The rapid advancement of artificial intelligence (AI) has fundamentally reshaped the workplace, sparking growing scholarly interest in its impact on employee behavior (Bankins et al., 2024a; Gupta et al., 2024; Kanbach et al., 2024; Wael Al-khatib, 2023; Singh et al., 2024). Extant research has primarily examined the organizational implementation of AI technologies—such as algorithms, robotic services, and intelligent systems—and their effects on various employee outcomes, including proactive service and withdrawal behaviors (Liu et al., 2024), individual productivity (Tan et al., 2024), workplace thriving (Leong et al., 2025), AI engagement and helping (Li et al., 2024), innovative behavior (Dong et al., 2025), and proactive career development (Lin et al., 2024). However, this line of inquiry is largely rooted in a recipient-oriented perspective, which conceptualizes AI adoption as an organizationally driven process and portrays employees as passive responders to managerial decisions and system changes (Cheng et al., 2023; He et al., 2024). This view underplays a rapidly emerging trend: employees increasingly act as proactive agents, independently integrating generative AI (GenAI) tools into their work without formal mandates or oversight (Korzynski et al., 2024). As GenAI becomes more accessible and intuitive, employees are using it autonomously to streamline tasks, enhance creativity, and reconfigure their work experiences (Tomas et al., 2023). This bottom-up pattern of AI adoption represents a significant departure from traditional top-down implementations—yet surprisingly little is known about how self-initiated GenAI use shapes employees’ attitudes and behaviors at work.

To better understand the behavioral implications of self-initiated GenAI adoption, we draw on the concept of job crafting—employees’ proactive efforts to reshape their jobs to better fit their needs, values, and aspirations (Wrzesniewski and Dutton, 2001; Harju et al., 2016). As a bottom-up job redesign, job crafting offers a theoretically robust lens to examine how employees actively engage with GenAI to reconfigure their roles and work experiences (Petrous et al., 2015, 2018; Xiao et al., 2025; Shi et al., 2025). Existing research has shown that organizationally driven AI implementation can trigger job crafting. For instance, studies based on transactional stress theory (Cheng et al., 2023; Zhao et al., 2025) and the challenge–hindrance framework (Tan et al., 2024) suggest that employees engage in different forms of crafting—promotion- or prevention-focused—depending on how they appraise AI-driven changes. Similarly, symbolic leadership cues (He et al., 2023) and top-down job redesigns (Li et al., 2024) have been found to induce adaptive crafting behaviors in response to mandated AI systems. However, this reactive orientation understates the increasingly proactive nature of employee behavior in the era of GenAI. Drawing on the proactive behavior literature (Parker et al., 2010; Grant and Ashford, 2008), employees who take initiative—rather than merely react—are more likely to generate positive outcomes. Yet, we still know little about how employee voluntary GenAI adoption enables such proactive job crafting, and how these crafting behaviors contribute to longer-term attitudinal and behavioral outcomes. Addressing this gap is essential to understanding the broader consequences of GenAI in the workplace.

Among attitudinal outcomes, we focus on career commitment, which is defined as “one’s attitude towards one’s profession or vocation” (Blau, 1985) and reflects the emotional and cognitive investment an individual makes in their career path (Zhu et al., 2020). We consider career commitment a key construct through which job crafting may translate into longer-term engagement. As GenAI reshapes the contours of knowledge work and expands the scope of individual contribution, it can prompt employees to reconsider the value, meaning, and trajectory of their careers (Brougham and Haar, 2018; Brynjolfsson and Mcafee, 2017; Jarrahi, 2018; Manyika et al., 2017; Wach et al., 2023). Prior research suggests that job crafting fosters career commitment by allowing employees to regain autonomy, accumulate resources, and enhance the meaningfulness of their work (Dubbelt et al., 2019; Hakanen et al., 2018; Harju et al., 2016; Wang et al., 2024a). Yet, it remains unclear whether AI-enabled job crafting produces the same career-sustaining benefits as more traditional forms of proactive work design.

Beyond attitudinal outcomes, we also examine how GenAI adoption may influence employee behaviors through the sequential mechanisms of job crafting and career commitment. Specifically, we investigate its effects on voice quality, cyberloafing, and cheating behavior—three behaviors that reflect employees’ constructive, avoidant, and unethical responses to their work contexts, respectively. Although career commitment has been linked to ethical conduct and prosocial behavior (Fu, 2014; Mrayyan and Al-Faouri, 2008), research has yet to explore how GenAI adoption may indirectly shape these behaviors through employees’ crafting efforts and evolving career orientations. By modeling this dual-stage pathway, we aim to provide a more nuanced understanding of how self-initiated GenAI adoption unfolds into meaningful workplace consequences.

Finally, we introduce liking of AI—employees’ positive affective orientation toward AI technologies (Rubin, 1970; Pan et al., 2024)—as a psychological boundary condition that shapes the extent to which job crafting fosters career commitment. Drawing on the resource spiral principle of Conservation of Resources (COR) theory (Hobfoll, 2001), we argue that while job crafting represents a proactive effort to generate new resources, its motivational returns are not uniform. Rather, the degree to which crafting efforts translate into deeper career commitment depends on how employees appraise these experiences. Specifically, employees who exhibit a strong liking of AI—those who find value, enjoyment, and utility in engaging with the technology—are more likely to perceive GenAI-enabled crafting as intrinsically meaningful and consistent with their long-term career goals. In contrast, employees with low liking of AI may view the same crafting efforts as burdensome or externally imposed, thereby dampening the perceived resource gains and undermining the motivational force required to reinforce career commitment. By examining the liking of AI as a moderator, we aim to highlight the affective–motivational context that conditions the outcomes of job crafting in technology-driven environments.

Taken together, this study addresses the overarching question: How does GenAI adoption shape employees’ workplace behaviors through job crafting and career commitment? Our research offers several key contributions. First, we advance the GenAI literature by introducing job crafting as a proactive mechanism through which employees actively translate GenAI tools into meaningful work experiences. While job crafting has traditionally been studied in response to job demands or leadership behaviors (Wrzesniewski and Dutton, 2001; Tims et al., 2012; Bakker et al., 2020), we extend its application to the context of GenAI, illustrating that employees are not merely passive recipients of technological change but rather agentic users who reshape their jobs and outcomes through self-initiated digital engagement. Second, we contribute to the emerging conversation on AI and career development by theorizing and testing the link between GenAI adoption and career commitment—a dimension of attitudinal investment often assumed to be undermined by transformative technologies such as AI (Amankwah-Amoah et al., 2024; Ooi et al., 2025; Wach et al., 2023). Drawing on COR theory, we show that job crafting functions as a resource investment process that enables employees to derive motivational value from GenAI and sustain long-term career engagement—an underexplored pathway in prior AI research. Third, we develop and empirically test a dual-pathway model that connects GenAI adoption to both desirable (e.g., voice) and undesirable (e.g., cyberloafing, cheating) employee behaviors through job crafting and career commitment. Although prior studies have linked career commitment to positive work behaviors (Fu, 2014; Mrayyan and Al-Faouri, 2008), little is known about how GenAI may activate or mitigate such outcomes through psychological mechanisms. While much of the existing literature remains conceptual, our study provides empirical evidence that GenAI—when appraised and used proactively—can function as a resource that promotes adaptive behaviors and reduces counterproductive ones. The conceptual model is summarized in Fig. 1.

