Claude Meetings puts an AI colleague at the table. Not with clients

Here's the awkward moment every team knows. The most important call of the week is the one with the client, and that's exactly where a note-taking bot shows up in the participant list and somebody asks who invited the robot. Anthropic made a clear call on that question for its new meeting feature. The call is more interesting than the feature.

What's new

• Claude Meetings: According to TechCrunch, a beta went live on November 10, 2026 for Team and Enterprise plans, working only with Google Meet at launch. More plans are expected later. Anthropic's own write-up still calls the feature "coming soon" without a date, so expect a gradual rollout.
• On the call: Claude joins as a visible participant, announces itself in the meeting chat, follows the discussion, drafts the plan being talked through, maps a timeline on a whiteboard and pulls information from connected work apps.
• Afterwards: notes, action items and anything Claude created live in claude.ai.
• The rule: Claude won't join calls with external guests unless an admin allows it, and it leaves automatically when an external guest joins.

The positioning is a team room, not a client room

To me the guest rule isn't fine print. It's the product decision. Anthropic is building Claude Meetings for internal planning: sprints, campaign kickoffs, budget rounds. There, a visible AI participant that builds the timeline while people talk is real leverage. Nobody types up minutes, and the plan already exists when everyone hangs up.

On a client call, the visible bot is the problem. That's why my tool page on Granola leads with the big difference from the status quo: no bot sits in the call. Claude Meetings goes the opposite way and says it out loud: I'm here. That's honest. It also means that for sales, agency pitches and client calls, this release changes very little for now.

The catch: who decides what makes the plan?

A plan drafted live in the meeting quickly feels like an approved plan. It isn't. My rule of thumb: judgment, context and accountability stay in-house, meaning priorities, budget and final sign-off. Claude can write the draft. Who gets to say no still has to be settled before the call, or the whiteboard ends up making the decision.

Then there's the platform question. Google Meet only at launch. Teams and Zoom shops are out, and per TechCrunch Anthropic hasn't given a timeline for other platforms.

What I'd do now

• → Split your meetings into two lists: internal (Claude candidate) and with outsiders (bot stays out). That's your use case, cleanly separated.
• → If you run Google Meet and Claude Team: pick one recurring internal planning meeting and check for four weeks whether the draft actually gets used.
• → Decide up front who signs off the AI draft. One person, one line in the calendar invite.

Bottom line: Claude Meetings is a colleague for the team meeting, not the client call, and that boundary is exactly what makes it credible.