When Not to Use an AI Notetaker at All
Almost everything written about AI notetakers is written by people who want you to use one. This site has published a great deal about using them well, which is exactly why it is worth being clear about the other side: there are meetings where the right configuration is off, and recognizing them is part of using these tools competently rather than a failure of nerve. The cases divide into three kinds with three different reasons behind them, and conflating those is why people either record everything or abandon the tools entirely.
This is judgment and practice, not legal advice. Some of the situations below touch areas where formal obligations may apply, and those vary by location and by organization. Where it matters, get proper advice.
Three different reasons to leave it off
The three categories are worth separating, because they call for different judgments:
- Recording changes the conversation. The meeting depends on candor, and a recorder suppresses it. The cost is the meeting itself.
- The record is the liability. The conversation is fine to have, but a durable searchable copy of it creates a problem that outlives its usefulness.
- It adds nothing. No harm, no benefit, just an extra artifact nobody reads.
Most people only think about the second, which is why they end up recording conversations that quietly stop being useful.
When recording changes the conversation
Some meetings work because people speak without weighing every sentence. Introduce a recorder and you get a more careful, less useful version of the same meeting.
The clearest cases:
- Anything where someone might admit a problem. A project going badly, a mistake, a deadline that will slip. If the point of the meeting is to surface bad news early, a recorder is working directly against it.
- Genuine brainstorming. Half-formed ideas need to be safe to say. People edit themselves when a transcript is being produced, and you get fewer bad ideas and therefore fewer good ones.
- Negotiations. Both sides become more positional and less exploratory.
- One-to-ones, for reasons covered in one-on-ones vs group meetings.
- Skip-levels, where the entire purpose is that someone can speak freely about their own manager. Recording these defeats the format completely.
The test: ask whether you want people to say the unpolished version. If yes, do not record.
When the record is the liability
Here the conversation is unremarkable and the durable copy is the problem.
Anything involving someone's personal circumstances: health, family situations, financial difficulty, an accommodation request. Anything about a complaint, a dispute, or an investigation. Compensation and performance discussions. And any conversation where a participant is sharing something with you in confidence.
The issue is that meeting archives are rarely deleted. A transcript created for convenience persists, is searchable, and may be accessible to more people than you assume, as covered in where your meeting recordings live. A sensitive conversation that would have faded from memory becomes a permanent, retrievable document instead.
Ask a simple question: would this transcript be a problem in two years, read by someone who was not there and does not have the context? If plausibly yes, take notes by hand.
When it genuinely adds nothing
The least discussed and most common category. No harm, no value.
- Short transactional meetings. A ten-minute check-in with one outcome does not need a transcript. You will remember, or a one-line message covers it.
- Meetings already documented elsewhere. If decisions land in a ticket, a project board, or a document as part of the process, a parallel transcript is a second record nobody consults.
- Presentations you are attending. The deck is the record. A transcript of someone reading their slides is not useful.
- Meetings whose notes you have never once reopened. This is the honest test. Look at your archive and check which recurring meetings you have gone back to. Most people find the answer is one or two, and everything else was captured out of habit.
Turning these off costs nothing and shrinks the archive that creates the risks in the previous section.
The cost nobody counts: attention
There is a subtler cost worth naming, and it is the one the marketing never mentions.
The promise is that recording frees you to be present instead of taking notes. That is often true. But there is a version where knowing everything is captured makes you listen less carefully, because the transcript is the safety net. You stop doing the synthesis that happens while writing a note by hand, which is where a lot of understanding actually forms.
Writing something down is not just storage; it is a decision that it mattered. Handing that entirely to a tool means nobody makes that decision during the meeting, and you get an accurate record of a conversation you engaged with less. For meetings where your thinking is the point rather than the record, that trade is a bad one.
Recording selectively is a skill, not a compromise
The most common failure is treating this as binary. Either every meeting is recorded automatically, or the tool is abandoned as untrustworthy.
Selective recording is the competent position. It looks like: project work and client calls recorded by default, one-to-ones and sensitive conversations never, and a live judgment on the rest. It requires turning auto-record off and deciding per meeting, which is mildly inconvenient and takes about three seconds.
This is also easier to explain to colleagues and clients than either extreme. "We record working sessions but not one-to-ones" is a policy people find reassuring, and it makes the tool's presence meaningful rather than ambient. It is a core part of the rollout advice in rolling a notetaker out to a team.
What to do instead
When you decide not to record, you still need something.
Write four lines afterward. What was decided, who owns what, and anything to follow up. Two minutes, and it captures more of what mattered than a transcript you will never reopen.
Say the decision out loud before the meeting ends, so everyone leaves with the same understanding. This is good practice whether or not you are recording.
Send a short follow-up message where a shared record matters. A three-line message people actually read beats a transcript nobody opens.
If someone asks you to stop mid-meeting, that is its own situation, handled in when someone in the meeting objects to the bot, and how you disclose recording in the first place is covered in telling people you are recording. For commercial engagements specifically, see recording a client call.
None of this argues against the tools. Used on the right meetings they save real time and catch things you would have lost, which is what the rest of our guides cover, starting with how to choose an AI meeting note taker. Knowing where they stop being useful is what separates using them well from using them everywhere.
FAQ
Should I record every meeting?
No. Selective recording is the competent position: default on for project work and client calls, off for one-to-ones and sensitive conversations, and a judgment call on the rest.
Which meetings should never be recorded?
Conversations that depend on candor, such as skip-levels, genuine brainstorming, and meetings where someone might admit a problem. Also anything involving personal circumstances, complaints, disputes, performance or compensation.
How do I know if a recording is a liability?
Ask whether the transcript would be a problem in two years, read by someone who was not there and lacks the context. If plausibly yes, take notes by hand.
Is there a downside to recording a harmless meeting?
Two. It grows an archive that is rarely deleted and often more widely accessible than assumed, and it can make you listen less carefully, since deciding what to write down is part of how understanding forms.
What should I do instead of recording?
Write four lines afterward covering what was decided and who owns what, state decisions out loud before the meeting ends, and send a short follow-up where a shared record matters.