Fig. 1
figure 1

The research model.

Full size image

Literature review and hypothesis development

Conservation of resources theory

This study draws on COR theory (Hobfoll, 1989; Hobfoll et al., 2018) to explain how GenAI adoption influences employee behavior through job crafting and career commitment. COR theory posits that individuals strive to acquire, retain, and protect resources they value, and that those with greater access to such resources are more likely to initiate a gain spiral—an upward trajectory in which accumulated resources promote further investment and generate positive outcomes (Halbesleben et al., 2014).

We extend this logic by conceptualizing GenAI as a valued resource that employees can proactively leverage to reduce repetitive work, enhance efficiency, and stimulate creativity. These benefits represent both instrumental resources (e.g., time, information) and psychological resources (e.g., autonomy, efficacy), which in turn can activate job crafting behaviors—including seeking resources, embracing challenges, and optimizing demands (Korzynski et al., 2024). In the workplace, job crafting is a key strategy through which individuals reinvest their resources to generate further resource gains, especially under conditions of uncertainty and change (Meijerink et al., 2020; Petrou et al., 2015, 2018).

Through job crafting, employees reshape tasks, relationships, and perceptions to enhance control, meaning, and alignment with their personal values and aspirations (Demerouti et al., 2021; Buonocore et al., 2023). As a result, they experience greater agency and coherence between their work and long-term vocational goals, leading to stronger career commitment—defined as the psychological investment in one’s profession (Dubbelt et al., 2019; Hakanen et al., 2018; Harju et al., 2016). Thus, we propose that job crafting and career commitment function as sequential mediators linking GenAI adoption to employee outcomes.

Furthermore, COR theory emphasizes that resource accumulation is not purely behavioral, but also perceptually shaped by how individuals interpret the value of their actions (Hobfoll et al., 2018). In this light, we introduce liking of AI as a psychological resource that moderates the job crafting–career commitment link. Employees with a higher liking of AI are more likely to view AI-enabled crafting as personally meaningful and growth-congruent, thereby experiencing stronger motivational returns and increased career investment. Conversely, employees with low liking of AI may perceive crafting efforts as imposed or draining, attenuating the gain spiral and weakening the link to career commitment (Niu, 2010; Singhal and Rastogi, 2018; Zhu et al., 2021).

In sum, COR theory provides a coherent framework for understanding how GenAI adoption functions as a resource that triggers proactive job redesign (i.e., job crafting), strengthens long-term attitudinal engagement (i.e., career commitment), and shapes downstream behavioral outcomes—conditioned by employees’ affective orientation toward AI.

GenAI adoption and job crafting

In an era where adaptability to external environmental changes is paramount, organizations increasingly expect employees to proactively reshape their roles to unleash their full potential (Wrzesniewski and Dutton, 2001). GenAI, with its ability to enhance work efficiency and transform the scope and depth of employee roles, presents a unique opportunity for such proactive adjustments. However, research exploring how GenAI influences employees’ roles and behaviors remains limited.

Job crafting occurs when employees proactively adjust their job demands and resources to better align with their capabilities, needs, and aspirations (Petrou et al., 2012). It involves self-initiated behaviors such as seeking resources, seeking challenges, and optimizing demands (Demerouti and Peeters, 2018; Petrou et al., 2012). Seeking resources involves mobilizing additional job resources to manage demands, achieve goals, and enhance personal effectiveness (Tims and Bakker, 2010). Examples include soliciting advice from colleagues or supervisors (Carver et al., 1989), seeking feedback on performance (Ashford et al., 2003), or pursuing learning opportunities (Demerouti, 2014). Seeking challenges refers to actively taking on new tasks or responsibilities, such as participating in new projects, experimenting with innovative tools, or volunteering for extra assignments (Harju et al., 2016). This behavior is particularly prominent in dynamic work environments, where employees strive to master their roles and derive motivation from embracing challenges (Karasek and Theorell, 1990). Finally, optimizing demands focuses on streamlining tasks and improving efficiency, embodying the principle of “working smarter, not harder” (Demerouti and Peeters, 2018). This strategy enables employees to identify more efficient ways of completing tasks, conserving energy and time while achieving personal and organizational goals (Freund and Baltes, 2002; Bandura, 2023). Similar to workarounds, optimizing demands often results in secondary benefits, such as increased time for other activities (Demerouti and Peeters, 2018).

From the perspective of COR theory, employees are motivated to accumulate and safeguard valuable resources to cope with potential losses and remain competitive (Hobfoll, 2001). With the introduction of GenAI, employees are prompted to engage in job crafting to stay relevant and avoid obsolescence (Felten et al., 2021). As employees incorporate AI into their roles, they may seek resources by acquiring new skills and knowledge, ensuring they can use AI tools effectively (Kraus et al., 2022; Noy and Zhang, 2023; Santana and Díaz-Fernández, 2023). This could include seeking advice, training, or feedback to enhance their competencies (Carver et al., 1989; Ashford et al., 2003; Demerouti, 2014). Additionally, employees may seek challenges by taking on AI-related projects or new responsibilities that demonstrate their adaptability and proactivity—qualities that are especially valued in today’s rapidly evolving work environments (Budhwar et al., 2023; Harju et al., 2016). Embracing these challenges helps employees foster mastery, encourage innovation, and improve workflows (Gilardi et al., 2023; Haefner et al., 2021). Finally, by optimizing demands, aligning with the “working smarter, not harder” principle (Demerouti and Peeters, 2018), employees can restructure their tasks and improve efficiency using GenAI, enabling them to achieve personal goals while conserving energy and resources (Freund and Baltes, 2002; Bandura, 2023).

Through these three dimensions—seeking resources, seeking challenges, and optimizing demands—employees can proactively align their roles with the evolving technological environment. This alignment not only facilitates personal development but also enhances organizational performance. Supporting this view, Li et al. (2024) demonstrated that GenAI enhances job crafting by enabling employees to analyze their work data, identify areas for improvement, and adjust their roles to align with their strengths and preferences. Similarly, Hessari et al. (2024) found that the use of GenAI tools significantly boosts employee adaptability. Thus, we propose the following hypothesis:

H1. GenAI adoption is positively related to job crafting: (a) seeking resources, (b) seeking challenges, and (c) optimizing demands.

Job crafting as a mediator between GenAI adoption and career commitment

Career commitment refers to an employee’s dedication to their career path and their intent to maintain long-term engagement within their profession (Blau, 1985). In the face of rapid technological advancements, career commitment becomes especially critical as employees must continuously adapt to evolving job roles (Zhu et al., 2020). Previous research highlights that job-related resources, such as autonomy and supervisor support, contribute to career satisfaction and success, while high job demands—such as excessive workloads or job insecurity—can elevate stress levels and increase turnover intentions (Ekmekcioglu et al., 2020; Salmela-Aro and Upadyaya, 2018; Van Steenbergen et al., 2018). As GenAI reshapes workplace dynamics, employees’ ability to adapt to these changes is key to sustaining their career commitment in this rapidly transforming environment.

