Meetings generate work before, during and after — and the after is where most of it gets lost. AI handles the whole cycle well, provided you get the consent question right at the start.
Get the consent question right first
Before you record or transcribe anything: tell everyone in the meeting, every time, and get their agreement. This is both a legal and a trust matter — recording people without their knowledge damages relationships in ways no efficiency gain justifies. Say it plainly at the start: 'I'm using a tool to take notes — everyone happy with that?' If anyone isn't, don't.
Also think about what's in the room. Client confidential information, staff matters, commercially sensitive discussion — these need a properly contracted business-grade tool, not a free consumer one. See is it safe to put business data into AI.
Before: the agenda nobody writes
Most small business meetings have no agenda, which is why they overrun and decide nothing.
Prompt — agenda from a brain dump
I have a meeting with [who] about [topic]. Here's everything on my mind about it: [dump it all out, unstructured]. Turn this into a clear agenda with timings for a [30/60] minute meeting. Order the items so the most important decisions get made while people are still fresh. Flag anything that doesn't actually need this meeting and could be an email instead.
That last instruction is quietly valuable — it regularly removes a third of the agenda.
Prompt — preparing for a difficult meeting
I have a difficult meeting about [situation, described generically]. Help me prepare: what are the likely objections, what questions should I be ready for, what's my strongest argument, and what would I do if they say [likely response]? Be direct — I want to be prepared, not reassured.
During: capturing without missing the conversation
Two routes. Transcription tools capture everything automatically — accurate, but requires consent and a suitable tool. Or take rough notes yourself and let AI tidy them afterwards, which needs no consent and no tool beyond what you have.
For most small businesses the second is the pragmatic default: scribble badly, fix later.
After: where the real value is
Turning rough notes into something usable
Prompt — notes to summary
Turn these rough meeting notes into a clear summary. Structure it as: decisions made, actions agreed (with who and by when), open questions, and anything deferred. If an owner or deadline wasn't specified, mark it 'not stated' rather than guessing. Notes: [paste, with names removed if the content is sensitive].
The "not stated" instruction is what makes this useful rather than fictional — it surfaces exactly the actions that were agreed vaguely and will otherwise never happen.
The follow-up email
Prompt — follow-up email
Write a short follow-up email from these meeting notes confirming what we agreed, listing each person's actions clearly, and noting the next meeting date. Professional and brief — people should be able to read it in 30 seconds and know what they owe: [paste summary].
Sending this within an hour of the meeting is one of the highest-leverage habits in a small business. It converts a conversation into commitments while everyone still remembers.
Different summaries for different people
Prompt — tailored versions
From this meeting summary, produce two versions: a two-line update for someone who wasn't there and only needs the headline, and a detailed action list for the person who has to deliver most of it: [paste].
Client and sales meetings
Prompt — client meeting follow-up
Turn these notes from a client meeting into a follow-up email that confirms my understanding of what they want, sets out next steps and what I need from them, and gives a timescale. Warm and professional. Flag anything in my notes that's ambiguous so I can check rather than assume: [paste notes, client name removed].
The ambiguity flag prevents the most expensive mistake in client work — proceeding confidently on a misunderstanding. See summarising client calls with AI.
Making meetings themselves better
Beyond the paperwork, AI is a useful critic of your meeting habits.
Prompt — meeting audit
Here's a list of our recurring meetings with attendees and durations: [paste]. Which of these look like they could be shorter, less frequent, have fewer attendees, or be replaced by a written update? Be blunt — I'd rather hear it.
What to keep human
- Reading the room — AI can't tell you someone disagreed but didn't say so
- Deciding — a summary of options is not a decision
- Difficult conversations — prepare with AI, hold them yourself
- Checking accuracy — AI misattributes who said what surprisingly often
That last point matters in any meeting with consequences. Read the summary against your memory before circulating it — an incorrect record is worse than none.
Meetings generating more work than they solve?
We help businesses build habits that make the follow-up automatic. AskColin gives small businesses low-cost monthly support with one-to-one help whenever you need it — so when you're stuck on a real job, there's someone to ask. Practical, jargon-free, built around how you actually work.
Get one-to-one AI support for your businessFrequently asked questions
Do I need consent to record a meeting with AI?
Yes — tell everyone in the meeting, every time, and get their agreement before recording or transcribing. It's both a legal and a trust matter; recording people without their knowledge damages relationships in ways no efficiency gain justifies.
How can AI help with meeting notes?
Take rough notes yourself and ask AI to structure them into decisions made, actions agreed with owners and deadlines, open questions and deferred items. Instruct it to mark anything unspecified as 'not stated' rather than guessing — that surfaces the vague actions that never happen.
What's the highest-value meeting habit?
Sending a follow-up email within an hour confirming what was agreed and listing each person's actions. It converts a conversation into commitments while everyone still remembers, and AI makes it a two-minute job rather than one you skip.
Can AI write my meeting agenda?
Yes — dump everything on your mind about the meeting unstructured and ask AI to turn it into an agenda with timings, ordering items so key decisions happen while people are fresh. Also ask it to flag anything that doesn't need a meeting at all.
What should stay human in meetings?
Reading the room — AI can't tell you someone disagreed silently. Making decisions, since a summary of options isn't a decision. Difficult conversations, which you prepare for with AI but hold yourself. And checking accuracy, as AI misattributes who said what surprisingly often.