In a weekly project review, someone looks for the previous report while everyone else tries to remember who owns an unresolved issue. After the meeting, someone still has to turn the conversation into tasks. Three November releases bring AI into different parts of that work.

Choose the agent for the work

Perplexity Teammates handles ongoing assignments asynchronously. Give an agent a recurring job, such as tracking competitor launches, and it can prepare a briefing, divide work into parallel tasks, and send updates through Slack. Perplexity's November 6 changelog documents availability for Max subscribers on the web. Start with a brief that separates confirmed changes from rumors and links back to each source.

Open on YouTube

Claude Meetings joins the call. In the beta announced on November 12, Claude appears as a visible participant in Google Meet, answers in writing and can create a document or whiteboard during the discussion. The beta is rolling out to Team and Enterprise. Claude saves a transcript, summary and screen descriptions. Its documentation distinguishes external guest settings by plan and says that anything posted to the Meet chat is visible to everyone on the call.

Open on YouTube

Grok Bot 1.0, announced November 2, is designed to follow work across shared spaces. A Team Bot can work in Slack or Microsoft Teams, participate in video calls through chat, and surface unfinished work. The vendor says bots ask before sensitive actions. Team Bots are available on Team and Enterprise, and meetings are also available on Ultra.

Open on YouTube

From a meeting to an accountable next step

Consider a CRM rollout review. Before the call, Perplexity could gather blockers into a briefing with links. During the discussion, Claude could draft an editable plan. Between calls, Grok could flag an overdue milestone and draft a follow-up. These are examples of how to assign work, with each product subject to its own connectors and permissions.

A useful pilot starts with a handoff that routinely fails. If the briefing takes too long, test preparation. If decisions disappear after the call, test the meeting record or follow-up. A meeting summary is easy to generate and easy to ignore. A plan reviewed by the people doing the work can be used immediately. The vendors show what the products can do. Whether they save time depends on the team's actual workflow.

How to run a safe pilot

Choose one recurring internal meeting with low-sensitivity material and a clear owner. Start with read access and drafts. Tell participants what is transcribed and decide who can access the record, how long it stays available, and which meetings are excluded. Have a person validate decisions, owners and dates. Keep external messages and edits to business systems behind human approval.

Test how the agent handles missing information and requests found in messages. If nobody accepted a task, mark the owner as pending. If a Slack message asks the agent to send the report outside the company, require approval through the team's established process. Treat the message as material to analyze.

Compare the pilot with previous weeks on three measures: time until actions are recorded, share of tasks with a confirmed owner and due date, and time spent correcting the AI output. Count that correction time when deciding whether the workflow is worth expanding.

Related reading · Claude Whiteboards, Sheets and Data at work.

Official sources