Anthropic adds three office tools to Claude for working with data

Discuss a product launch with AI, copy the calculations into a spreadsheet, transfer the decisions to a working document. At every handoff, you have to explain the task again and check that nothing has been lost. Anthropic is trying to eliminate that friction: on November 10, the company introduced three tools inside Claude — Sheets, Whiteboards and Data. For a product team, this is a reason to test how many manual transfers can be removed from a single workflow.

From discussion to working document

Sheets is designed for collaborative work with spreadsheets, formulas and multiple tabs. Whiteboards gives teams a board for discussing ideas that Claude can then summarize. Data, previously called Dashboards, works with connected data sources and helps create dashboards, visual explanations and short videos with voiceovers. The company describes this combination in its announcement of the new features.

Imagine a small online service preparing a new pricing plan. The product manager gathers hypotheses on a whiteboard: who needs the plan, which features to include and what to test before launch. They then ask Claude to move the selected options into a spreadsheet and calculate several scenarios. After launch, the team wants to compare its expectations with the actual results.

This is an example use case, not a published test result. Its purpose is to show where savings might emerge: fewer repeated explanations between stages. Both the formulas and the underlying assumptions will still need checking. A neatly assembled spreadsheet can easily make a solution feel complete, even when an error lies in the very first assumption about user behavior.

The app opens alongside the conversation

The second part of the update concerns services the team already uses. A supported tool can open alongside the chat or expand to full screen. For Notion and Confluence, Anthropic describes a full editor running under the user's account. Claude can modify the open page, row or file with a person's approval. Availability depends on the specific integration and the rollout of the update. Details are available in the description of interactive connectors.

For example, after discussing requirements, the product manager opens a document and asks Claude to make the agreed changes. The document is visible in the same window: they can read the edit, adjust it manually and continue the conversation. The practical value depends on how well the assistant understands the selected item and how easy it is to review its actions.

Developers should take a look at MCP Apps, an MCP extension through which a server provides its own interface inside an AI application. Anthropic says such an app can also work in other clients that support the extension. The expanded capabilities are available to developers in public preview.

For a SaaS founder, this raises a separate product question: what complete operation would a customer want to perform from a conversation with an assistant? A good candidate is an action with clear input, a visible result and a straightforward way to review it. Updating a specific document, for example. Simply displaying the entire interface inside another window is not enough to make the work easier.

What to check before moving workflows

The original post presents the launch as a ready-made replacement for an entire set of work services. The confirmed capabilities paint a more specific picture. In particular, the official announcement says Meetings will arrive later. It is premature to expect Claude to join the next team call on the basis of this message.

For a pilot, choose one process the team repeats every week: preparing a project status update, recalculating scenarios or updating requirements. Record how long the work currently takes, then measure the time with Claude, including reviews and corrections. Check access permissions separately: the convenience of a shared window does not remove the need to understand which data and actions the assistant can access.

My conclusion: the update deserves attention primarily as an attempt to connect discussion with execution. There are no grounds yet to declare the end of analysts or specialized services. An adoption decision is best made after a small pilot that shows which manual steps disappeared and how much work reviewing the result added.