Skip to content

Signals · Workplace Trust

Known AI use cut supervisor trust from 68% to 24%, and explicit encouragement to use AI removed the penalty

Managers trust staff less once AI is involved, a six-study paper finds, unless the company has explicitly encouraged its use.

by ·4 min read·

New here? Start with the free AI Survival Kit →

Known AI use cut supervisor trust from 68% to 24%, and explicit encouragement to use AI removed the penalty

0:00 / 6:27

This voice is generated by AI.

Known AI use cut supervisor trust from 68% to 24%, and explicit encouragement to use AI removed the penalty

Supervisors trust employees less once they know AI was involved in the work. HR Katha reported on 6 November on a six-study paper in The Leadership Quarterly. When an employee did a task alone, supervisors accepted their account of how long it took 68% of the time. When the employee had used AI-generated output, that fell to 24%. The lower trust then spread to other judgements. Across the studies it went with poorer perceived performance, weaker ratings of organisational citizenship and lower promotability. One condition removed the effect. Where the company explicitly encouraged staff to use large language models, the penalty was no longer statistically significant. Where AI was banned, or where there was no stated policy, the penalty stayed.

The penalty sits in the silence

The obvious response is to protect yourself: use the tools and say nothing. The study did not test it. The researchers measured whether supervisors would believe an employee's account of how long a task took, or insist on checking it. The quality of the work was not part of the question. The design suggests the drop came from knowing AI was involved. Supervisors' own experience with AI did not reliably soften it, and neither did how important the task was. AI-assisted work did score better than work copied from another person, which is a low bar.

That leaves policy as the one thing that changed the result. Where there was no clearly stated policy, the trust penalty remained, and it is the employee who is trusted less.

HR Katha notes that employees may feel compelled to conceal their AI use, which can contribute to shadow AI, where staff use tools without saying so. Hidden use is a bet that nobody finds out. When someone does, you are still working under a silent policy, and now you have a credibility problem as well.

The work nobody sees

A second survey shows why the silence is expensive. In October 2026, Founder Reports surveyed 2,052 US employees who use AI at least once a week. Among the 1,856 who had trained a new employee, 60% said training, overseeing, checking and fixing AI's work takes at least as much effort as training a person. One in three said it takes more. The survey calls this oversight work botsitting. Of all respondents, 31.6% spend at least three hours a week on it and 8.4% spend more than ten.

The load grows with seniority. Founder Reports found that 40.7% of managers and above spend at least three hours a week overseeing AI and fixing its errors, against 25% of individual contributors. The same report cites Microsoft's 2026 Work Trend Index, in which workers ranked quality control of AI output (50%) as the top emerging skill, ahead of critical thinking (46%).

Put the two findings side by side. A supervisor who learns that AI touched a piece of work rates the person who delivered it lower. Founder Reports counts the hours behind work like that: 31.6% of weekly users spend at least three hours a week overseeing AI and fixing its output. Neither study asked whether supervisors see those hours.

AI does the routine work. You do the thinking. When the thinking leaves no trace, a supervisor can only score the routine part, and the Leadership Quarterly paper shows how they score it.

Your Next Move

  1. Get the policy written down this month. Ask your manager, in writing, which uses of AI are encouraged on your team and for which tasks. Explicit encouragement from the company was the one condition in the study that removed the trust penalty. A short team rule, agreed and recorded, takes you out of the no-policy condition. If you manage people, write it for them before the next review cycle opens.

  2. Record how you used the tool. When you hand over AI-assisted work, attach three lines: what the tool produced, what you checked it against, and what you changed. The supervisors in the experiment doubted the employee's account of the work. A short record of your checks answers that doubt before it forms. It also shows the judgement that the finished work hides.

  3. Log your oversight hours for four weeks. Record the time you spend briefing, checking and correcting AI output. Founder Reports found almost a third of weekly users already spend three hours or more on it. That is a measurable share of your week that currently counts for nothing in how you are assessed. Bring the log to your next one-to-one and ask for that time to count when your work is judged.

I build systems like the one publishing this site. → Work with me

About the author

Jo

Jo runs The War Room: one signal a day on how AI is changing work, and what to do about it.

Sources

Get the Briefing

Want more intelligence like this?

The weekly briefing, free — and the AI Survival Kit with it.

Takes 30 seconds. No spam.

More in Signals

See all