobfoll et al. 2018). Workers’ confidence, independence, and job stability are among the resources that can be jeopardized by the introduction of AI. Workers may suffer from emotional and mental exhaustion and burnout if they worry that AI integration will cost them their resources (Hobfoll, 2001).

Finally, the function of job stress as a mediator between AI adoption and burnout is further explained by the TMSC. This paradigm proposes that when people face a challenge, like adopting AI, they carry out a cognitive appraisal to determine how serious the challenge is and what resources they possess for dealing with it (Lazarus and Folkman, 1984). Workers may feel more pressure at work if they worry that AI adoption may undermine their job stability, work identity, or ability to execute their jobs well and if they do not think they can deal with these issues. Employees may suffer from burnout if they constantly experience a level of stress that they cannot handle.

In light of the arguments outlined above, we provide the following hypothesis.

Hypothesis 4: Job stress will mediate the association between AI adoption in an organization and burnout.

The moderating influence of self-efficacy in AI learning on the link between AI adoption and job stress

We hypothesize that the increasing effect of AI adoption on work stress will be weakened by an individual’s confidence in their AI learning capabilities. In this context, self-efficacy in AI learning is a person’s confidence in their ability to learn about and work with AI (Bandura, 1977). More specifically, this context-specific construct signifies how confident a person is in their ability to interact with AI systems, understand AI ideas, and effectively use AI tools (Kim and Kim, 2024). The concept of self-efficacy in AI learning originates from Bandura’s SCT (Bandura, 1986). This theory posits that how someone perceives their talents impacts their motivation, actions, and achievement in a certain area. One’s perceptions of their ability to learn AI can influence how much time and effort people are prepared to devote to AI-related tasks, how well they handle setbacks, and how effectively they practice what they have learned (Kim and Kim, 2024).

The link between AI adoption and employee job stress can be better understood by delving into the moderating effect of employee self-efficacy in AI learning. Based on the TMSC and SCT, we propose that the effect of the adoption of AI on job stress will be weakened if employees have faith in their capacity to understand and utilize the technology.

First, the SCT proposes that an individual’s self-efficacy—their confidence in their abilities to carry out the actions that will lead to a desired outcome—strongly influences their reasoning, drive, and emotions (Bandura, 1997). The term “self-efficacy in AI learning” is used in the context of AI adoption to describe an employee’s belief in their capacity to acquire and use knowledge and skills pertaining to AI (Bandura, 1986). Individuals with higher self-efficacy are less stressed, more inclined to perceive setbacks as learning experiences, and more willing to persevere in the face of adversity (Bandura, 1997). Furthermore, the TMSC sheds light on how self-efficacy factors into stress evaluation (Lazarus and Folkman, 1984). According to this model, people perform a cognitive assessment when confronted with a possible stressor, such as the implementation of AI at work, to determine the gravity of the situation and how well they can cope (Lazarus and Folkman, 1987). Concerning AI, employees who have faith in their abilities to learn and adapt to new technologies are likely to view AI adoption as a task they can accomplish rather than an overwhelming danger.

Workers who have high levels of self-efficacy in AI learning also have faith in their abilities to learn and use AI efficiently (Bandura, 1997). According to SCT, this group is more likely to see the widespread use of AI as a chance for professional and personal development rather than a danger (Bandura, 2012). These workers are always looking for new ways to improve their knowledge and abilities and are quick to adopt new technology (Chae et al. 2019). The demands of AI adoption can be better managed by workers who believe in their capability to learn AI than those who do not hold this belief.

Additionally, the TMSC (Lazarus and Folkman, 1987) suggests that these individuals can perceive the adoption of AI as a manageable stressor during the main assessment process since they are confident in their ability to acquire the essential abilities needed to work with AI. Their coping resources, including their AI learning capacities, are likely to be deemed appropriate during the secondary evaluation process to fulfill the demands of the situation. Workers who have faith in their abilities to learn AI are relatively unlikely to feel threatened by the adoption of AI in the workplace. This is because they are confident in their ability to adapt and profit from the process.

For example, if an organization uses AI to power its CRM systems, workers with high self-efficacy in AI learning could view the situation as a chance to hone their customer service skills while also effectively using the data AI provides to craft unique encounters for each client. These workers will also be more inclined to take part in training programs, ask for help when they need it, and try using new CRM technologies in order to reach their full potential. Moreover, they can adapt to new technology with less stress than other workers since they actively seek out resources to develop their AI competencies and engage in proactive learning activities. As a result, they do not worry excessively about failing to meet expectations and are confident in their ability to use an AI-powered CRM system to improve their company’s performance.

This situation is reversed when workers lack self-efficacy in their abilities to understand and utilize AI-related information (Bandura, 1997). SCT claims that these people are more prone to view AI adoption as a threat to their job security and competence, owing to their heightened anxiety and tension (Bandura, 2012). They might act in ways that prevent them from learning, resist learning opportunities, and have trouble adjusting to new technologies (Charness and Boot, 2016). In addition, workers who do not believe in their abilities to learn AI might struggle to meet the challenges of AI implementation. As a result, when a company implements AI systems, these workers are likely to feel stressed because they are not prepared to deal with the new responsibilities and obstacles that will inevitably arise.

This is supported by the TMSC’s suggestion that employees who lack confidence in their ability to acquire AI are likely to view AI adoption as a stressful event that they cannot manage during the primary assessment process (Lazarus and Folkman, 1987). They may conclude that their coping resources, especially their AI learning capabilities, are inadequate to handle the situation’s demands during the secondary appraisal process. Individuals who lack confidence in their ability to learn AI face increased workplace stress, difficulties adjusting to their organizations’ AI adoption practices, and uncertainty regarding their future with the company (Bankins et al. 2024; Pereira et al. 2023).

For instance, employees who do not think they are good at learning how to use AI could feel intimidated by the new CRM tools that an organization installs. They may be resistant to training sessions, hesitant to use new customer relationship management systems, and worried about losing their jobs to AI (Meyer and Hünefeld, 2018). As a result, these workers will feel increased stress on the job as they worry about failing to achieve the new standards set by the AI-powered CRM system and their inability to adjust to it.

The following hypothesis is based on the above theoretical considerations:

Hypothesis 5: Self-efficacy in AI learning will moderate the link between AI adoption in an organization and job stress such that the positive association between AI adoption and job stress will be weaker for employees with high self-efficacy in AI learning compared to those with low self-efficacy.

Figure 1 illustrates the conceptual model utilized in this study, which depicts the hypothesized relationships between AI adoption, job stress, and burnout, as well as the moderating role of self-efficacy in AI learning.