By Dex, Web, HUMAN HIVE
A new six-study paper in The Leadership Quarterly finds supervisors trust employees less when those employees use AI, even when the work is good. The fix is cheap: a clear, written "yes, use it, and here's how we show our work" from leadership shrank the penalty in the research.
The 60-second version
Supervisors took an employee's word without checking 68% of the time when the person worked alone, and 24% of the time when AI was involved.
The penalty was biggest where AI was banned or nobody had a stated policy, and smallest where leaders openly encouraged AI use.
The fix: write a one-paragraph AI use norm, judge the work instead of the tool, and make "show your work" normal for everyone, managers included.
What happened
On November 4, the Hebrew University of Jerusalem announced new research by Samuel Meyers, Dr. Yochanan Bigman and Dr. Roman Briker of WHU Otto Beisheim School of Management. Their paper, "Psychological reactions to subordinates' LLM use at work" in The Leadership Quarterly, runs six studies on one question: what happens to a boss's trust when an employee uses AI?
The headline number: employees who did the work themselves had a 68% chance of earning a supervisor's trust. When they used AI, it was 24%. As Bigman put it in the release, "AI is changing not only how we work, but also how we judge the people we work with."
Policy mattered. In one study of 601 supervisors, StudyFinds reported that trust in AI users sat at 19% when the company banned AI, 33% when the company said nothing, and 47% when the company encouraged it, against 58% for a task done alone. HR Katha's November 6 summary adds that lower trust was linked to lower ratings of performance and promotability, and that supervisors' own experience with AI did not reliably make the penalty go away.
Why it happened: the human side
I build and fix websites for a living, so I think about this like a broken page. The code can be perfect and the visitor still bounces if something feels off. Trust works the same way.
Silence reads as hiding. When there's no stated rule, using AI feels like a secret, and secrets cost trust. The study's own words: "the trust penalty remained when organisations prohibited AI use or had no clearly stated policy on its adoption."
Managers judge the process, not just the output. As Bigman put it, an employee "may use AI to produce better work more efficiently yet still lose their manager's trust" simply because AI was involved.
It cuts both ways. A separate 2025 study of 1,100 professionals by Anthony Coman and Peter Cardon found that only 40% to 52% of employees saw supervisors as sincere when their messages leaned heavily on AI, compared with 83% for low-AI messages. Nobody gets a pass. The problem isn't who uses AI. It's that nobody agreed on what good use looks like.
What Mitch has taught on this
Mitch has been loud about this for a while. In his Mitch's Moment on the "AI slop" panic, he framed the real test this way: "The real question is not, 'Did someone use AI?' It is, 'Did this create value for another human being?'"
And on what actually matters now: "The competitive advantage is wisdom. It's judgment. It's knowing what decisions we should and shouldn't give away to machines." About people who use AI on real work with real value: "if they happen to use AI in the process, who cares for real?"
Watch the full episode: Why the AI slop panic is off base (Mitch's Moment).
Best practices: 5 moves for this week
Write the norm in one paragraph. Leadership, not IT, states it: where AI is encouraged, where it's off limits, and what people disclose. The research gap between "silence" and "encouraged" was 14 percentage points of trust.
Judge the work, not the tool. Managers review against three questions: is it accurate, is it useful to the reader, and can the person explain it? If yes, how it was drafted is a footnote.
Make "show your work" normal. One line at the end of any AI-assisted deliverable: what AI did, what the person checked, what they decided. Disclosure stops being a confession and becomes a habit.
Managers go first. Leaders share their own AI-assisted drafts with the same one-line note. If the boss hides it, the team will too.
Ask "walk me through it." Before trusting or dismissing anything, a manager asks the person to explain one decision in it. That's how you check judgment, which is the part AI can't supply.
What's driving people crazy this week
"Botsitting." The Founder Reports Botsitting Report, an October 2026 survey of 2,052 U.S. workers who use AI weekly, found that 60% of those who have trained an entry-level employee say getting good work from AI takes at least as much effort. Now picture doing all that checking and then being trusted less for it. The human fix: count the checking as real work. Name it in the deliverable note, and give people credit for the judgment it takes.
What this means for the next 1 to 3 years
"Proof of judgment" becomes a work product. We expect the short note on what a person checked and decided to become as normal as a file name. The best teams will treat it like a signature: this is mine, I stand behind it.
Performance reviews shift from output to decisions. When anyone can produce a clean draft, reviews will ask what you caught, what you changed and what you chose not to do. That's a better measure of a person anyway.
Trust becomes a team skill, not a feeling. Teams will rehearse it. Picture a quarterly session where a team reviews three AI-assisted decisions together and talks honestly about where the human judgment made the difference.
How we evolve
The skill to build is explaining your judgment out loud. As AI writes more of the first draft, the human value moves to the choices: what you kept, what you cut and why. People who can explain those choices clearly will earn trust fast, with or without AI in the mix. Mitch put the question every one of us should be asking this way: "What we should be doing is saying, 'Well, what skills do I need to evolve?'"
Daily practice: at the end of each day, write one sentence about a decision you made on an AI-assisted task. "AI suggested X. I chose Y because Z." Thirty seconds. In a month, you'll have a record of your judgment and a habit of noticing it.
Try this today
Ask at your next team meeting (5 minutes): "Where are we using AI right now that we haven't said out loud?" Write down what you hear. No judgment.
Draft your team norm (15 minutes): copy the template below, fill in the brackets, send it to your team.
Add the one-line note (2 minutes): on the next thing you send, add "AI helped with [part]. I checked [what] and decided [what]."
Free tool: the one-paragraph AI norm
"On this team, we encourage AI for [drafting, research, summaries]. We don't use it for [client data, final decisions on people, anything confidential]. When AI helps with a deliverable, add one line: what AI did, what you checked, what you decided. We judge the work by whether it's accurate, useful and something you can explain. Managers follow the same rule."
FAQ
Do managers really trust employees less when they use AI?
Yes, in this research. Across six studies in The Leadership Quarterly, supervisors trusted employees' accounts 68% of the time when they worked alone and 24% when AI was involved.
How can a company reduce the AI trust penalty?
State a clear policy that encourages AI use. In one study, trust in AI users rose from 33% with no policy to 47% when the company encouraged AI. Pair the policy with a simple "show your work" habit.
Should employees disclose when they use AI at work?
Where the team has agreed on a norm, yes, in one plain line: what AI did, what you checked and what you decided. It turns disclosure into proof of judgment.
Where HIVE can help
Adoption is a behavior problem. Find your team's friction points in one short conversation with AI enablement from HIVE. Not sure where your team stands? Take the free readiness assessment.