On November 10, Anthropic announced Claude Meetings, an assistant designed to join a video call and begin drafting the work being discussed. In the preview described by Axios and TechCrunch, Claude can answer questions in Google Meet chat and build plans or timelines while people talk. A team could leave the call with a draft already taking shape, even as decisions about its contents remain open.
A shared plan gives a discussion consequences beyond the people who attended. A proposed launch date can become the date a colleague uses to book staff. An action item can look settled even when nobody has accepted it. Claude’s ability to put that draft in front of the group creates an opportunity to catch those gaps while the people who can resolve them are still together. The review can happen before the plan reaches colleagues who missed the discussion.
Axios and TechCrunch described a beta for Claude Team and Enterprise customers. Anthropic’s own announcement still calls Meetings “coming soon.” Its current Whiteboards help page explains the shared canvas for diagrams and plans but gives no instructions for using it in Meetings. The rollout timetable and the rates or limits for using Meetings remain unclear.
What the room can see
Nicholas Lin, Anthropic’s product lead for knowledge work, told Axios reporter Ina Fried that the company had been testing Meetings internally for about two months. He said Claude could retrieve information people needed, make materials as they talked and track follow-up. Lin said the assistant mostly waits for instructions today. His expectation that it will act more proactively describes a direction for the product.
The whiteboard is a concrete example. Anthropic’s announcement describes a shared canvas where teams can organize ideas with Claude. Its help page explains how collaborators can rearrange a board and turn it into a document or slide deck. The meeting preview extends that idea to the conversation happening on a call. TechCrunch’s Aisha Malik reports that after the call, the notes, action items and materials Claude made are available in Claude. A person who already has the board open could point out that a launch date depends on a supplier’s confirmation while everyone still remembers the exchange. An absent colleague may only see the version they are sent.
That is a change from reading a tidy recap at the end of the day. The value depends on showing the draft at a point when people can still do something about it. In a planning meeting the group could see that a proposed schedule requires two teams on the same morning and adjust it immediately. If everybody is looking at the participants’ faces and no one opens the board, the same draft might simply be another thing to check after the call.
Google Meet already offers automatic notes on eligible plans. Its native feature can show a summary during a call, put the recap in Google Docs and send a link when the meeting ends. Google lets hosts choose whether notes go to hosts only, invited people in the organization or all invitees. And Otter announced a voice-activated assistant that could field questions and help with tasks in 2025. The useful comparison for Claude is with the work a team is already doing. A group that only wants to catch up may be well served by a recap. A group that will spend another hour building a timeline has something to gain from making a first version together.
The difference between a date and a promise
Consider a project meeting about a supplier. Maya offers to get a quote by Thursday. Someone else floats Friday as a launch date if the quote arrives and the price is acceptable. A third person asks for a draft customer email so the team can move quickly if everything comes through. Those are three different kinds of follow-up. All three could look like tasks on a polished board.
| What happened in the room | What the team needs in its plan |
|---|---|
| Maya volunteered to request a quote by Thursday | An accepted task with Maya as owner |
| Friday was proposed if the supplier confirms and the price is acceptable | A provisional date with both conditions attached |
| Someone asked for a draft customer email | A draft that stays internal until the launch is confirmed |
On a timeline, a tentative date gets the same neat box as a deadline. Without the condition next to it, a colleague opening the board the next day may reasonably start booking staff. A blank space for an owner can also be more accurate than a plausible name inferred from who happened to be talking. A good board keeps the gaps visible so the people in the room can decide which ones to close.
The same tool can make an unresolved question easier to settle. Put the quote, staffing and customer email on the same board and the team might see that staffing has to be booked before the supplier can respond. The board is doing useful work before the group has decided anything. The human judgment comes in when someone says which date to move and who has authority to accept the change.
There is a strong case for starting certain work while a decision is pending. Maya can sketch the email using a placeholder for the date, and the team can check that the wording works in either case. That saves a real step. If the system or a colleague treats that same draft as ready to send, it creates a promise the group has yet to make. A review just before a draft changes hands matters more than polishing a transcript afterward.
Who gets to decide what leaves the call
To make the plan, the assistant has to follow the conversation. That makes its presence relevant to everyone who speaks, including someone who never uses Claude. TechCrunch reported that Claude identifies itself as a participant, posts a notice in chat and is restricted from meetings with outside guests unless an administrator permits it. The restriction gives administrators some control over which calls Claude can attend. A guest’s presence or absence tells you little about whether coworkers are comfortable having a sensitive discussion captured.
In July, Cathy Bussewitz of the Associated Press reported on how professionals handle those boundaries with AI note takers. One practical suggestion was to use a note taker for part of a meeting and switch it off for more sensitive discussion. Google’s own controls illustrate a different distinction. An administrator can require participants to explicitly agree to certain native features such as recording, transcription or Google’s note taking. That documentation concerns Google’s tools. The reported notice from Claude is a different mechanism.
After the call, coworkers may want to know who can open the notes or board and how long those records, or the information used to make them, will be kept. Anthropic says it generally excludes commercial customer data from model training, with an exception for customers who opt into its Development Partner Program. That policy addresses training. How long Meetings keeps any call audio or transcripts is a separate question. So is who ends up with a copy of a plan that a colleague takes into another app. Those questions can matter even in a meeting where everyone works for the same organization.
The distinction also affects what people are willing to say. In the supplier meeting, a teammate might want to flag a concern about the relationship without circulating it alongside the launch plan. A written record can help someone who could not attend, while a more widely shared draft can make a passing observation travel further than the speaker expected. The useful amount of detail depends on who needs to act on it.
What counts as time saved
In a public discussion about the announcement, one Reddit commenter saw a chance to finish follow-up before the workday ends. Another liked action-item tracking but wanted participants to know when they were being recorded or analyzed and to have people review the results. Their reactions point to two ways the work could shift. The worker who used to assemble notes after dinner might get time back. The worker whose name appears next to a task they never accepted has a new job correcting it.
Irwin Lazar, an analyst interviewed by Computerworld, pointed to the benefit of avoiding the step of copying meeting context into another tool. That can save effort, especially when the next task depends on details people would otherwise have to reconstruct. For a team that already edits a common plan during calls, the gain may be much smaller. If each person opens a different copy later, the team can be worse off even when the first draft took seconds.
The work continues after a board appears. The time spent checking a date or asking who owns a task belongs in the same calculation as the time saved typing the plan. Lin’s account of internal use shows that Anthropic has been bringing Claude into its own calls. How much time the tool saves once teams account for corrections remains open.
The Friday timeline earns its place if Maya leaves knowing what she agreed to do, and the person who missed the meeting knows which date is still provisional. If they can settle that while everyone is looking at the same draft, they can spare the next morning’s round of messages asking whether the launch is on.