Claude Meetings puts an AI in the call as a visible participant, and its consent rules are the point

TechCrunch reports that Claude Meetings lets teams add Claude to a video call as a visible participant. It listens, drafts the plan being discussed, maps timelines on a whiteboard, answers questions in the meeting chat and fetches information from connected work apps. When the call ends, notes, action items and anything it produced live in claude.ai. It is in beta for Team and Enterprise customers and works with Google Meet first.

It arrived alongside Claude Whiteboards and Claude Sheets, general availability of Claude Data and interactive connectors for apps like Adobe, Atlassian and HubSpot (our analysis of the full office push). TechCrunch notes the launch came weeks after OpenAI’s own workplace tools, with Meta’s Muse agent chasing the same territory.

Meeting transcription is old news. An AI that acts on a meeting while it is happening is different, and it changes what a meeting is for. The most common complaint about meetings is that nothing happens until everyone hangs up, and then it happens slowly or not at all. A participant that drafts the plan as people talk, and pulls up last year’s numbers when someone asks, attacks that problem directly. If it works, some meetings get shorter and more of them end with an actual document.

Anthropic has its own data point on why this matters. Its study of its AI-heavy workplace found people spending more time in meetings (our coverage). A tool that makes those meetings produce something is a logical next step.

Claude identifies itself, posts a notice in the chat and, by default, leaves when someone from outside the company joins. It will not join calls with external guests unless an admin explicitly permits it. That is the right default. The worst version of this product is a silent bot recording clients, candidates or partners who never agreed. Anthropic built the default around the people in the room who did not sign up for the software, and that should become an industry floor rather than a selling point. It echoes the approval-first design of Anthropic’s Duet agent (our Duet launch coverage).

An AI participant with access to connected apps is a new route for sensitive data, and meetings are where people say things they would never write down. Everything Claude produces is stored, which makes meetings more searchable and more discoverable. Legal and HR teams will notice. There is also a subtler effect: when a machine drafts the plan in real time, the first draft tends to win. Teams should make sure the AI’s version of a decision does not quietly become the decision.

Longer term, meetings become an input stream for agents. That can be a real productivity gain, and it also turns a company’s internal conversation into structured data held by a model provider. The admin controls are a good start. The test is whether they survive pressure to make the product more autonomous. See more Anthropic coverage and our enterprise AI coverage.