AI Task Substitution Quietly Erodes Employee Learning, Study Finds

When artificial intelligence quietly takes over discrete tasks within a workflow, employees do not simply lose those tasks. According to new research published in Current Psychology, they may also lose something less visible but arguably more valuable: the motivation to learn. A team of psychologists from Central China Normal University, led by Yuyan Wang, Ruining Wang, Yang Yang, and Ye Li, has found that partial AI task substitution can inhibit employee learning behavior by eliciting job anxiety, and that a job-growth mindset can attenuate this indirect negative process. The findings arrive at a moment when organizations worldwide are reallocating work between humans and machines, often without a clear picture of the psychological consequences.

The study defines AI task substitution with unusual precision. It is not wholesale replacement, in which a machine assumes an entire job, nor is it augmentation, in which AI merely assists. Instead, AI task substitution is a form of task reallocation in which AI assumes responsibility for discrete tasks while employees remain actively engaged in other interdependent components of the same workflow. A customer service agent, for example, may keep handling complex complaints while a conversational AI system takes over routine inquiries. This partial substitution is now the dominant form of AI-driven work transformation, and it creates a psychologically ambiguous situation: employees keep their jobs, but the boundaries of those jobs are shifting beneath them.

To understand how employees appraise this situation, the researchers drew on cognitive appraisal theory, the classic framework developed by Richard Lazarus and Susan Folkman, which holds that emotional responses to events depend on how individuals evaluate those events rather than on the events themselves. From this perspective, AI task substitution functions as a stressor that employees appraise as threatening. When a machine takes over tasks that once signaled competence and contribution, employees may interpret the reallocation as evidence that their skills are becoming obsolete, triggering job anxiety. That anxiety, in turn, consumes the cognitive and motivational resources needed for learning behavior, the proactive activities through which employees acquire new knowledge and skills at work.

The empirical case was built in three stages. Study 1 recruited 469 Chinese employees who had experienced partial AI task substitution in real workplace settings and surveyed them online. The correlational data supported the proposed indirect relationship: employees who reported greater AI task substitution also reported more job anxiety, and that anxiety was associated with less learning behavior. But correlation alone cannot establish the direction of the causal chain, so the team turned to experiments.

Study 2 brought 204 participants into a scenario-based experiment set in a clerical work context. Participants read a vignette describing an employee whose tasks were partially reallocated to an AI system, with the degree of substitution manipulated across conditions. The experimental evidence confirmed the successive links of the proposed model: AI task substitution increased job anxiety, and job anxiety reduced learning behavior. Critically, Study 2 also tested a moderator. The researchers measured participants’ job-growth mindset, the belief that one’s job capabilities are malleable and can be developed through effort, and found that the indirect negative relationship from AI task substitution to learning behavior was weaker among employees with a stronger job-growth mindset.

Study 3 extended the design to a customer service context with 40 participants and offered preliminary evidence consistent with the buffering role of job-growth mindset. Across the three studies, the pattern held together as a moderated mediation model: AI task substitution elicits job anxiety, job anxiety suppresses learning behavior, and job-growth mindset attenuates this indirect process. The consistency across correlational and experimental methods, and across two different occupational contexts, strengthens the case that the relationship is not an artifact of a single research design.

The theoretical significance lies in what the model reveals about the psychology of partial automation. Previous research on AI at work has documented both bright sides, such as augmented creativity and improved service performance, and dark sides, such as self-esteem threat, burnout, and unethical behavior. This study adds a specific mechanism to that literature. It suggests that the threat is not triggered by AI itself but by the appraisal of task loss within an interdependent workflow. Because employees remain engaged in other components of the same work, they cannot dismiss the substitution as irrelevant, yet they cannot fully control it either. That ambiguity is precisely what cognitive appraisal theory predicts will produce anxiety, and what the data confirmed.

The practical implications are equally pointed. Organizations often frame AI adoption as a productivity story, assuming that employees freed from routine tasks will invest the recovered time in learning and development. The research suggests this assumption may be backwards. If AI task substitution elicits job anxiety, employees may respond by withdrawing from learning rather than embracing it, undermining the very adaptability that organizations need for long-term competitiveness. The authors argue that managers should treat the psychological experience of task reallocation as a central variable in AI transformation, not a side effect to be ignored.

The buffering role of job-growth mindset offers a concrete intervention target. Employees who view their job capabilities as developable, rather than fixed, appear to appraise AI task substitution less as a verdict on their worth and more as a change to be navigated. Organizations can cultivate this mindset through development-oriented feedback, training programs that emphasize skill malleability, and communication that frames AI reallocation as an opportunity for capability growth. The study’s authors note that such practices align with prior work showing that growth mindsets about the self and job predict happiness at work and proactive behavior.

There are limits to what the findings can claim. The samples were drawn from Chinese employees, and cross-cultural research suggests that responses to workplace threat may vary across contexts. Study 3 involved a small sample, and the authors describe its evidence as preliminary. The scenario-based experiments, while valuable for establishing causality, cannot fully reproduce the accumulated experience of living through AI transformation over months or years. The dataset is available from the corresponding author upon reasonable request, and the study was approved by the Central China Normal University Ethics Committee and funded by the National Natural Science Foundation of China.

Even with these caveats, the research lands with force at a time when generative AI tools are spreading through offices faster than any previous workplace technology. The message for organizations is that the human response to machines taking over tasks is not a footnote to the transformation but its psychological core. If companies want employees who keep learning as AI reshapes work, they must manage the anxiety that substitution produces and build the growth mindsets that allow employees to see machines as partners in development rather than as judges of their obsolescence.

Subject of Research: Psychology

Article Title: When machines take over: how AI task substitution affects employees’ learning behavior

Article References: Wang, Y., Wang, R., Yang, Y., & Li, Y. (2026). When machines take over: how AI task substitution affects employees’ learning behavior. Current Psychology, 45(20), Article 1623. https://doi.org/10.1007/s12144-026-10150-4

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10150-4

Keywords: artificial intelligence, employee learning, job anxiety, growth mindset, workplace psychology, AI task substitution, machines, take, over, task, substitution, affects