When AI is in your meetings, and when to say it helped
The silent attendee, and the rule that keeps your reputation.
By the end of this lesson
After this lesson, you'll be able to recognise when AI is taking notes in your meetings and decide whether to keep it on — and you'll have a one-sentence rule for when to say AI helped with your work.
This is the lesson with two halves, because there are two separate calls to make.
First — when an AI is sitting in your meeting, transcribing every word, drafting action items, and sometimes flagging your tone. You're allowed to ask it to leave. A surprising number of people don't realise that.
Second — when AI helped you write something other people will rely on, when do you mention it? Saying nothing has cost more careers than over-disclosing ever has. Here's the honest rule for both.
Part one — AI in the room
Most video-call platforms in 2026 — Microsoft Teams, Zoom, Google Meet, Webex — now have an AI that can transcribe, summarise, and draft action items in real time. Some join automatically if a colleague has set their calendar to send an AI notetaker to every meeting they accept. They are quiet. They take notes. They remember.
What gets recorded is more than you think.
- The transcript. Every word, attributed to a speaker. Including the muttered aside, the half-sentence you walked back, the joke that didn't land.
- Action items. Surprisingly accurate. Sometimes attributed to people who didn't volunteer.
- A summary. Often shared automatically with all attendees afterwards. Depending on the settings, the summary may be shared beyond the people who spoke in the room.
- Tone or sentiment notes on some platforms — flagging utterances as supportive, sceptical, defensive, etc. If your workplace uses those, it's worth knowing.
For internal meetings on your own company tenant, this is bound by your organisation's data policies. For mixed meetings — external partners, client calls, a candidate interview — the recording can sit on a different tenant than you'd expect. The notetaker that joined was on whose calendar?
A Helsinki HR manager joins what she thinks is a quiet wellbeing check-in with a struggling employee. A small icon at the top of the call shows the team's AI notetaker has auto-joined, because the manager who scheduled it has the AI set to join all her meetings by default. The wellbeing conversation is now being transcribed. She asks for it to be stopped. The manager apologises — they'd forgotten the setting was on.
When to switch the AI off — and how to say it
Some conversations need to feel safe being half-baked, sceptical, off-record, or genuinely confidential. The transcript turns those conversations into something else.
Categories where your company policy may require AI notes off — and where it's usually wise to check first:
- HR conversations — pay, performance, complaints, wellbeing, a difficult one-on-one.
- Salary negotiations and procurement negotiations where you don't want every offer you tested logged.
- Brainstorms where people need to feel safe being half-formed.
- Anything covered by legal privilege (advice from your company's lawyer, anything subject to discovery).
- Candidate interviews — many companies require explicit consent.
- Anything where you'd be unhappy if the transcript leaked.
Most platforms now show a small icon when AI note-taking is active. If you don't see it announced at the start of a meeting, the polite, professional version of the question is:
"Quick check before we start — is this being recorded or transcribed by an AI assistant?"
If yes, and you'd rather it weren't, the polite version of the ask is:
"Could we turn the AI notes off for this one? I'd like us to be off-record."
That's it. No drama. Most colleagues will be relieved you asked, because they hadn't thought to.
A Cape Town finance team is mid-quarter-end. Their weekly stand-up has an AI notetaker on for the action-item summary — that's fine. The CFO joins and starts talking about a confidential supplier renegotiation. The team lead says: <em>"Quick one — let me turn the notes off for this part."</em> It takes four seconds. Nobody objects. The conversation continues, off-record.
Part two — AI in the work
Different question. Different rule.
You wrote something. AI helped — drafted it, restructured it, tidied the grammar, summarised three documents into the version you sent. Should you mention it?
The honest rule, and it's the one that's quietly saved a lot of careers in 2024 and 2025:
If AI helped, mention it. If you'd be embarrassed to be asked, mention it loudly.
The public stories that damaged trust in 2024 and 2025 often had the same shape: people used AI heavily, pretended they hadn't, and were caught later. The using wasn't the problem. The hiding was.
This doesn't mean disclose every AI-assisted sentence. That's like disclosing every spell-check correction. The line is whether the audience would feel they're being misled if they found out later.
Almost certainly disclose:
- Anything graded, assessed, or attributed to your individual work.
- Client deliverables that were promised to be human-made.
- Photos, illustrations, or audio you publish as your own.
- Anything where the audience is paying for your judgement (legal advice, medical advice, journalism, creative writing).
- Code that goes into critical systems — say what AI generated and what you reviewed.
Generally don't need to:
- Routine drafts (the same way you wouldn't disclose using spell-check).
- Summarising your own meeting notes for your own use.
- Brainstorming ideas you then properly developed yourself.
- Translation tools you've always used.
The pragmatic test: if you'd be unbothered to have someone watch you produce the work, you don't need to disclose. If you'd close the laptop, you probably do.
An Amsterdam consultant sends a client a strategy memo. AI drafted the structure and the first version of section 3, then she rewrote roughly half of it and added the recommendations herself. She adds one short line at the end: <em>"Drafted with AI assistance, reviewed and edited by [her name]."</em> The client appreciates the honesty. Three months later when the topic comes up at a senior review, no one is surprised. No quiet damage.
The trigger table you can keep
One short table covers most situations you'll meet at work.
| Situation | What to say |
|---|---|
| AI notetaker has joined the meeting | "Quick check — is this being transcribed?" Then: "Could we turn it off for this part?" if needed. |
| HR, performance, salary, or sensitive one-on-one | AI off, by default. Don't rely on the other person to remember. |
| External meeting — partner, client, supplier | Confirm whose tenant the recording lives on, before the meeting starts if you can. |
| You used AI to draft a colleague-facing document | Usually no disclosure needed — same as spell-check. |
| You used AI for client work, graded work, or work attributed to your judgement | One-line disclosure. "Drafted with AI assistance, reviewed by me." |
| Photos, illustrations, audio you publish as your own | Disclose if AI-generated. The pattern of "caught later" is consistent and damaging. |
The shorter version: if AI was in the room, ask. If AI was in the work, mention it when it matters.
Try this for 5 minutes
Open your calendar for the next two weeks. Find one meeting where AI notes really shouldn't be on (HR, sensitive one-on-one, candidate interview, confidential negotiation). Mark a reminder to yourself five minutes before: "Check the notetaker. Switch it off if it's on."
Then look at one piece of work you'll send out this week — a memo, a slide deck, a customer email, a report. Did AI help materially? If yes, write the one-line disclosure now: "Drafted with AI assistance, reviewed by [your name]." See how it reads. Most readers find it reassuring, not alarming.
Keep this
- Rule — AI in the room is a consent question — ask, then ask for it off when needed. AI in the work is an honesty question — disclose if you'd be embarrassed to be asked.
- Phrase — "Quick check — is this being transcribed?" — the four-second question that protects sensitive conversations.
- Don't — Don't hide AI use on work that matters. The using is fine. The hiding is what costs the career.
Pop quiz, no marks
-
You're about to start a one-on-one with a struggling team member. The AI notetaker icon is on. What's the one-line ask?
Show answer
"Could we turn the AI notes off for this one? I'd like us to be off-record." Most platforms switch it off in two clicks. The conversation goes better.
-
AI drafted half of a client deliverable — you wrote the recommendations and signed off. What goes at the end?
Show answer
One line. "Drafted with AI assistance, reviewed and edited by [your name]." Done. The client knows. You're protected if it ever comes up.
Want to go deeper?
Each of these topics on Plain AI explores one idea from this lesson in more detail: