When an employee becomes exceptionally productive with artificial intelligence, colleagues may react in different ways. Some ask how the person achieved the result. Others compare themselves with the successful coworker and become more concerned about what they deserve.
New research suggests that the way work is organized helps explain which reaction becomes more likely. Across a randomized experiment involving 240 full-time employees and a separate three-wave study of 360 employee–supervisor pairs, responses to a colleague’s strong AI-assisted performance depended on how closely employees’ tasks were connected.
Where work was interdependent, stronger AI-assisted coworker performance was associated with more advice seeking, which was linked to better subsequent work performance. Where tasks were largely independent, the same performance signal was associated with greater psychological entitlement and a less favorable pathway to work performance.
Published on 30 October 2026 in Frontiers in Psychology, the research adds a social dimension to AI productivity. An organization’s benefits from technology cannot necessarily be understood by measuring only the employee using it: colleagues observe and respond to those gains.
When productivity becomes a comparison
Much workplace AI research has asked whether employees who use generative tools work faster or produce better outputs. Other studies examine whether colleagues see AI users as diligent, competent or overly reliant on software. This study asks a different question: how does someone react to the quality of another employee’s AI-assisted result?
That distinction is increasingly relevant where AI use is becoming ordinary. If many workers use similar tools, using AI alone may reveal little about individual skill. What stands out is whether the final result meets a demanding standard.
Researcher Yi Wang of Suzhou City University in China drew on social comparison theory. People evaluate their abilities and standing partly by observing others, particularly those in similar positions. A successful colleague can therefore be an example to learn from or a benchmark that feels personally threatening.
The study investigated two responses. Interpersonal knowledge seeking means asking a coworker for useful advice or techniques. Psychological entitlement means a heightened feeling of deserving more recognition or favorable treatment than someone else.
The researcher proposed that task interdependence could change which response prevailed. In a linked workflow, one person’s work is an input to another’s task. In independent roles, employees may complete separate assignments without needing each other’s deliverables.
Where colleagues must collaborate, the successful person’s methods may be directly useful. Where their tasks are independent, the same success may function more as an unfavorable comparison than a practical learning opportunity.
The first study randomized four workplace scenarios
The experiment recruited 240 full-time employees associated with technology-sector companies in China. Participants averaged about 38 years of age, and 141 were men. Each was randomly assigned to one of four conditions.
Everyone read a hypothetical workplace scenario featuring a colleague who used AI to complete a business assignment. Two conditions described a high-quality final result, while the other two described inaccurate or incomplete work. The hypothetical colleague carefully reviewed and revised AI output in every condition.
The second factor was whether the participant’s own work depended on that colleague’s output. In the high-interdependence scenario, the coworker produced a necessary input to the participant’s task. In the low-interdependence scenario, they worked on parallel assignments.
Participants then rated their willingness to seek advice and their sense of entitlement on seven-point scales. Random assignment strengthens the experiment’s ability to test how the described situations influenced those immediate responses, although the participants did not complete real assignments with the hypothetical coworker.
Linked work encouraged advice seeking
Under high task interdependence, the high-performing AI-assisted colleague prompted an average knowledge-seeking score of 5.23 out of seven, compared with 3.60 when the colleague’s performance was low. The difference was statistically significant (p < 0.001).
When tasks were independent, knowledge-seeking means were much closer: 3.41 in the high-performance scenario and 3.17 in the low-performance scenario. That direct comparison did not reach the conventional five-percent significance threshold (p = 0.060).
The interaction between performance and task interdependence was significant. In simpler terms, seeing a colleague succeed with AI had a much stronger association with wanting their advice when the observer needed to build on that colleague’s work.
That result is not proof that employers should make every task interdependent. It describes the reactions produced by particular experimental scenarios, not the long-term effectiveness of a management intervention.
Independent tasks produced a different response
The pattern changed for psychological entitlement. Under low interdependence, participants observing a highly successful AI-assisted coworker reported an average entitlement score of 4.52, versus 3.14 after observing a lower-performing coworker (p < 0.001).
When tasks were closely connected, the corresponding means were 3.33 and 3.15. That smaller difference was not statistically significant at the usual threshold (p = 0.094).
This does not mean colleagues always become resentful when someone else uses AI successfully. The researchers measured average responses to hypothetical situations, not actual attempts to undermine another worker or demands for higher pay.
It does suggest that visible success can have distinct social meanings. An interdependent colleague may see valuable expertise, while an independent coworker may see an uncomfortable standard against which to evaluate their own position.
A second study followed working employee–supervisor pairs
The researcher then examined whether similar patterns appeared in existing workplaces. An initial 380 employee–supervisor pairs were recruited; 360 completed all three survey waves.
In November 2025, employees rated a specific coworker’s AI-assisted performance and their task interdependence with that person. In December, they answered questions about knowledge seeking and entitlement. In January 2026, supervisors evaluated the employees’ work performance.
The staggered measurements and separate supervisor ratings reduced some limitations of relying on a single self-report survey. However, the field component was observational. Employees selected the coworkers they rated, and the researchers did not randomly assign their work structures.
Unmeasured factors such as personal ambition, team culture, salary or management practices could have affected both employees’ perceptions and their supervisors’ performance evaluations.
