Overview
- A peer‑reviewed paper in The Leadership Quarterly led by Samuel Meyers synthesised six studies and found supervisors trusted employees who worked without large language models 68% of the time versus 24% when employees used AI.
- Lower trust for AI‑assisted work translated into poorer performance ratings, reduced perceptions of an employee’s contribution, and lower suitability for promotion in the experiments.
- The trust penalty disappeared when organisations explicitly encouraged responsible AI use, while prohibition or the absence of clear policy made supervisors significantly more likely to distrust AI users.
- Authors warn the gap between productivity gains and managerial distrust could push employees to conceal AI use as 'shadow AI' and they recommend clear pro‑AI policies plus manager training to avoid reputational harms.
- The findings come from controlled experiments and a graduate‑student sample and the authors say more field research is needed before generalising across all jobs, and there is no evidence yet of widespread employer policy change.