What you’ll be able to do
Right, so the goal here is simple: you sit in a meeting on Teams, Zoom or Google Meet, you actually pay attention, and something else writes down what happened. Not just a wall of text either. Decisions. Actions. Who’s doing what by when. The stuff you’d normally scribble on a notepad and then lose.
You won’t need to type a thing during the call. That’s the whole point.
Before you start
A few things worth knowing before you switch anything on.
First, tell people you’re recording. In the UK it’s lawful to record a call you’re part of, but that’s not really the point, is it. Workplaces and clients expect a heads up, and some regulated industries (legal, healthcare, financial services) actually require consent before you record anything. So just say it out loud at the start: “I’m going to let AI take notes on this one.” Takes two seconds, saves an awkward conversation later.
Second, match the tool to the platform you’re actually using. Teams has its own AI recap built in. Zoom has AI Companion. Google Meet has a Gemini note-taking feature. Then there’s a separate category of bots, like Otter, Fireflies and tl;dv, that join any call regardless of platform. And then there’s Granola, which doesn’t join the call at all, it just listens to your computer’s audio. More on all of that in a second.
Third, and this one matters: don’t put confidential meetings into a consumer-grade AI tool without thinking it through. A casual team catch-up, fine. A conversation involving client data, HR issues, legal matters or anything commercially sensitive? Check your company’s policy first. Some tools store recordings and transcripts on their own servers. That’s not automatically a problem, but it’s something you should know before you hit record, not after.
Get set up
The setup depends entirely on what you’re using to have the meeting in the first place. I’ll go through the main three, then the bot alternative.
Microsoft Teams. If your organisation has a Copilot licence or Teams Premium, you get access to Intelligent Recap, which gives you AI-generated notes, mentions of you, and a searchable transcript after the call. To turn it on, start recording or transcription during the meeting (there’s an option in the meeting controls, usually under “Record and transcribe”). The recap appears afterwards in the meeting chat or calendar entry. Without one of those licences, you still get basic transcription in many Teams plans, just not the AI summary layer on top.
Zoom. AI Companion comes included with paid Zoom plans. During a meeting, look for the AI Companion icon in the toolbar and turn on meeting summary. It’ll generate a summary automatically once the call ends, sent to the host and anyone else who was invited to receive it. No bot joins as a separate participant here, it’s built into Zoom itself.
Google Meet. On eligible Google Workspace plans, there’s a “Take notes for me” option, powered by Gemini. You’ll find it in the meeting controls before or during the call. It captures notes and shares them afterwards with attendees, usually as a Google Doc linked in the calendar event.
Bots that join any call. Otter.ai, Fireflies and tl;dv all work differently, they don’t care what platform you’re using, because they join the meeting as a separate participant (you’ll see them listed as an attendee, usually with a name like “Otter Notetaker”). You connect your calendar, and the bot dials in automatically, or you invite it manually by adding its email or joining link. Useful if your organisation doesn’t have Copilot or Zoom’s paid tier, but it does mean an extra “person” is visibly present on the call, which is exactly why you need to tell people beforehand.
Granola. This one’s different again. It’s an app for Mac and Windows that listens to your computer’s audio directly, so there’s no bot joining the meeting and nothing visible to other participants. That can feel less intrusive, but it also means the people you’re meeting with may not realise notes are being taken unless you tell them, so the same rule applies: say something.
Recorded calls on iPhone. If you’re on a recent version of iOS, the Phone app can transcribe recorded calls, and Notes can transcribe voice recordings you make directly. Handy for a quick one-to-one, less practical for a full video meeting with several people.
Pick whichever fits your situation. If your company already pays for Copilot or Zoom’s paid plan, use what’s built in, it’s the path of least resistance. If not, a bot like Otter or Fireflies is the quickest way in.
Try it yourself
Once the notes or transcript exist, most of these tools let you ask follow-up questions, either through a chat feature or by pasting the transcript into ChatGPT, Claude, Copilot or Gemini. Here are three prompts that each do something different.