Two pathways also appeared in real workplaces
Among the 360 matched pairs, the interaction of coworker AI-assisted performance with task interdependence was significant for knowledge seeking (coefficient 0.42, p < 0.001).
When task interdependence was high, stronger coworker performance was positively associated with advice seeking (b = 0.50, p < 0.001). When interdependence was low, the association was negative (b = −0.16, p < 0.05).
Advice seeking itself was positively associated with supervisor-rated performance (b = 0.25, p < 0.001), consistent with a learning-oriented pathway.
Entitlement showed the contrasting pattern. Under low interdependence, perceived coworker AI-assisted performance was positively associated with entitlement (b = 0.33, p < 0.001). Under high interdependence, that relationship was negative (b = −0.15, p < 0.05).
Entitlement was also negatively related to supervisor-rated performance (b = −0.26, p < 0.001). These results support an association between the measured psychological responses and performance, but they do not establish that either response caused a change in productivity.
The indirect effects are not productivity percentages
The paper used 20,000 bootstrap resamples to evaluate statistical pathways between coworker performance and the observer’s own performance.
Under high task interdependence, the indirect estimate through knowledge seeking was 0.13, with a 95% confidence interval of 0.07 to 0.20. Under low interdependence, the indirect estimate through psychological entitlement was −0.08, with a 95% confidence interval of −0.13 to −0.04.
Those are model coefficients, not percentage gains or losses in output. A value of 0.13 does not mean employees became 13% more productive.
The pattern was consistent with the proposed explanations, but indirect statistical associations in an observational study are not evidence that the intermediate psychological responses necessarily caused the later supervisor ratings.
What the findings suggest for AI adoption
Organizations often judge AI implementation by licenses purchased, employees trained and hours saved. These measures can overlook how visible performance gains affect coworkers who must collaborate or compete for recognition.
In teams with connected tasks, successful AI users may become practical sources of expertise. Sharing their techniques could help colleagues finish the next stage of a shared workflow. In largely independent roles, performance comparisons may become more prominent without offering an obvious route to learning.
This adds a useful dimension to Research Today reporting on changing management skills in an AI-enabled workplace. The new paper examines interpersonal responses rather than the skills employers seek.
Similarly, research on knowledge sharing and hotel service innovation considered another industry where exchanging information was associated with useful organizational outcomes.
And Research Today coverage of positive coworker support highlights why workplace relationships deserve separate attention from individual job performance.
More collaboration is not automatically the answer
One tempting conclusion would be to make employees depend more heavily on one another. The study’s author explicitly cautioned against that interpretation.
Creating unnecessary task dependencies can increase meetings, slow decisions and obscure accountability. Interdependence should suit the work itself, not be imposed simply to change how people respond to a colleague’s AI success.
A more reasonable implication is to examine existing workflows. Where employees share inputs, managers might offer opportunities to explain effective AI practices. Where roles remain independent, clear performance expectations and access to learning could help prevent unproductive social comparisons.
These are possible applications, not proven interventions. The research did not randomly assign businesses to different training, recognition or collaboration policies and then measure their effects.
Where the evidence has limits
The experimental component used hypothetical scenarios rather than actual collaboration over time. Its causal evidence applies to participants’ immediate responses to the described situations.
Both scenarios involved a colleague using AI. Therefore, the results cannot isolate what was unique to AI assistance. Employees might react similarly when a coworker achieves exceptional performance without AI tools.
The field study was non-randomized, and employees selected which coworker to describe. The entitlement scale was adapted to the coworker context and could partly reflect stable personal characteristics as well as situational feelings.
Both studies took place in China. Workplace hierarchies, collaboration norms, and attitudes toward AI differ across countries and sectors. Replication would be needed before treating these patterns as universal.
Finally, supervisors rated employee performance in the field study. Those ratings provide an external perspective, but are not direct records of profit, completed tasks per hour or financial productivity.
The hidden organizational side of AI success
The study challenges the assumption that one employee’s AI-driven success will automatically raise everyone else’s performance. Its findings suggest that colleagues may seek useful knowledge when their tasks are linked, while less connected roles may invite other kinds of comparison.
For businesses, the practical question is not simply whether an employee can produce better work with AI. It is how the organization allows that expertise to spread, how contributions are recognized and whether coworkers have genuine opportunities to learn.
Those factors deserve investigation alongside the usual measures of speed and efficiency as generative AI becomes a more routine part of work.
Looking for studies on this topic? Use the Research Today Research Paper Finder to search scholarly research by topic, question, author or keyword.
Source Information
Study title: Inspired or entitled? The contingent effects of coworker AI-augmented performance through the lens of social comparison
Authors: Yi Wang
Journal: Frontiers in Psychology, Volume 17, Article 1969612
Publication date: 30 October 2026
Study design: Randomized four-condition online workplace-scenario experiment followed by a three-wave observational study of employee–supervisor pairs, with moderated pathway analysis and bootstrapped indirect effects
Sample or scope: 240 full-time employees in China in the experiment, and 360 matched employee–supervisor pairs in a field study conducted across November 2025–January 2026
Key finding: Under high task interdependence, stronger coworker AI-assisted performance was linked to greater knowledge seeking and a positive indirect association with supervisor-rated performance (0.13, 95% CI 0.07–0.20); under low interdependence, the entitlement pathway showed a negative indirect estimate (−0.08, 95% CI −0.13 to −0.04)
