Claude gains enterprise Office access—Excel, PowerPoint and Word available to businesses
- Anthropic opened Claude’s Microsoft 365 add-ins wider on May 7, making Excel, PowerPoint, and Word generally available while Outlook entered public beta. - The sharpest detail is cross-app memory: edit a number in Excel and Claude can update the linked chart, memo, and deck without re-briefing. - This pushes Claude from chat window to work surface — useful, but much more sensitive for IT, audit, and document-control teams.
Office software is where a lot of real work still happens. Models live in Excel, contracts get marked up in Word, decks get built in PowerPoint, and the inbox glues the whole mess together. The gap, until now, was that Claude mostly sat outside those files — useful, but separate. On May 7, Anthropic changed that by making Claude for Excel, PowerPoint, and Word generally available, while putting Claude for Outlook into public beta for paid plans. ### What actually launched? Anthropic’s new setup is a set of Microsoft 365 add-ins. They put Claude directly into Excel, Word, and PowerPoint through one AppSource listing, with Outlook handled through a separate beta listing. Anthropic says the generally available add-ins work across paid Claude plans, and admins can deploy them from the Microsoft admin center. (claude.com) ### Why is this different from “AI in Office”? The important bit is not just that Claude appears in the sidebar. It carries conversation context across apps. Anthropic’s own example is basically the whole white-collar workflow in miniature — start from an email in Outlook, open the attachment in Word, build the analysis in Excel, then turn it into a PowerPoint deck without re-explaining the task each time. (claude.com) ### What can Claude do in each app? In Excel, Claude can read multi-tab workbooks, build models with formulas, and track cell-level changes. In PowerPoint, it works inside an existing template, using native charts and the company’s layouts, fonts, and colors. In Word, edits land as tracked changes, which matters because legal and policy teams usually need a visible review trail, not silent rewrites. Outlook is more about inbox triage, draft replies, and scheduling. (claude.com) ### Why does cross-app memory matter so much? Because app-switching is where a lot of AI friction lives. Most assistants can help inside one box, but work rarely stays in one box. Anthropic says if you change an assumption in Excel, the related chart in PowerPoint and the number in a Word memo can update too, as long as those files are open. That turns Claude from a chatbot into something closer to a working layer across documents. (marketplace.microsoft.com) ### Why is finance showing up first? Because finance teams already hop constantly between spreadsheets, memos, decks, and email — and small mistakes travel fast. Anthropic spent the same week pushing ten finance-focused agent templates for things like pitchbooks, KYC screening, valuation review, and month-end close. It also said Claude Opus 4.7 leads Vals AI’s Finance Agent benchmark at 64.37%, which is the number it is using to argue this is more than a UI expansion. (claude.com) ### What’s the catch for companies? Access control. The Microsoft marketplace listing is blunt: the app can read and make changes to documents, and it can send data over the internet. That does not make it unsafe by itself, but it means procurement, security, and compliance teams will care a lot about where prompts go, what files Claude can touch, and how actions get logged. (anthropic.com) ### Did Anthropic build any enterprise controls around that? Yes — and that is probably why this launch matters more to businesses than to casual users. Anthropic says admins can stream prompts, tool calls, and document references through OpenTelemetry to their own collector, and the Analytics API breaks usage out by user, app, and day. Anthropic’s trust center also lists enterprise compliance programs including SOC 2 Type 2, ISO 27001, and ISO 42001 for Claude Enterprise. (marketplace.microsoft.com) ### So what changed in practical terms? Claude moved closer to the document itself. That sounds small, but it is the difference between asking an AI for help and letting it participate in the workflow where numbers, language, and approvals actually live. If this works well, Anthropic gets a real foothold in enterprise productivity. If it goes badly, the failure mode is not a weird chat answer — it is a weird edit in a spreadsheet, contract, or board deck. (claude.com)