Yes, running an AI notetaker in every meeting does wear down how well your team listens. The tool captures everything, so people stop holding the thread themselves, and responses arrive a day late in a comment box instead of live in the room. The repair is facilitation: name the problem out loud, call people in, and protect time that nothing records.
I run a company of embedded generalist operators, and we sit in a lot of other people's meetings [1]. Over the past year the same pattern has shown up across small businesses, startups, nonprofits and larger teams: the transcript quality went up and the conversation quality went down. Here's the method we use to pull it back.
how We Do: build culture on purpose in an AI-notetaker world
Culture in a small company doesn't just happen because two people work well together. It has to be built on purpose, and right now most of that work is change management. These six steps are what we actually do.
- Name the trade out loud. Tell the team what the recorder is for and what it can't do. People behave differently once the trade-off is spoken rather than assumed.
- Decide per meeting type whether the bot is on. A standup and a one-on-one about someone's career are not the same event. Set a default for each and write it down.
- Assign a live listener. One named human is accountable for responding in the moment, not for capturing anything. Rotate the role so it doesn't become a job.
- Facilitate for response. Ask open questions, call people in by name, and leave silence long enough for someone to fill it.
- Put analog time on the calendar. Unstructured, unrecorded conversation is where the good tangents live. If it isn't scheduled, it won't survive a busy quarter.
- Codify how agents show up. Make it obvious when a teammate is talking and when an agent drafted the message on their behalf.
Why the recorder changes behavior in the room
The notetaker is a good tool. What it changes is incentive. Once everyone in the room knows the record exists, the cognitive work of holding the conversation quietly moves off the humans and onto the transcript.
"we sit in a meeting and we throw on an AI noteaker and because we know the AI noteaker is there and capturing all of the information, we forget to actively listen to each other." Britt Gage
Attention shifts from the person to the retrieval
Listening well is an active behavior. It shows up as questions that build on what the other person said, and it's visible to the speaker in real time [2]. When the record is guaranteed, that behavior gets deferred. You'll read it later. Later rarely arrives with the same energy.
The lag becomes normal
My colleague Sarah Matthew, who works in mental health alongside her operating work, watched this shift happen in two stages: first when teams moved from rooms to Zoom rooms, and again when the notetakers arrived.
"with this notetaker um people are losing the capacity to actually pay attention and respond in the moment." Sarah Matthew
A text or an email gives you time to process, which sounds like an advantage until you notice that nobody is reacting to the thing actually happening.
Signals your team has stopped listening live
You can diagnose this without a survey. Watch one week of meetings for these.
| Signal | What it looks like | First move |
|---|---|---|
| Nobody builds on anyone | Each person delivers an update, no one references the previous speaker | Ask one person to respond directly to the last point before adding theirs |
| Decisions get re-litigated | The same question returns two meetings later | End every meeting with the decision said aloud by a human |
| Silence after a question | Long pauses that someone fills by summarizing the agenda | Call a specific person in by name |
| The transcript is the only follow-up | Action items exist only in the AI summary | One owner writes the follow-up in their own words |
| Tangents have disappeared | Meetings end early and on-agenda every time | Leave the last ten minutes with no topic |
Two cultures, one team: humans with humans, humans with agents
Something newer is underneath the notetaker problem. Most teams now run two cultures at once. There's the culture the humans build with each other, and there's the culture forming between humans and the agents they've built to help them.
Culture creates process without rules. So when one person is having Claude scan their inbox and draft replies to teammates, and another person is writing everything by hand, you now have two working norms and no shared agreement about either. The gap isn't in the technology. It's in how people codify what they're doing with it, which is usually nowhere.
| Human-to-human culture | Human-to-agent culture | |
|---|---|---|
| How norms form | Watching how people treat each other | Watching what output shows up |
| What builds trust | Response in the moment, follow-through | Accuracy and consistency of the tool |
| Failure mode | Conflict avoided, resentment stored | Confusion about who actually wrote the message |
| Who owns it | The leader, explicitly | Usually nobody |
The goal is understanding the competency spread across your team, then making human interaction the baseline for engagement rather than the exception. AI adoption at work has moved faster than the norms around it, which is exactly when leaders have to be explicit [3].
Concrete facilitation moves that work
These are cheap and they work in the next meeting you run.
- Raise the bar for attendance. As Sarah puts it: "if you have nothing to add to the meeting, then we probably shouldn't have had the meeting." Fewer meetings with fewer people restores the pressure to participate.
- Call people in, not out. "Priya, you've done this migration before, what are we missing?" gives someone a reason to be present.
- Ask questions that can't be answered with a status. Open-ended beats yes or no every time.
