On the other hand, PAIA may further influence employee thriving at work. Consistent with Self-Determination Theory (SDT), AI-supported autonomy can trigger both resource-enhancing and resource-depleting psychological processes, with important consequences for employee outcomes. Specifically, when employees perceive AI-supported autonomy through the automation of monotonous tasks, the provision of insights that support better decision-making, and the enablement of more tailored interactions, they are more likely to feel more engaged and subsequently begin to thrive at work.

Furthermore, drawing on Social Cognitive Theory (SCT), we consider individual differences in self-construal as a key boundary condition that shapes how employees respond to AI-supported autonomy. If AI-supported autonomy reduces AI techno-exhaustion while enhancing thriving at work, it should also lead to improved performance in service jobs and increased innovative behavior. Based on this reasoning, it becomes essential to understand the conditions under which employees differ in their responses to AI-supported autonomy, as well as the factors that regulate the strength of these relationships. To address this, we draw on the concept of interdependent self-construal. We argue that the negative relationship between PAIA and AI techno-exhaustion, and the positive relationship between PAIA and thriving at work, will be weaker for employees with lower levels of interdependent self-construal (relative to those with higher levels).

By developing and testing this research model, our study contributes to the literature in three important ways. First, we develop and test a dual-path framework showing how AI-supported autonomy can enhance employees’ service job performance by helping them thrive, while simultaneously generating strain by increasing AI techno-exhaustion. This extends beyond the common view that AI is either purely beneficial or purely harmful. Second, by applying SDT to the AI context, we explain how autonomy-enhancing technologies can simultaneously fulfill and exhaust psychological resources, thereby challenging overly optimistic beliefs about autonomy-driven digital work design. In doing so, we advance theorizing on autonomy-enhancing digital work designs. Third, using SCT, we identify interdependent self-construal as a significant boundary condition influencing how workers interpret and respond to AI-supported autonomy. By integrating SDT and SCT, we provide a multilevel psychological explanation of AI-enabled autonomy in organizations.

Finally, we test our model using time-lagged data from service workers in China. This service-sector context is particularly important for our study because service organizations rely heavily on human–technology interaction and customer-facing tasks, where AI-supported autonomy is especially salient for shaping service job performance and innovation outcomes. This context offers a theoretically meaningful setting for analyzing AI-supported autonomy and its consequences.