Basic summary, if you’ve just got a raw transcript:
Here's a transcript from a meeting. Summarise it in plain English,
under 200 words, covering what was discussed and why it mattered.
Don't include filler or small talk.
[paste transcript]
Pulling out decisions and actions specifically, which is usually what you actually need:
From this meeting transcript, list:
1. Every decision that was made
2. Every action item, with who owns it and any date mentioned
3. Anything left unresolved that needs a follow-up
If a name or date wasn't clearly stated, say "unclear" rather than guessing.
[paste transcript]
Correcting names before you send notes on:
This transcript has some misheard names. Replace "Sarah" with "Zara",
"Mark" with "Marc", and "the Dublin office" with "the Cork office"
throughout. Keep everything else exactly as it is.
[paste transcript]
That last one matters more than it sounds. Transcription tools are decent but they’re not psychic, and names get mangled constantly.
Check the result
Don’t just trust the output and forward it on. Check it, properly, every time to start with.
Read the names first. Genuinely, this is where most errors live. “Sian” becomes “Shauna”, “Aoife” becomes “Eva”, that sort of thing. Fix it before it goes anywhere near a client.
Then check the decisions section against your own memory of the call. Did it capture the actual decision, or just the discussion around it? AI tools sometimes note that something was “discussed” when actually the room agreed and moved on. Worth a second look.
Check the action items have owners and dates, not just tasks floating in space. “Follow up with the client” is useless without a name attached. If the AI couldn’t work out who owns what, that’s a sign the conversation itself was vague, not necessarily that the tool failed.
And ask yourself what’s missing. Every summary tool drops something. Sometimes it’s a throwaway comment that turns out to matter later. Compare the notes against what you remember happened in the first ten minutes especially, that’s often where context gets lost because people are still settling in.
If you’ve got a transcript as well as a summary, skim it for anything that got cut that shouldn’t have.
If it doesn’t work
A few common problems and what to do about them.
The bot won’t join. Usually a permissions issue, either the meeting host needs to admit external participants, or your organisation’s settings block bots from joining calendar invites automatically. Check with whoever manages your Teams or Zoom account.
Transcription is garbled. Background noise, people talking over each other, or a poor internet connection all make this worse. There’s not much you can do mid-call beyond asking people to speak one at a time, which, let’s be honest, rarely happens.
Names keep coming out wrong. Some tools let you build a custom vocabulary or spell out names in advance. Otherwise, fix it afterwards with a prompt like the one above.
On privacy: think before you paste. Don’t put anything into a consumer AI chatbot that you wouldn’t be comfortable seeing outside the room, client financials, health information, legal disputes, HR complaints, anything under an NDA. If a meeting covers that kind of ground, check whether your company has an approved enterprise tool with proper data handling, rather than reaching for a free consumer app. And if you’re not sure whether something counts as sensitive, treat it as sensitive. Better to ask first than explain later.
Keep exploring
If you’ve got a recording rather than a live meeting, have a look at How to get AI to transcribe an audio recording. If what you actually need is the key points pulled from a long document rather than a call, there’s How to get AI to summarise a document. And if the actions from your notes need turning into an actual plan for the week, try How to get AI to organise my schedule.
Sources and review notes
- Zoom Support — AI Companion meeting summary: https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0065750
- Google Meet Help — “Take notes for me”: https://support.google.com/meet/answer/14754931
- Microsoft Support — Catch up on meetings with Copilot in Teams: https://support.microsoft.com/en-us/office/get-started-with-copilot-in-microsoft-teams-meetings-0bf9dd3c-96f7-44e2-8bb8-790bedf066b1 · Teams Premium: https://www.microsoft.com/en-gb/microsoft-teams/premium
- Fireflies guide: https://guide.fireflies.ai/ · Granola: https://www.granola.ai/ · Otter.ai help: https://help.otter.ai/hc/en-us
Review date: 15 September 2026 — every source above was opened and checked against the text on that date.