Drawing on COR theory’s gain spiral logic (Hobfoll et al., 2018), GenAI provides employees with new external resources—such as AI tools, data, and automated workflows—that they can leverage to align their work with shifting organizational priorities (Aguinis et al., 2024). When proactively invested through job crafting, these resources accumulate into deeper psychological engagement—namely, career commitment (Goulet and Singh, 2002)—which in turn supports desirable behavioral outcomes. For example, through job crafting, employees can adapt by reshaping their tasks to fit their career aspirations, thereby boosting satisfaction, motivation, and engagement (Chen et al., 2014; Tims et al., 2016). This upward spiral highlights how resource-rich individuals are more likely to engage in resource-enhancing behaviors, leading to sustained professional adaptation (Kundi et al., 2024; Halbesleben et al., 2014). Within the context of this study, we propose that job crafting mediates the relationship between GenAI adoption and career commitment by enabling employees to (1) seek new resources that support their tasks, (2) pursue challenges that promote both personal and professional growth, and (3) optimize demands to enhance work efficiency. These proactive behaviors allow employees to align their work with their career goals, thereby strengthening their sense of ownership over their career development and reinforcing their commitment to their profession (Erdogan and Bauer, 2005; Berg et al., 2008).

Furthermore, in fast-evolving work environments, job crafting functions as an effective coping mechanism that helps employees manage the stress and uncertainty associated with AI-related changes (Solberg and Wong, 2016). By proactively adjusting their roles to accommodate the demands of new technologies, employees can reduce work-related anxiety and enhance their sense of control (Van den Heuvel et al., 2015). This reduction in anxiety has been shown to foster stronger career commitment (Niu, 2010). As a result, when employees engage in job crafting effectively, they are more likely to perceive the adoption of GenAI as an opportunity for career advancement rather than a threat to their professional standing.

In summary, GenAI adoption provides employees with opportunities to build and safeguard their professional resources through job crafting. By actively seeking resources, pursuing challenges, and optimizing demands, employees improve their work experiences and align their roles with both personal and career goals, ultimately enhancing their career commitment. Based on this, we propose the following hypothesis:

H2. GenAI adoption has indirect effects on career commitment via job crafting, specifically in the areas of (a) seeking resources, (b) seeking challenges, and (c) optimizing demands.

The mediating role of job crafting and career commitment

Although much research has focused on how AI can enhance organizational performance, there is limited exploration of the relationship between GenAI adoption and employee organizational behavior. Positive organizational behaviors are essential for driving individual performance, engagement, and well-being. We argue that GenAI adoption triggers a sequence of behaviors beginning with job crafting, which in turn strengthens career commitment, ultimately influencing employee outcomes such as voice quality, cyberloafing, and cheating behavior.

Career commitment plays a vital role in shaping employee behaviors and attitudes at work (Zhu et al., 2021). Employees with high career commitment are more engaged in their work, demonstrating positive outcomes such as higher job involvement (Womack et al., 2016), willingness to make personal sacrifices for their careers (Fu, 2014), and greater participation in organizational citizenship behaviors (Afsar et al., 2018). Career commitment is also linked to better productivity (Ahmed, 2019), work-related well-being (Duffy et al., 2011), sportsmanship, and altruism (Carson and Carson, 1998). Furthermore, committed employees are more likely to deliver superior service quality, achieve higher customer satisfaction, and align their efforts with business goals (Karatepe et al., 2024). They experience higher organizational commitment, job satisfaction (Kim et al., 2021; Zhang et al., 2014), and lower withdrawal intentions (Kim et al., 2015; Weng and McElroy, 2012). Employees committed to their careers actively enhance their employability to thrive in dynamic job markets (Van der Heijden et al., 2022) and tend to develop favorable work outcomes compared to employees who are solely committed to their organizations (Somers and Birnbaum, 2000).

Career-committed individuals exhibit professionalism and adaptability—qualities essential in today’s AI-driven work environments. They proactively seek opportunities to stay aligned with industry trends and maintain cutting-edge competencies (De Vos et al., 2020). These individuals are also more likely to contribute to their workplaces beyond formal expectations (Carson and Carson, 1998). As a result, employees with high career commitment are more likely to exhibit high-quality voice, along with lower levels of cyberloafing and cheating behaviors.

Voice quality, defined as the value of employee input characterized by rationale, feasibility, organizational focus, and novelty, reflects the impact of discretionary communication intended to improve organizational functioning (Brykman and Raver, 2021; Morrison, 2011). Research suggests that voice quality, rather than frequency, leads to more positive evaluations, influencing employee performance and career progression (Burris, 2012; Brykman and Raver, 2021; Xu et al., 2020). Career-committed employees, motivated by a desire for professional development, are more likely to engage in high-quality voice to secure resources and enhance career opportunities (Carson and Carson, 1998).

Cyberloafing, defined as the use of employer-provided internet access during work hours for non-job-related purposes, is a common challenge in many organizations (Lim, 2002). As a form of deviant behavior, cyberloafing can undermine productivity and organizational effectiveness (Chen et al., 2022; Lim and Teo, 2022; Ohana et al., 2023; She et al., 2022). Factors such as job involvement, managerial support, and workplace stressors can predict cyberloafing (Liberman et al., 2011; Zhou et al., 2021). From a resource perspective, depleted psychological resources often drive employees to engage in cyberloafing to cope with stress (Agarwal and Avey, 2020; Henle, 2023). However, career commitment acts as a psychological resource that can reduce the likelihood of cyberloafing by fostering motivation and focus on productive tasks (Elrehail et al., 2021; Mercado et al., 2017). Career-committed employees view work as a means of achieving professional success, which decreases the temptation to engage in cyberloafing during work hours.

Cheating behavior in the workplace refers to unethical actions intended to create unfair advantages, such as misrepresenting productivity or lying about absences (Mitchell et al., 2018; Kundro et al., 2023; Spoelma, 2022). These behaviors are often linked to job dissatisfaction and disengagement (Fu, 2014). However, research suggests that career-committed employees are less likely to engage in unethical practices because they derive satisfaction from meaningful work (Zhang, 2020). Such employees are motivated to align their behavior with organizational goals, viewing their work as a pathway to career success (Zhang, 2020). GenAI adoption, by fostering job crafting behaviors, enhances career commitment and promotes ethical workplace conduct.

Thus, a sequential mediation model suggests that GenAI adoption indirectly influences employee outcomes—such as voice quality, cyberloafing, and cheating behavior—through job crafting and career commitment. Specifically, employees who engage in job crafting behaviors (e.g., seeking resources, seeking challenges, and optimizing demands) are more likely to strengthen their career commitment, which in turn leads to positive behavioral outcomes. As employees align their work with AI-driven opportunities, they are more likely to engage in high-quality voice, reduce cyberloafing, and refrain from unethical behavior. Based on this, we propose the following hypothesis:

H3. The indirect effects of GenAI adoption on employee outcomes (i.e., voice quality, cyberloafing, and cheating behavior) are mediated sequentially by job crafting behavior—(a) seeking resources, (b) seeking challenges, or (c) optimizing demands—and career commitment.

The moderating role of liking of AI

While COR theory originally emphasized objective resource acquisition and protection (Hobfoll, 1989, 2001), later developments highlight the role of subjective resource appraisal in shaping individuals’ stress responses and motivational patterns (Hobfoll et al., 2018). According to COR theory, individuals are motivated not only to acquire and protect valued resources but also to interpret and evaluate the utility of their investments (Hobfoll et al., 2018). Psychological resources—such as positive attitudes—can play a critical role in shaping how individuals perceive and derive value from their proactive efforts. In this study, we conceptualize liking of AI as a psychological resource reflecting employees’ positive affective orientation toward AI technologies (Rubin, 1970; Pan et al., 2024).

