AI Notetakers in 1-on-1s: What They're Actually Costing Your Team's Honesty
AI Notetakers in 1-on-1s: What They're Actually Costing Your Team's Honesty
Quick Bridge: Somewhere on your team right now, a direct report is choosing a smaller answer than the one they actually have, because a recording tool is sitting in the meeting and nobody has said out loud what happens to it afterward.
The next version of this is already happening. I have seen it come up often enough now that it stopped feeling like an edge case. Some managers have started using AI to score the sentiment of 1-on-1 conversations. Not what was discussed, not action items, but whether the person on the other side sounded positive or negative. No conversation about it beforehand, no explanation of who sees the score or what it gets used for.
Just a running record of how someone came across, building quietly in the background of a meeting that used to be theirs. Once you know this pattern exists, it explains itself in one sentence: the person across from you now has to manage not just what they say, but how a machine will grade the way they said it.
When leaders bring this into a coaching conversation, the first question usually turns legal: can my employer record me on company property? In most places, yes. That is also beside the point. The real question is not whether it is legal. It is what it feels like to walk into a conversation that used to be safe and realize it is being graded.
The AI Notetaker Trust Problem Nobody's Naming
Most of what gets written about AI notetakers is a legal or IT conversation: consent requirements, data retention, which vendor has the least risky terms of service. That conversation matters, and it is also missing the actual damage.
A 1-on-1 only works because a direct report believes they can say the true version of something and not have it come back to hurt them. That belief is fragile. It does not survive a "yes, but" the way a project deadline does. The moment someone has to wonder who else might read a transcript, or how a sentiment score might get referenced three weeks from now, they stop answering the real question and start answering the safe one.
I have watched this happen from the other side of a very similar setup. Early in my coaching training, recording client calls was standard practice, every client agreed to it, and I recorded so my supervisor could review my work.
What I noticed had nothing to do with my coaching skill. Clients gave shorter answers once the recording started, less emotional depth, the same person and the same question producing a visibly smaller answer than the one I knew was actually in there.
Why Consent Doesn't Fix the Chilling Effect
None of that changed once I moved to a recorder my clients could not see as easily. The consent was identical. What changed was how present the recording felt in the room, and the answers got longer again the moment it stopped feeling present.
That gap between agreeing to something and forgetting it is happening is the whole mechanism. Consent is a real requirement, and it is also not the fix leaders think it is. A signature does not make someone forget there is a second audience.
One in three U.S. workers now say an AI notetaker or transcription bot has sat in on their meetings. Only about a third of that group say they were reliably asked for permission first, according to a 2026 survey commissioned by Kolmogorov Law and reported by ABC17 News. That means most of the workers experiencing this did not get a clear conversation about it at all, let alone a chance to ask what happens to the recording afterward.
And even in the minority of cases where someone was asked, the research on how people actually behave once they know a recording exists tells you consent alone will not restore candor. People do not stop talking. They start managing what they say, which looks identical from the outside and is not the same thing.
How to Keep 1-on-1s Honest When an AI Notetaker Is in the Room
If you are serious about protecting the one meeting built for honesty, the work is behavioral, not procedural. A policy in an employee handbook does not do it.
Say what the tool is for, out loud, every time. Silence about a recording tool is what makes people fill in the worst-case explanation themselves. Tell them who can see the notes, how long they are kept, and what they get used for.
Ask what kind of tool you actually have running. A notetaker that only captures what was decided is a different animal from one that scores tone, and if you cannot explain in one sentence why the second kind needs to exist in this specific meeting, it does not belong there.
The chilling effect rarely looks like someone refusing to talk. It looks like an answer that used to run three minutes now running thirty seconds, and that shift is the signal, not just a quiet week. Watch for the shorter answer, not just the missing one.
You do not need artificial intelligence to solve a trust problem. I switched from a visible recorder to a handheld one set near my speakers and got longer, more honest answers back, with the exact same consent in place. A notebook works fine for the conversations that matter most.
When AI Notetaking Goes Too Far: The Sentiment-Scoring 1-on-1
The sentiment-scoring pattern described above is not a rare edge case. It is where this practice goes once nobody puts a boundary on it, and it is worth sitting with as its own example rather than a footnote.
A tool that only logs action items is annoying at worst. A tool that scores whether someone "sounded positive" turns every word choice into something with a running tally attached, and it does that without the direct report ever agreeing to be evaluated that way. The person on the receiving end of this is not upset about being recorded. They are upset that a machine is quietly deciding how they came across, in a meeting where showing up unpolished was supposed to be allowed.
This is also where "just get consent" fully falls apart as a fix. Even a direct report who signs off on a sentiment tool is not signing off because they trust it. Plenty of people will consent out of obligation to their job, then quietly adjust what they say to avoid anything that could read as career-limiting later, exactly the behavior the tool was never supposed to cause.
You cannot tell that group apart from the group that is genuinely fine with it, because the transcript looks identical either way. That is the actual risk: the tool cannot show you what it is costing you.
If a tool like that is already running in your organization, the fix is not a better consent form. It is deciding, out loud and in advance, that nothing from the record ever gets quoted back at someone in a performance conversation, and telling your team that is the rule before they have to wonder about it themselves.
On Monday, Look For This
Pick the 1-on-1 this week where the conversation actually matters most, the one where you need the real version of what is going on. Turn off any recording or notetaking tool for that single meeting, and tell the person why you turned it off. Then pay attention to whether their answers get longer.
If a notetaker sat in on your last 1-on-1, would you have said everything the same way without it?
You're great at the work. Let's make you impossible to ignore.
If you are working out how much of your team's honesty a well-meaning tool might be quietly costing you, consider reaching out.
Comments
Post a Comment