02

What Copilot and AI can actually do for your meetings right now

A practical guide — including the parts that don't work yet.

There are two kinds of people talking about AI and meetings at the moment.

The first kind says AI will fix meetings entirely. Fewer attendees needed, better notes, smarter follow-up, lower cost per minute. Everything sorted.

The second kind says AI will make meetings worse. People will check out while the bot takes notes. Nobody will pay attention. The meeting becomes theatre and the AI becomes the only real participant.

Both of these are partially right and mostly incomplete. So here is what AI — specifically Microsoft Copilot and Claude's Cowork mode — can actually do for meetings right now, along with an honest account of the limits.

What it does

If your organisation has a Microsoft 365 Copilot licence, Copilot is available inside Teams meetings. Here is what it does in practice.

Microsoft 365 Copilot · Teams
During and after the meeting
  • Intelligent meeting recap. After a meeting ends, Copilot produces a structured summary: what was discussed, what was decided, what the action points are, and who owns them. Accuracy is good on factual content and variable on intent and implication.
  • Real-time questions. During a meeting you can ask Copilot "what has been agreed so far?" or "has anyone mentioned the budget?" It searches the transcript and returns a summary. Useful if you join late or if a long meeting has drifted.
  • Follow-up drafting. Copilot can draft a follow-up email from the meeting summary. A decent starting point — always edit it. The draft tends to be complete but flat: accurate without being useful.
  • Catch me up. If you join late or miss a meeting entirely, Copilot can summarise what happened before you arrived. This one genuinely saves time.
"What Copilot does not do: it cannot tell you whether a meeting was necessary. It cannot tell you whether the decision made was the right one. It cannot replace the judgment of the people in the room."
The honest limit of AI note-taking

What it adds — before the meeting

Where Copilot is strong during and after a meeting, Claude's Cowork mode is most useful before one.

Claude · Cowork Mode
Before the meeting
  • Pre-meeting briefing. Pull together background on the people attending, the topic being discussed, any relevant documents, or context from earlier conversations. Means you walk in prepared rather than relying on a vague memory of the last meeting.
  • Agenda preparation. Give Cowork a topic and a list of attendees and it will draft a structured agenda with time allocations and the question each item is trying to answer. Forcing each agenda item to be a question — not a topic — is more useful than it sounds.
  • Note synthesis. If you take rough notes during a meeting — even badly formatted ones — Cowork can structure them into a clear summary, extract decisions, and draft follow-up tasks. Takes under a minute.
  • Research connected to your documents. Because Cowork has access to your project folders, it can connect what happened in a meeting to what exists in your files. The continuity from meeting to document to action is where most meeting value gets lost.

What AI cannot fix

Things no AI meeting tool solves

  • AI note-taking does not make a bad meeting better. It makes the record of a bad meeting crisper.
  • Disengagement risk is real. If someone knows a summary is coming, the incentive to pay attention drops. The meeting becomes something to have survived rather than participated in. Nobody has fully solved this yet.
  • Governance is unsettled. Copilot requires a transcript, which means meetings need to be recorded. Not everyone is comfortable with this. Who sees the transcript, how long is it kept, what happens if someone says something indiscreet — most organisations haven't answered these questions.
  • Trust cannot be automated. Neither Copilot nor Cowork can build trust between people who have never met. They cannot produce the accidental connection that happens when two people talk about something that wasn't on the agenda. They cannot replace the moment when someone reads the room and changes what they were about to say.

What to try this week

1
If you have Copilot in Teams: use the intelligent meeting recap on your next three meetings. Read the summary before you write your own notes. See how much of what you thought mattered made it in — and what didn't.
2
If you use Claude or another AI assistant: write your next meeting agenda as questions, not topics. Use the AI to draft it. Then cut anything for which you cannot articulate the decision you are trying to reach.
3
Both experiments take less than 30 minutes. They will tell you more than any vendor demo.

The goal is not to make meetings more efficient. Efficiency is not the problem with most meetings. The problem is clarity: what is this for, who needs to be here, what are we deciding? AI can help with the scaffolding. The judgment is still yours.


← Prev: The problem Next: The 2030 meeting →