Employees with a high liking of AI are more inclined to view AI as a supportive and enabling tool (Pan et al., 2024), which facilitates the effective enactment and meaningful interpretation of their job crafting behaviors. For instance, individuals who appreciate AI are more likely to perceive resource-seeking behaviors as productive and growth-aligned, reinforcing their emotional investment in long-term career goals. Similarly, those with favorable attitudes toward AI are more open to AI-related challenges (Dong et al., 2024), interpreting them as opportunities for professional development rather than burdens (Chiu et al., 2021). In the case of demand optimization, liking of AI promotes cognitive reframing—employees are more likely to perceive AI-induced changes as manageable or even energizing (Albarracin and Shavitt, 2018; Chiu et al., 2021), thus reducing strain and enhancing perceived career alignment (Kundi et al., 2024).

Through this lens, liking of AI functions as a boundary condition that shapes the extent to which job crafting behaviors foster career commitment. When individuals hold a positive attitude toward AI, they are more capable of harnessing AI as a resource, thereby amplifying the motivational impact of their crafting efforts. Conversely, those less favorable toward AI may struggle to perceive these same behaviors as valuable or rewarding, weakening the positive effect on career commitment. Therefore, we propose the following hypothesis:

H4. Liking of AI moderates the relationship between job crafting and career commitment such that the positive effects of (a) seeking resources, (b) seeking challenges, and (c) optimizing demands on career commitment are stronger when individuals have higher liking of AI (versus lower liking of AI).

Building on the preceding hypotheses, we further predict that the indirect effect of GenAI adoption on career commitment through job crafting will vary depending on employees’ liking of AI. Specifically, GenAI adoption creates opportunities for employees to acquire new resources, such as skills and knowledge related to AI technologies (Budhwar et al., 2023). Employees with a high liking of AI are more inclined to explore and leverage these resources to their fullest potential. Their positive attitude towards AI fosters proactive engagement with the technology, strengthening the relationship between resource acquisition and career commitment.

Additionally, GenAI introduces new challenges within work roles, such as adapting to novel workflows or developing innovative solutions (Sedkaoui and Benaichouba, 2024). Employees with greater appreciation for AI view these challenges not as obstacles but as opportunities for personal and professional growth. Their enthusiasm for AI encourages them to embrace these challenges, reinforcing their career commitment as they see themselves progressing within an AI-enhanced work environment.

GenAI also changes the nature of tasks, potentially rendering some tasks obsolete while creating new demands. Optimizing demands involves adjusting job requirements to better align with employees’ abilities, reducing stressors, and enhancing work-life balance (Zhang and Parker, 2022). Employees with a positive attitude towards AI are better equipped to adapt their roles to incorporate AI efficiently. This positive adaptation promotes alignment between job demands and personal skills, resulting in a more fulfilling work experience and stronger career commitment.

In sum, employees with a high liking of AI are more likely to engage proactively in job crafting behaviors such as seeking resources, embracing challenges, and optimizing demands. This proactive engagement amplifies the positive impact of AI adoption on career commitment. Conversely, employees with a low liking of AI may not fully utilize the opportunities AI offers, leading to a weaker relationship between job crafting behaviors and career commitment. Therefore:

H5. Liking of AI moderates the indirect effects of GenAI adoption on career commitment via (a) seeking resources, (b) seeking challenges, and (c) optimizing demands, such that the effects become stronger when individuals have higher liking of AI (versus lower liking of AI).

Based on the hypotheses outlined above, we propose that variations in employees’ liking of AI could influence the pathway from GenAI adoption, through job crafting and career commitment, to employee outcomes.

H6. Liking of AI moderates the sequentially indirect effects of GenAI adoption on employee outcomes via (a) seeking resources, (b) seeking challenges, or (c) optimizing demands and career commitment, such that the effects become stronger when individuals have higher liking of AI (versus lower liking of AI).

Methods

Research design and sampling technique

Because this study aimed to examine the outcomes of employees’ GenAI adoption, the target population comprised full-time employees who had experience using GenAI tools, such as ChatGPT or Ernie Bot. We recruited participants from a range of industries in China via Credamo (https://www.credamo.cc/#/), a widely used online survey platform similar to MTurk. Credamo has been validated as a reliable source of high-quality data for social science research in China (Del Ponte et al., 2024) and has been employed in numerous peer-reviewed studies published in leading journals (e.g., Yang and Zhang, 2025; Zhou et al., 2025; Zhu et al., 2024).

Participant recruitment and screening

We first published a recruitment notice on the survey platform, targeting full-time employees who were currently using GenAI at work and inviting them to voluntarily participate in the study. This initial screening phase consisted of two questions. The first was a single-choice item: “Do you use GenAI (e.g., ChatGPT, Ernie Bot, MarsCode) for your work?” Participants who answered “No” were excluded from the study, and their participation was automatically terminated. The second was an open-ended prompt: “Please describe a scenario in which you use GenAI in your work.” Responses to this item were manually reviewed by the research team to ensure that participants met the inclusion criteria before proceeding to the main survey. Qualified participants were then asked to forward the survey link to a coworker in order to establish a dyadic pairing. Pairing validity was verified through two mechanisms: (1) Forward-forwarding: participants were required to forward the registration link directly to a colleague to ensure the relationship between the pair; (2) Two-way verification: both participants were required to input the last four digits of each other’s phone numbers to confirm mutual identity and prevent random matching. Only successfully matched dyads were included in the final sample. In total, 400 employee–coworker dyads were established during this stage.

First-stage survey

Credamo sent the formal questionnaire to the 400 matched employee-colleague respondents. Employees provided personal information, covering age, gender, educational level, organizational tenure, and dyadic tenure, and reported their GenAI adoption and liking of AI, while their colleagues completed the seeking resources, seeking challenges and optimizing demands for employees. A total of 391 pairs completed the survey. Each participant received a 5 RMB reward upon completing the questionnaire. After data cleaning, we excluded careless responses, questionnaires that failed the attention check questions, and those with unusually short or long response times, resulting in 364 valid questionnaires.

Second-stage survey

One month later, we sent the survey to the same 364 pairs and asked employees to complete their career commitment, while their colleagues completed the voice quality, cyberloafing, workplace cheating behaviors for employees. We received 318 pairs of responses in the second phase. Again, each participant received a 5 RMB reward upon completion. After the same data cleaning process, we excluded careless responses, those that failed the attention checks, and those with unusual response times, resulting in 291 valid questionnaires, yielding a response rate of 72.75%.

Data sample

The sample size of 291 employees meets the minimum participant-to-item ratio of 2:1 recommended by Hinton et al. (2014), providing a solid foundation for subsequent analysis. Of the participants, 161 were male (55.3%) and 130 were female (44.7%). Gender was coded as “1” for males and “2” for females. Participants reported their age, organizational tenure, and dyadic tenure in terms of both years and months; the months were converted into decimal fractions of a year for analysis. The average age of participants was 35.31 years, with ages ranging from 24 to 55. A large majority (83.8%) were under the age of 40. Education level was measured as the number of years of formal education completed, ranging from 15 to 19 years. The most frequently reported value was 18 years (25.8%), while the least frequent was 15 years (18.9%). Organizational tenure ranged from 0 to 22 years. Approximately 55% of participants had 1–5 years of tenure, 31% had 5–10 years, and around 14% had more than 10 years. Dyadic tenure (i.e., time spent working with the paired coworker) ranged from 0 to approximately 9 years, with 86.3% of dyads having less than 5 years of shared experience and 13.7% having more than 5 years.