- Start with a check-in that isn't work. Two minutes, everyone speaks, nobody reports.
- Turn the recorder off for the last segment. Announce it. The shift in the room is immediate.
- Write your own recap. A human summary in human words tells the team the meeting mattered enough to think about afterward.
When to go fully analog
Some sessions should have no tooling in them at all. One of our operators recently ran a client retreat with this rule, and the results are worth copying.
What analog protected
No session happened on a computer. Paper, pens and a giant sticky board. A laptop with a recorder sat in the corner for the record, and everything else stayed off screens. Nobody could quietly Google the best way to phrase something, so people said what they actually thought. One of the favorite exercises was naming the adjective each person brought to the company: all the nice things people think and never say out loud.
Where the good ideas came from
The conversations that changed the product most happened poolside, between two people chitchatting about a client, one story reminding them of another. That tangent structure is the thing AI tooling quietly removes, because tooling optimizes toward the agenda and the tangent is off it.
A simple rule of thumb
| Meeting type | Recorder default |
|---|---|
| Status, project sync, cross-team handoff | On |
| One-on-ones and performance conversations | Off unless both people want it |
| Brainstorms and strategy sessions | Off for the generative half |
| Retreats and offsites | Off, with one device in the corner if you need a record |
If you don't have a people function, you are the people function
Most small businesses and startups don't have a dedicated people and culture person. That's not an excuse to leave culture to chance while everything about how people work is changing. When there's no HR team, the founder is the one helping people through it.
- Say what you're adopting, and why, before the tool appears in a meeting invite.
- Give people a named person to bring their confusion to.
- Treat fear of missing out on AI as a real feeling with a real cost, and address it directly.
- Understand the relationships and the handoffs first, then implement. Chasing the productivity of a tool you don't understand yet costs you the relationship it was supposed to support.
This mirrors early SaaS closely. Teams got excited about bots they hadn't thought through, and the tool became the point. AI magnifies whatever was already happening between people. If coordination was weak, it gets weaker faster.
Remote teams and the in-person tone
Remote work is not the villain here. Our operators are distributed nationally and it works [1]. Video does change what you can read off another person, and the mental load of reading engagement through a screen is well documented [4]. Add a recorder to that and the signal drops again.
- Get the team in one room on some cadence, even if it's rare.
- When you do, leave a block with no topic and no outcome. Lunch counts.
- For distributed teams, schedule one recurring call with no agenda and no bot.
- Trust built in person survives the months of screen time that follow.
FAQ
Should we ban AI notetakers from meetings entirely?
No. AI notetakers are genuinely useful, especially for generalists who hold context across many workstreams and clients. The better approach is setting a default per meeting type, so recording is a decision rather than a reflex. One-on-ones, performance conversations and the generative half of a brainstorm are the usual candidates for keeping the tool off.
How do you know if your team has stopped actively listening?
Watch whether people build on each other's points or simply deliver updates in sequence. Other reliable signals include long silences after open questions, decisions getting re-litigated weeks later, and follow-ups that exist only inside an AI summary. Active listening is visible behavior, and its absence is equally visible once a leader looks for it [2].
What is the difference between using AI notes and outsourcing listening to AI?
Using AI notes means the transcript serves as a backup for details while humans stay responsible for responding in the moment. Outsourcing listening means the transcript becomes the only place the meeting exists, and participants mentally check out because retrieval feels guaranteed. The tool is identical in both cases. The behavior around it is what differs.
Does remote work make this worse than in-person work?
Remote work raises the difficulty, because reading engagement through a screen takes more effort than reading a room, and sustained video calls carry a documented cognitive cost [4]. Adding an AI recorder to a video meeting compounds that effect. Distributed teams can offset it with periodic in-person time and at least one recurring unrecorded conversation with no agenda.
What is one low-cost way to re-train a team to listen live?
Run one meeting per week with the recorder off and announce it in advance. Open with a two-minute check-in where everyone speaks and nobody reports status. The shift in attention is usually noticeable inside a single session, and it costs nothing beyond the discipline of leaving the tool switched off.
References
- of All Trades. weofalltrades.com. https://weofalltrades.com
- Zenger, J. and Folkman, J. "What Great Listeners Actually Do." Harvard Business Review, July 2016. https://hbr.org/2016/07/what-great-listeners-actually-do
- Microsoft WorkLab. "Work Trend Index." https://www.microsoft.com/en-us/worklab/work-trend-index
- Stanford News. "Four causes for 'Zoom fatigue' and their solutions." February 2021. https://news.stanford.edu/2021/02/23/four-causes-zoom-fatigue-solutions/