Claude Meetings joins the call, and leaves when a stranger walks in
Anthropic put Claude into Google Meet on Thursday as a visible participant: https://claude.com/resources/articles/claude-meetings. It is a beta for Team and Enterprise plans, Google Meet only for now. Claude answers in writing by default, builds docs and whiteboards live while people edit alongside it, writes notes and action items, and leaves a page for each meeting in claude.ai. It saves the transcript, a summary and descriptions of shared screens. It does not save audio.
The interesting part is how it decides whose orders to follow. Claude uses each person's connected tools with that person's own permissions, and asks the first time. When someone gives it a spoken request, it confirms who is speaking before acting, every time. And by default it stays out of calls that include outside guests, and leaves if one joins. Admins can override that; then it posts a notice and stops replying in chat, though guest speech still lands in the notes. That is a real answer to a question most meeting bots skip: in a room of six people, who is the agent working for? Orgs on zero data retention, HIPAA-ready or customer-managed keys can't use it during the beta. Help page: https://support.claude.com/en/articles/18325154
The same day Anthropic shipped Claude for Manufacturing (https://claude.com/resources/articles/claude-for-manufacturing): 17 skills such as quoting a job, investigating a failed inspection, diagnosing a breakdown or re-sourcing a supplier, plus connectors for Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. The notable piece is Infor's own plugin, which can make changes such as releasing a purchase order or rescheduling production, limited by the user's existing Infor permissions. Anthropic cites a 2025 survey that 79% of manufacturers still enter data by hand. Claude doesn't operate equipment, and each skill ends at a human review.
Both launches answer the same question from two sides. In the meeting, authority follows the speaker. In the ERP, it follows the account. Neither trusts the agent to work it out on its own, which is the right instinct. Disclosure: clauday is written with Claude.
← Back to all articles
The interesting part is how it decides whose orders to follow. Claude uses each person's connected tools with that person's own permissions, and asks the first time. When someone gives it a spoken request, it confirms who is speaking before acting, every time. And by default it stays out of calls that include outside guests, and leaves if one joins. Admins can override that; then it posts a notice and stops replying in chat, though guest speech still lands in the notes. That is a real answer to a question most meeting bots skip: in a room of six people, who is the agent working for? Orgs on zero data retention, HIPAA-ready or customer-managed keys can't use it during the beta. Help page: https://support.claude.com/en/articles/18325154
The same day Anthropic shipped Claude for Manufacturing (https://claude.com/resources/articles/claude-for-manufacturing): 17 skills such as quoting a job, investigating a failed inspection, diagnosing a breakdown or re-sourcing a supplier, plus connectors for Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. The notable piece is Infor's own plugin, which can make changes such as releasing a purchase order or rescheduling production, limited by the user's existing Infor permissions. Anthropic cites a 2025 survey that 79% of manufacturers still enter data by hand. Claude doesn't operate equipment, and each skill ends at a human review.
Both launches answer the same question from two sides. In the meeting, authority follows the speaker. In the ERP, it follows the account. Neither trusts the agent to work it out on its own, which is the right instinct. Disclosure: clauday is written with Claude.
Comments