Measures

We utilized Brislin’s (1980) back-translation technique to ensure the fidelity of our survey items translated from English to Chinese. Before deploying the main survey, a pilot test was conducted with 30 participants to refine the wording, question order, and response options. We followed Singh et al. (2024) in using a 5-point Likert scale for GenAI adoption to ensure consistency with the original measurement framework. Given that GenAI adoption is the core construct of our study, we retained this format to preserve its validity and comparability. All other variables—including job crafting (seeking resources, seeking challenges, optimizing demands), liking of AI, career commitment, voice quality, cyberloafing, and cheating behavior—were measured using 7-point Likert scales (1 = strongly disagree, 7 = strongly agree). The use of 7-point scales provides finer granularity, which is particularly valuable for constructs involving nuanced psychological or motivational aspects (Preston and Colman, 2000). The measurement items can be found in Appendix 1.

GenAI adoption was measured using a five-item scale developed by Singh et al. (2024). A sample item is, “Generative AI makes it easier for me to perform my organizational responsibilities.” The scale demonstrated adequate reliability, with Cronbach’s alpha of 0.92.

Job crafting was measured following Demerouti’s (2023) scale, covering three dimensions: seeking resources (Cronbach’s alpha = 0.91, 6 items), seeking challenges (Cronbach’s alpha = 0.84, 3 items), and optimizing demands (Cronbach’s alpha = 0.88, 5 items). Example items include: “This colleague tries to learn new things at work,” “This colleague asks for more tasks when he/she finishes his/her work,” and “This colleague simplifies work processes to make his/her job easier.” Colleagues rated the frequency of these behaviors over the past three months on a seven-point scale from 1 (“never”) to 7 (“often”). In this study, we adopt the three-dimensional framework of job crafting—seeking resources, seeking challenges, and optimizing/reducing demands—which has been conceptually overlapping in both Tims et al. (2012) and Petrou et al. (2012). Although these two frameworks differ slightly in terminology and focus, they converge theoretically on these three core dimensions. Several scholars have acknowledged this conceptual overlap, noting that the three-factor model offers a parsimonious yet comprehensive structure that is particularly well suited for studies of motivational mechanisms (e.g., Bakker et al., 2020; Demerouti et al., 2021; Lu et al., 2025). Consistent with recent practice (Demerouti, 2023), seeking resources and seeking challenges were measured using items from the Job Crafting Scale developed by Petrou et al. (2012). Optimizing demands—which captures employees’ proactive efforts to make their work more efficient—was measured using the scale by Demerouti and Peeters (2018). Given that employees’ adoption of GenAI in our study is primarily driven by the desire to enhance efficiency and reconfigure work tasks, optimizing demands is particularly relevant to our research context. Consequently, we adopt this measurement approach in our study. Furthermore, Costantini et al. (2021) provide empirical support for the construct validity of this job crafting scale, reinforcing the robustness of our measurement choices.

Liking of AI was assessed using a revised four-item scale based on Wayne and Ferris (1990). A sample item is, “I think Generative AI makes my work more enjoyable,” with Cronbach’s alpha of 0.92.

Career commitment was measured using a seven-item scale by Qu et al. (2023). A sample item is, “If I could do it all over again, I would not choose to work in my present industry,” with the scale showing high reliability (Cronbach’s alpha = 0.85).

Voice quality was measured using a four-item scale from Brykman and Raver (2021). A sample item is, “This colleague suggests ideas that fit with the organization’s core values,” with Cronbach’s alpha of 0.92.

Employee cyberloafing was measured through a nine-item scale by Lim and Chen (2012), with responses ranging from 1 (“never”) to 7 (“constantly”). A sample item is, “This colleague visits non-work-related websites,” with Cronbach’s alpha of 0.94.

Workplace cheating behavior was evaluated using a seven-item scale developed by Mitchell et al. (2018). A sample item is, “This colleague misrepresents work activity to make it look as though they have been productive,” with Cronbach’s alpha of 0.93.

To mitigate potential confounding effects and enhance the robustness of our findings, this study incorporated gender, age, education level, organizational tenure and dyadic tenure as control variables. These factors were selected based on their established relevance in previous research (Cao et al., 2023; Chaudhary, 2020; Ozturk et al., 2021), ensuring a thorough analysis of the primary variables of interest. Since the control variables, with the exception of gender, tenure and dyadic tenure, were not significantly related to the dependent variables, we excluded them from the analysis, following the best practices for control variable selection recommended by Bernerth and Aguinis (2016). However, the results remain consistent when these variables are included.

Results

Table 1 presents the descriptive statistics, scale reliabilities, and bivariate correlations. To address potential common method variance (CMV), we employed several procedural and statistical measures, following the recommendations of Podsakoff et al. (2003). Procedurally, we informed participants of the study’s purpose, provided detailed information about the confidentiality of their responses, and scrambled the order of items in the questionnaire. Statistically, we conducted Harman’s one-factor test (Podsakoff et al., 2003) to assess the potential for CMV. The results showed that the first factor accounted for only 23.729% of the variance, well below the 50% threshold, indicating that no single factor accounted for the majority of the variance. Therefore, CMV is unlikely to be a concern in this study. Additionally, as shown in Table 2, a comparison of the nine-factor (proposed) model with the single-factor model demonstrated that the single-factor model provided a poor fit, further supporting the conclusion that CMV is not an issue.

Table 1 Descriptive statistics and bivariate correlations.
Full size table
Table 2 Confirmatory factor analysis results.
Full size table

We verified the factor structure of our items using confirmatory factor analysis (CFA) following the procedure outlined by Anderson and Gerbing (1988). The CFA results indicated that each observable indicator loaded significantly (p < 0.01) on its intended factor, supporting the convergent validity of the scale items. The maximum likelihood estimation procedure was employed to assess model fit. We compared the hypothesized nine-factor measurement model (baseline model) with eight alternative models. The statistical indices for the baseline model indicated an adequate fit (χ²[1139] = 1301.212, χ²/df = 1.142, CFI = 0.983, TLI = 0.981, RMSEA = 0.021, SRMR = 0.037). As shown in Table 2, the baseline model demonstrated superior fit compared to the eight alternative models, indicating that the variables were adequately distinct.

Mediation analysis

To test Hypotheses 1, 2, and 3, we specified a mediation model using AMOS 26.0 software, employing 1000 bootstrap samples to ensure statistical robustness. Specifically, we modeled the effect of GenAI adoption on the three dimensions of job crafting (i.e., seeking resources, seeking challenges, and optimizing demands), career commitment, and employee outcomes (i.e., voice quality, employee cyberloafing, and workplace cheating behaviors). Additionally, we examined the effects of the three dimensions of job crafting on career commitment and employee outcomes, as well as the effect of career commitment on employee outcomes. The path model demonstrated a good fit with the data (χ²[980] = 1207.202, p < 0.001, CFI = 0.973, TLI = 0.972, RMSEA = 0.028, SRMR = 0.0749). As shown in Table 3, the results indicated that GenAI adoption was positively related to seeking resources (γ = 0.213, 95% CI [0.134, 0.295]), seeking challenges (γ = 0.207, 95% CI [0.117, 0.297]), and optimizing demands (γ = 0.241, 95% CI [0.141, 0.344]). Therefore, Hypotheses 1a, 1b, and 1c were supported.

