Skip to content
IFAI

AI tools

Claude can now build and edit actual files

Published Reach 42

Anthropic just gave Claude the ability to create and edit Excel spreadsheets, documents, PowerPoint slide decks, and PDFs directly in-app. Instead of returning plain text, Claude runs code in a private sandboxed environment to tackle multi-step analysis and data formatting. The feature is now rolled out to paid plans.

Image: anthropic.com

For the longest time, asking an AI to build a spreadsheet meant copying messy markdown tables into Excel, fixing the broken columns, and manually rebuilding every formula.

Anthropic is finally fixing that handoff. Claude can now directly create and edit real files, including Excel spreadsheets, PowerPoint slide decks, formatted documents, and PDFs across both Claude.ai and its desktop app.

The feature initially rolled out as a preview for Max, Team, and Enterprise tiers before expanding to Pro subscribers, followed by a full release on October 21 with dedicated network and egress controls.

Giving Claude its own sandbox

Instead of just spitting out text responses or in-app artifacts, Claude now operates inside a private computer environment behind the scenes.

When you describe what you need, Claude writes and executes code in the background to build the file from scratch, run calculations, and format charts just like a junior analyst would.

That sandbox setup makes complex cross-format workflows surprisingly smooth. You can hand Claude a dense PDF report and ask for presentation slides, feed it a pile of raw invoices to get an organized multi-sheet spreadsheet with working formulas, or upload meeting transcripts for a clean brief.

Once Claude finishes crunching the data, you can download the final files directly or save them straight to Google Drive.

To try it out, eligible users have to manually toggle on Upgraded file creation and analysis under Settings > Features > Experimental. Anthropic does note that enabling this gives Claude internet access to fetch and analyze data, urging users to keep a close eye on chat sessions.

Why it matters

We think this marks a critical transition from AI that advises to AI that actually produces finished work.

For knowledge workers, consultants, and founders, the value is immediate. Getting a working financial model with real formulas instead of text you have to reconstruct saves hours of tedious grunt work, making this an easy recommendation to test right away.

Strategically, Anthropic is taking a direct swing at Microsoft 365 Copilot and Google Workspace. Instead of relying on rigid enterprise office integrations, Claude uses raw Python code in an isolated sandbox to spit out native file formats, delivering a far more flexible workflow without the corporate bloat.

The real test will be security. Granting an autonomous model an active code-execution environment with web connectivity opens the door to data egress risks and prompt injection bugs. Anthropic adding outbound traffic guardrails in October is reassuring, but enterprise IT teams will likely remain cautious before letting employees pipe sensitive quarterly ledgers through it.

Sources (1)

  1. [1]
    anthropic.com(opens in a new tab)Originalanthropic.com

Best of AI, daily