RecodeAI Explore AI Intelligence

AI Agents & Productivity

Claude moves into the document, the spreadsheet and the meeting

Anthropic shipped a working surface rather than a chat window. Most of it is beta, some of it is off by default, and one piece turns itself on.

2 min read Editorially reviewed 22 NOV 2026

In brief

On 10 November Anthropic added Whiteboards, Sheets and Meetings to Claude, and made interactive connectors generally available. Whiteboards and Sheets are in beta on paid plans and start disabled on Enterprise; Claude Data switched on by default from 19 November. Meetings is limited to Google Meet. No post-beta pricing has been announced.

What happened

Anthropic introduced Claude Whiteboards, Claude Sheets and Claude Meetings on 10 November, alongside general availability for interactive connectors that open third-party applications inside Claude. Reported connector partners include Adobe, Atlassian and HubSpot; source lists differ beyond those.

Availability is uneven and worth reading carefully. Whiteboards and Sheets are in beta on paid plans across web, desktop and mobile. On Enterprise plans both start switched off, requiring an administrator to enable them — but Claude Data turned on by default from 19 November. Meetings is described as beta for Team and Enterprise customers starting with Google Meet; sources disagree on whether it has shipped or remains “coming soon”, and no Zoom or Microsoft Teams support has been announced. Anthropic has not said what any of these will cost once they leave beta.

Why it matters

The move is from assistant to workspace. A chat interface requires the person to carry context in and carry output back out. A whiteboard, a spreadsheet and a meeting transcript are where the context already lives — which changes what the model can see without being told.

That is also the governance question. A model invited into the meeting has access to what people say before it is minuted, and a model inside the spreadsheet has the working numbers rather than the published ones.

Business and industry implications

The default settings are the story for anyone running an enterprise tenant. A capability that starts off is a decision you get to make. A capability that turns itself on is a decision that has been made for you, and Claude Data's 19 November default is exactly that. Enterprises with data-residency, confidentiality or works-council obligations should audit current tenant state rather than assume it matches the state they configured.

Unpriced beta features are also a planning hazard. Teams build workflows on them during the free period and inherit the pricing later.

Leadership perspective

Decide deliberately which surfaces an AI system is invited into, and write the decision down. The useful question is not whether these tools are good — they appear to be — but which rooms your organisation is content for a model to sit in, and whether that answer was chosen or inherited from a default.

Three questions for business leaders

  1. Which of these surfaces are enabled in our tenant right now — and did we choose that, or did a default?
  2. Will AI become the primary interface to our enterprise information, and are we designing for that or resisting it?
  3. What is our exposure if a feature we have built on is priced differently after beta?