Table 3 Results for hypothesized direct and indirect effects.
Full size table

To assess the mediation model as outlined by Preacher et al. (2010), we utilized the PROCESS macro for SPSS 23.0 (Hayes and Scharkow, 2013) to obtain confidence intervals for the indirect effects through bootstrapping. We applied PROCESS Model 4 to individually examine the mediating effects of the three dimensions of job crafting—seeking resources, seeking challenges, and optimizing demands—on the relationship between GenAI adoption and career commitment. As reported in Table 3, using 1000 bootstrap samples and 95% confidence intervals, the results revealed significant indirect effects of GenAI adoption on career commitment through each job crafting dimension (seeking resources: effect = 0.049, SE = 0.014, 95% CI = [0.026, 0.079]; seeking challenges: effect = 0.042, SE = 0.013, 95% CI = [0.018, 0.070] and optimizing demands: effect = 0.033, SE = 0.013, 95% CI = [0.012, 0.063]). These findings provide strong support for Hypotheses 2a, 2b, and 2c.

Similarly, we employed PROCESS Model 6 to evaluate the sequential mediation model, addressing Hypotheses 3a, 3b, and 3c. As shown in Table 3, the results indicated that GenAI adoption significantly influenced employee outcomes—specifically, voice quality, cyberloafing, and workplace cheating behaviors—through the sequential mediation of the job crafting dimensions (i.e., seeking resources, seeking challenges, and optimizing demands), followed by career commitment. Notably, the 95% confidence intervals for these sequential mediation models did not include zero, providing support for Hypotheses 3a, 3b, and 3c.

Moderated mediation analysis

Building on the mediation model, we specified a moderated mediation model by incorporating interaction terms between the job crafting dimensions (i.e., seeking resources, seeking challenges, and optimizing demands) and liking of AI. The moderated mediation model demonstrated a good fit with the data (χ²[983.918] = 686, p < 0.001, CFI = 0.954, TLI = 0.950, RMSEA = 0.039, SRMR = 0.078). To test the moderating role of liking of AI in the relationship between the job crafting dimensions and career commitment (Hypotheses 4a, 4b, and 4c), we employed PROCESS Model 1 with 1000 bootstrap samples. As presented in Table 4, liking of AI significantly moderated the effects of seeking resources (effect = 0.142, p < 0.01, 95% CI = [0.358, 0.602]) and optimizing demands (effect = 0.080, p < 0.05, 95% CI = [0.018, 0.143]) on career commitment (see Figs. 2 and 3). However, the moderating effect of liking of AI on the relationship between seeking challenges and career commitment was not significant (effect = 0.046, p > 0.05, 95% CI = [−0.002, 0.094]). Thus, Hypothesis 4b was not supported.

Table 4 Results for moderating role of liking of AI (Process Model 1).
Full size table
Fig. 2
figure 2

Moderating effect of liking of AI on the relationship between seeking resources and career commitment.

Full size image
Fig. 3
figure 3

Moderating effect of liking of AI on the relationship between optimizing demands and career commitment.

Full size image

Simple slope tests further revealed that the positive relationship between seeking resources and career commitment was stronger when liking of AI was high (1 SD above the mean; effect = 0.480, p < 0.01) than when it was low (1 SD below the mean; effect = 0.108, p > 0.05). Similarly, the positive relationship between optimizing demands and career commitment was stronger at high levels of liking of AI (1 SD above the mean; effect = 0.252, p < 0.01) than at low levels (1 SD below the mean; effect = 0.042, p > 0.05). Therefore, Hypotheses 4a and 4c were supported.

We evaluated the moderated mediation hypotheses using PROCESS Model 14, specifically testing Hypotheses 5a, 5b, and 5c with 1000 bootstrap samples. As presented in Table 5, the results indicated that liking of AI moderated the indirect effects of GenAI adoption on career commitment through either seeking resources (effect = 0.027, SE = 0.008, 95% CI = [0.013, 0.044]) or optimizing demands (effect = 0.019, SE = 0.009, 95% CI = [0.004, 0.041]). However, the moderating effect of liking of AI on the indirect relationship between GenAI adoption and career commitment via seeking challenges was not statistically significant (effect = 0.009, SE = 0.008, 95% CI = [−0.005, 0.026]). Hypothesis 5b was thus supported. Specifically, the mediation effects of seeking resources were stronger when liking of AI was high (effect = 0.090, SE = 0.020, 95% CI = [0.054, 0.135]), compared to when it was low (effect = 0.018, SE = 0.012, 95% CI = [−0.003, 0.044]). The mediation effects of optimizing demands were stronger when liking of AI was high (effect = 0.048, SE = 0.017, 95% CI = [0.020, 0.087]), compared to when it was low (effect = 0.000, SE = 0.017, 95% CI = [−0.038, 0.030]). Therefore, these findings provide support for Hypotheses 5a and 5c.

Table 5 Results of moderated mediation model (Process Model 14).
Full size table

To evaluate the sequential moderated mediation hypotheses (i.e., Hypothesis 6), we employed PROCESS Model 91 with 1000 bootstrap samples. As detailed in Table 6, the results revealed that liking of AI moderated the sequential indirect effects of GenAI adoption on employee outcomes through two distinct pathways: first, via seeking resources followed by career commitment (voice quality: effect = 0.011, SE = 0.005, 95% CI = [0.004, 0.023]; cyberloafing: effect −0.007, SE = 0.003, 95% CI = [−0.014, −0.002]; workplace cheating behavior: effect = −0.014, SE = 0.005, 95% CI = [−0.027, −0.005]); and second, via optimizing demands followed by career commitment (voice quality: effect = 0.006, SE = 0.004, 95% CI = [0.001, 0.016]; cyberloafing: effect = −0.011, SE = 0.005, 95% CI = [−0.023, −0.002]; workplace cheating behavior: effect = −0.005, SE = 0.003, 95% CI = [−0.012, −0.001]). However, the moderating effects of liking of AI on the sequential indirect relationship between GenAI adoption and employee outcomes via seeking challenges followed by career commitment were not statistically significant (voice quality: effect = 0.004, SE = 0.004, 95% CI = [−0.002, 0.013]; cyberloafing: effect = −0.004, SE = 0.004, 95% CI = [−0.013, 0.002] and workplace cheating behavior: effect = −0.002, SE = 0.002, 95% CI = [−0.008, 0.001]). Hypothesis 6b was therefore not supported.

Table 6 Results of sequential moderated mediation model.
Full size table

Specifically, the indirect effects of GenAI adoption on employee outcomes were stronger through the pathway of seeking resources followed by career commitment when liking of AI was high (voice quality: effect = 0.037, SE = 0.014, 95% CI = [0.014, 0.068]; cyberloafing: effect = −0.024, SE = 0.009, 95% CI = [−0.045, −0.008]; workplace cheating behavior: effect = −0.047, SE = 0.014, 95% CI = [−0.078, −0.022]), compared to when it was low (voice quality: effect = 0.007, SE = 0.006, 95% CI = [−0.002, 0.021]; cyberloafing: effect = −0.005, SE = 0.004, 95% CI = [−0.016, 0.001]; workplace cheating behavior: effect = −0.009, SE = 0.006, 95% CI = [−0.023, 0.002]). Moreover, the indirect effect through optimizing demands followed by career commitment was significant only when liking of AI was high (voice quality: effect = 0.017, SE = 0.008, 95% CI = [0.004, 0.034]; cyberloafing: effect = −0.028, SE = 0.011, 95% CI = [−0.052, −0.010]; workplace cheating behavior: effect = −0.013, SE = 0.006, 95% CI = [−0.028, −0.004]), and not when it was low (voice quality: effect = 0.000, SE = 0.005, 95% CI = [−0.012, 0.010]; cyberloafing: effect = 0.000, SE = 0.009, 95% CI = [−0.018, 0.018]; workplace cheating behavior: effect = 0.000, SE = 0.004, 95% CI = [−0.009, 0.009]). These findings provide support for Hypotheses 6a and 6c.

