SHEYU.AI舍予基业
AI Model Watch

Claude moves spreadsheets and whiteboards into the chat box, and the real cost isn't Office

Anthropic's new Claude Sheets, Whiteboards and Data tools show work shifting from software icons to the chat box, and firms must decide who holds their business knowledge.

·4 min read
Claude moves spreadsheets and whiteboards into the chat box, and the real cost isn't Office
AI-generated illustration, not a news photograph

Anthropic this week pushed Claude's office functions a step further, launching Claude Sheets, Claude Whiteboards and an upgraded Claude Data, while previewing Claude Meetings, which can join video calls. After the news, Microsoft's share price fell more than 3% intraday and closed down 2.2%, according to East Money Information (东方财富网), which reported the move took place on Tuesday local time.

Our view is that this is not a story about AI replacing Office. It is a story about the entry point to work moving from software icons to the chat box. For small and medium-sized business owners, the account that really needs calculating is not whether to save on a subscription, but who will hold your business knowledge, customer questions and content assets next, and in what form AI will call on them.

What Claude actually did this time

According to reports by Phoenix News (凤凰网) and East Money Information, Claude Sheets is a spreadsheet saved inside Claude Artifacts. It supports multiple tabs, cross-sheet formulas, conditional formatting, drop-down menus and charts. Team members can edit the same cells in real time, and Claude can keep making changes, outlining each edit.

But it is not a full replacement for Excel. The official positioning is a lightweight, shared spreadsheet, capped at 100,000 cells per sheet. It does not yet support pivot tables, macros, VBA, Power Query, DAX or live data connections, and mobile access is basic and read-only. Anthropic product lead Nicholas Lin told media that Sheets is currently "quite basic" and not designed for complex financial modelling.

Whiteboards targets team discussion: put ideas on a shared board, then let Claude summarise them. Claude Data, formerly Claude Dashboards, can connect to tools such as Snowflake and Salesforce to generate dashboards, visualisations or short narrated videos. Meetings is still at the preview stage and not open.

What enterprise users should care about

First, the entry point has changed. Employees used to open Excel to do the work; now they can go from discussion to building a sheet to revising it inside one conversational interface. Anthropic itself says Claude's applications are expanding from programmers to non-technical teams such as legal, finance and product management.

Second, knowledge is starting to settle inside conversations. According to East Money Information, Anthropic's legal team built a legal review tool inside Slack using Claude, which automatically flags the places that need changing according to existing review rules. Review time for marketing materials fell from three days to 24 hours. The key to this case is not speed, but that the review rules were written into the tool and can be reused.

Third, the boundary is clear. Complex financial models, production data pipelines and strictly regulated data should not, at this stage, be moved into Claude Sheets. The official guidance itself points to separate professional plug-ins for those needs.

A common counter-argument

Some will say these features are still so basic that it is better to wait until they mature, and that moving now is just paying tuition.

That argument is half right. The tools themselves are indeed still early, and rushing to move core finance into them is unwise. But it misses one thing: on the day the tools mature, have you organised your business rules, customer questions and brand facts into a form AI can use? In the Anthropic legal team case, the valuable part was not Claude. It was that they had first made their own review rules clear.

Under what conditions does this judgment not hold? If your business depends on a complex Excel model to make a living, or your data is heavily regulated, then you should indeed hold still for now and wait for the official professional plug-in route to stabilise.

One thing you can do now

Write down the twenty questions your business is asked most often, the few pieces of copy you repeat most, and the few rules that most need a unified wording. You do not need to buy anything. First get them out of people's heads and into text.

This has nothing to do with Claude and nothing to do with any vendor. But it is the precondition for whether any AI tool can later catch your business.

If you are already using content to acquire customers, SHEYU AGENT (舍予AI智能体) can take one part of this: it produces drafts based on brand facts and flagship products, files content under the corresponding topic, and can generate WeChat official account drafts. GEO monitoring watches Doubao, Wenxin, Kimi, Tongyi, DeepSeek and Yuanbao to see whether you are mentioned when customers ask related questions. What it cannot do is decide what your brand facts are, and it cannot guarantee that a platform will index or cite you; those still need separate verification. Tools can amplify what you have organised. They cannot amplify what you have not thought through.

What this round of competition really changes is not which software gets replaced, but whether companies are willing to write down what is in their heads and hand it to AI.

The 2.2% drop in Microsoft is the capital market's immediate reaction to the migration of the entry point. For an owner, more useful than watching the share price is spending an hour tonight writing out those twenty questions.

閱讀繁體中文版 →