Discussion

The primary goal of this study was to examine whether job crafting and career commitment sequentially mediate the relationship between employee GenAI adoption and employee outcomes, with a focus on the moderating role of liking of AI. This was motivated by the idea that employees should be seen as active players to understand how GenAI adoption relates to employees’ work-related outcomes. Although previous studies already showed the dynamic role of employees in engaging with technology to optimize work processes and advance their career trajectories (Bankins et al., 2024; Medici et al., 2023), our study extends this literature by illustrating how employees, as proactive job crafters, convert GenAI adoption into greater career commitment and, in turn, into distinct behavioral outcomes. We also explore how liking of AI shapes these processes, offering a more nuanced understanding of individual differences in technology acceptance. Below, we outline the theoretical and practical implications of these findings.

Theoretical implications

This study adds to the literature in several ways. First, our results have implications for GenAI research. Although showing the value of GenAI across various industries (Gupta et al., 2024; Shen et al., 2023; Singh et al., 2024; Wang et al., 2024b; Zhai et al., 2024), few studies have explored its impact on employee-centric outcomes. By identifying job crafting and career commitment as sequential mediators, we bridge this gap and show how AI adoption translates into both constructive (e.g., voice behavior) and dysfunctional (e.g., cyberloafing, cheating) outcomes. This implies that GenAI adoption relates to employee behavioral outcomes to the extent that it activates employees to proactively seek resources, seek challenges, optimize demands, and career commitment. This suggests that future research should focus on employees’ proactive involvement in resource accumulation to better understand how GenAI adoption relates top employee outcomes. Interestingly, existing studies predominantly adopt a top-down “recipient” perspective, portraying employees as passive adapters to AI-driven workflow changes (Cheng et al., 2023; He et al., 2024). Instead, our results imply that the emergence of GenAI empowers employees to shift from being passive recipients to proactive responders to new technology. GenAI research can benefit from applying COR theory as it proposes that GenAI represents a proactively integrated resource—adopted voluntarily by employees, which activates employees to acquire and reinvest additional resources (Hobfoll, 2001), notably job crafting and career commitment. This also reinforces the COR perspective that resource gain spirals depend not only on initial resource activation (e.g., AI-enabled job crafting), but also on the sustainability of reinvestment pathways embedded in employees’ roles and career trajectories.

Second, our results have implications for research into the understanding of the drivers of job crafting. Whereas existing studies predominantly view AI technologies such as robotics and algorithmic management as sources of uncertainty that drive top-down job design changes (Allal-Chérif et al., 2021; Kaushal et al., 2023; Pillai and Sivathanu, 2020; Prikshat et al., 2023; Votto et al., 2021), job crafting has often been conceptualized as a reactive adaptation to AI-induced work changes, portraying employees as passive recipients of organizationally implemented AI technologies (Cheng et al., 2023; Zhao et al., 2025). Our results show that GenAI provides employees with accessible tools to personalize and optimize their tasks, which enhances efficiency and allows them to focus on higher-order work that leverages their unique capabilities (Gupta et al., 2024; Pillai and Sivathanu, 2020). This shift positions GenAI as a catalyst for bottom-up job redesign (Grant and Parker, 2009; Oldham and Hackman, 2010; Tims et al., 2014), offering new theoretical insight into how emerging AI technologies can stimulate proactive changes in work roles and responsibilities.

Third, this study contributes to the career commitment literature by demonstrating how GenAI adoption can shape employees’ long-term attachment to their careers. While prior research has primarily examined dispositional and organizational antecedents of career commitment (Zhu et al., 2020), our findings reveal that employees’ proactive engagement with GenAI can also foster career commitment by enabling job crafting behaviors. This extends existing theory by showing that career commitment is not solely influenced by stable traits or contextual factors but can also emerge from employees’ self-initiated use of emerging technologies to enhance their work experience. Drawing on COR theory (Hobfoll, 2001), we show that GenAI serves as a resource that facilitates job crafting and, in turn, promotes career commitment. This motivational pathway leads to both desirable outcomes (e.g., enhanced voice quality) and the reduction of undesirable behaviors (e.g., cyberloafing and cheating). By examining both positive and negative behavioral consequences, our study also responds to recent calls for a more balanced understanding of AI’s impact on workplace behaviors (Dong et al., 2025; Park, 2024; Yin et al., 2024).

Lastly, this study contributes to a more nuanced understanding of individual differences in technology acceptance by examining the moderating role of liking of AI in the relationship between job crafting and career commitment. Our findings show that employees’ affective evaluations—specifically, their liking of AI—play a critical role in determining whether proactive behaviors such as job crafting translate into long-term career commitment. This aligns with prior research suggesting that personal meaning and positive emotional framing are key drivers of sustained career engagement (Zhu et al., 2021). More specifically, we find that liking of AI moderates the relationships between two job crafting dimensions—seeking resources and optimizing demands—and career commitment but not seeking challenges. This indicates that the relationship between job crafting and technology acceptance is more complex than previously assumed. In line with COR theory, liking of AI enhances the resource investment process by amplifying the effects of certain job crafting behaviors on attitudinal outcomes. However, its limited influence on challenge-seeking suggests the involvement of different motivational mechanisms. To explain this distinction, we draw on self-determination theory (Deci et al., 2017) and the job demands-resources (JD-R) model (Bakker et al., 2023). Seeking challenges typically involves engaging in tasks that promote personal growth, competence development, and intrinsic satisfaction. These behaviors are less dependent on external tools like AI and more driven by internal motives such as autonomy and mastery. From this perspective, liking of AI does not enhance the impact of challenge-seeking on career commitment because this dimension of job crafting is rooted in intrinsic motivation, which external resources cannot directly fulfill. The JD-R model further supports this view by suggesting that challenge-seeking is fueled by high intrinsic job resources, making the influence of technology-related affective evaluations less relevant.

Practical implications

Our results on the relationships among employees’ GenAI adoption, job crafting, career commitment, liking of AI, and employee outcomes offer valuable implications for managers and employees. AI adoption has become essential for business survival and growth (Giuggioli and Pellegrini, 2023; Raisch and Krakowski, 2021; Sedkaoui and Benaichouba, 2024). Our findings indicate that employees’ adoption of GenAI fosters positive attitudes and behaviors aligned with organizational objectives. However, many managers remain hesitant about integrating AI into employee tasks due to challenges related to change management and the lack of preparedness for technological transitions (Budhwar et al., 2023; Dwivedi et al., 2021; Sanders and Wood, 2023). The finding that job crafting and career commitment sequentially mediate the relationship between GenAI and employee behavioral outcomes suggests that managers can adopt at least three important roles to ensure that GenAI adoption produces desirable outcomes. First, given the organizational benefits of employees’ GenAI adoption, our research suggests that managers should actively promote GenAI adoption while ensuring it aligns with confidentiality protocols and organizational security (Chatterjee et al., 2024). Furthermore, besides integrating GenAI systems into workflow, they can also actively stimulate employees to seek resources, seek challenges and optimize needs. We observed differences in the extent to which employees’ GenAI adoption translated into job crafting. This variation could be reduced, and job crafting increased, if managers take an active role in motivating and enabling employees to convert GenAI use into meaningful job redesign across these three dimensions. For stimulating resource-seeking, organizations can provide quick and autonomous access to relevant knowledge and support. For example, GenAI assistants can facilitate real-time information retrieval (Feuerriegel et al., 2024; Jarrahi et al., 2023), while an organizational climate that encourages feedback-seeking and skill development further reinforces resource accumulation (Meijerink et al., 2020). To foster challenge-seeking, managers can support employees in identifying their skill gaps and guiding them toward growth-aligned opportunities, such as assigning stretch projects or enabling lateral moves (Debortoli et al., 2014; Ekuma, 2023). For optimizing demands, providing GenAI tools that assist with drafting, summarizing, or reorganizing routine tasks can reduce cognitive load, enabling employees to regain time and mental clarity—resources that can be reinvested in higher-value, meaningful activities (Maedche et al., 2019). Additionally, managers should provide complementary organizational resources that support the reinvestment of gains derived from GenAI use and job crafting, such as assigning enriched roles or developmental tasks that allow employees to apply newly acquired skills, reinforcing a cycle of capability building and sustained proactive engagement.

Our results also indicate that employees themselves can influence the outcomes of GenAI through engaging in job crafting and thus reinvesting resources that GenAI provides. Rather than viewing GenAI as a passive tool for task automation, employees can shape its value by adjusting how they seek resources, take on challenges, and optimize their work processes in response to new technological possibilities. For instance, employees can proactively solicit feedback from supervisors or colleagues to refine GenAI-generated outputs, ensuring alignment with organizational standards and expectations. Beyond task-level improvements, engaging in job crafting—by continuously seeking resources, embracing challenges, and streamlining demands—also enables employees to accumulate valuable psychological and instrumental resources that reinforce their career commitment. In turn, stronger career commitment aligns employee behavior with organizational goals, fostering constructive actions such as high-quality voice and reducing counterproductive behaviors such as cyberloafing and workplace cheating. This suggests that employees need not wait for formal structural changes to improve their career prospects; rather, by investing in job crafting as a deliberate and ongoing process, they become active agents in shaping their own development and value within the organization. At the same time, organizations can reinforce this process by offering tailored support for employee growth. For example, organizations can analyze employees’ skills, performance, and aspirations to help create personalized career paths (Tambe et al., 2019) or implement customized learning programs that identify specific skill gaps and recommend relevant courses, workshops, or developmental projects (Maity, 2019; Morandini et al., 2023). Such initiatives foster a continuous sense of progress and development, further strengthening employees’ commitment to their careers and enhancing the long-term impact of GenAI adoption.

Lastly, the finding that liking of AI moderates the relationship between certain job crafting dimensions (i.e., seeking resources and optimizing demands) and career commitment, but not seeking challenges, offers practical implications. To leverage this effect, organizations should implement targeted interventions aimed at increasing employees’ positive attitudes toward GenAI. For example, offering hands-on training sessions, showcasing practical success stories, and providing low-risk opportunities for experimentation can help employees build confidence and emotional comfort with GenAI tools (Wallace and Sheetz, 2014). These initiatives are especially valuable for promoting job crafting behaviors such as seeking resources and optimizing work demands, where liking of AI plays a strengthening role in reinforcing career commitment. However, because liking of AI does not moderate the relationship between seeking challenges and career commitment, organizations should adopt complementary strategies to support this distinct form of job crafting. These may include providing mentoring relationships, encouraging participation in stretch assignments, or offering career development workshops that address intrinsic growth needs independent of GenAI adoption. Such efforts ensure that all forms of job crafting are effectively supported, even those less influenced by employees’ attitudes toward GenAI.

Limitations and suggestions for future research

This study has several limitations that should be acknowledged. First, the correlational nature of this study limits the ability to draw causal conclusions. While we hypothesized that GenAI adoption influences career commitment through job crafting behaviors, it is also possible that employees who are highly committed to their careers may be more inclined to adopt and engage with GenAI, or they may craft their jobs in ways that align with GenAI. Although demographic variables were controlled for, the potential for reverse causality remains. Future research could employ experimental or quasi-experimental designs (Campbell and Stanley, 1963), comparing groups with different levels of GenAI adoption or job crafting behaviors, to establish clearer causal relationships. Longitudinal studies could also examine changes over time in both GenAI adoption and career commitment to provide stronger causal evidence (Zapf et al., 1996).

Second, while we focused on job crafting as a mediator in the relationship between GenAI adoption and career commitment, there may be additional mediators that could further clarify the underlying mechanisms. For instance, the fulfillment of the three basic needs—autonomy, competence, and relatedness—outlined in self-determination theory (Patrick et al., 2007), could influence how employees engage with GenAI and how these engagements, in turn, impact their career outcomes.

Third, this study focused on the moderating role of employees’ liking of AI in the relationship between job crafting behaviors and career commitment. However, there may be other moderating factors that influence this relationship. For example, individual characteristics such as technology self-efficacy (Cheng et al., 2023), digital literacy (Huu, 2023), or cognitive flexibility (Lemonaki et al., 2021) could impact how employees interact with GenAI, and how these interactions influence career commitment. Organizational factors such as leaders’ AI symbolization (He et al., 2024) and digital organizational culture (An et al., 2024) may also act as important moderators. A broader array of potential moderators can enhance the generalizability of the findings. Moreover, to ensure theoretical clarity and avoid overcomplicating the model, we focused on how the liking of AI moderates the downstream effects of job crafting. However, future research could investigate whether the liking of AI also shapes the extent to which GenAI adoption activates job crafting. As GenAI represents a self-initiated technological resource, employees’ affective orientation toward AI may influence not only how they channel crafted resources into career commitment, but also whether they engage in crafting behaviors in the first place (Jia et al., 2024; Park, 2024; Zheng et al., 2025). Examining this earlier stage would offer a more holistic understanding of how emotional receptivity to AI facilitates the translation of GenAI adoption into proactive work design.

Furthermore, a potential limitation of our study is that some of the peer-reported measures, such as seeking challenges, seeking resources, and voice quality, may have been slightly inflated due to social desirability bias. Despite our efforts to ensure anonymity during data collection, the lack of direct personal stakes in peer evaluations can lead to higher-than-actual scores, a phenomenon noted in previous research (e.g., Podsakoff et al., 2003). Moreover, our use of 7-point Likert scales (rather than 5-point scales, as initially described) means that these elevated mean values may be less pronounced than they appeared at first glance. To address this limitation, future research could triangulate these peer-reported findings with more objective measures, such as supervisor evaluations or behavioral observations, to provide a more nuanced and validated assessment of job crafting and voice behaviors.

Finally, the sample for this study was drawn from employees in China, and as such, the findings may be influenced by cultural and contextual factors unique to this population (Brutus et al., 2013). Given the increasing global adoption of GenAI, future research could explore cross-cultural differences in the relationship between GenAI adoption, job crafting, and career commitment to offer valuable insights into the universality and contextual specificity of the proposed model.