Claude Just Started Sending Emails and Building Spreadsheets. Here's What That Means

What actually happened

Anthropic expanded a feature called Cowork, which now works on the web and on mobile, not just desktop. Your files and chat history sync across all of them, so you can start something on your laptop and finish it on your phone.

The bigger news is what Cowork can now do. It's been connected to Microsoft 365, which means it can:

• Draft and send emails on your behalf
• Create or update calendar events
• Build and edit files in OneDrive and SharePoint, including Word docs and Excel sheets

Anthropic also rolled this into Claude for Government, with the same tools running in a more locked-down environment for public sector use. That's a signal they expect this kind of "AI that takes action" approach to stick around, not just a consumer gimmick.

Why you should care, even if you don't use Claude

Here's the pattern worth noticing: AI is moving from "tells you what to do" to "does it for you."

For years, AI tools have been really good research assistants and pretty mediocre employees. You'd ask ChatGPT or Claude to write an email, then you'd still have to open your email client, paste it in, check the recipient, and hit send.

That extra step matters more than it seems. It's the difference between an assistant and a really smart notepad.

When an AI tool can act directly inside the software you already use, meetings, docs, spreadsheets, and email, the amount of busywork it removes goes up dramatically. Not because the AI got smarter, but because the friction between "AI has an idea" and "task is done" disappeared.

If you run a business, manage a team, or just drown in email, this is the direction all the major AI tools are heading. Microsoft Copilot already works this way. Google's Gemini is moving the same direction inside Workspace. This isn't a Claude-only story. It's where the whole industry is going.

What changes tomorrow

Practically, not much changes overnight for most people. You'd need to actually turn on these connections and decide how much you trust an AI tool to act in your inbox and calendar. But here's what I'd start thinking about:

1. Which repetitive tasks are you still doing by hand? Scheduling. Meeting recaps. Weekly reports. Status updates. If a task involves reading something, deciding something small, and writing something back, that's exactly the kind of work this update targets.

2. Where do you actually want a human in the loop? Sending an email to a client is different from sending an email to yourself as a reminder. I'd start with low-stakes, reversible tasks, and keep review steps on anything client-facing or financial.

3. What's your approval process going to be? "Draft and let me approve" is a very different level of trust than "draft and send automatically." Most tools, Claude included, let you choose. Start conservative. Loosen up as you build confidence.

Is this hype, or is it actually important?

A bit of both, honestly.

The hype part: no, this doesn't mean AI is "running your business" now. It still needs clear instructions. It still makes mistakes. It still needs a human checking anything that matters. Anthropic itself has been public about building in extra safeguards specifically because giving AI the ability to act, not just talk, raises the stakes if something goes wrong.

The important part: the gap between "AI gives advice" and "AI does the work" is closing fast. That's the actual story here, not the specific feature list. Six months ago, most business AI tools were glorified chat windows. Now they're starting to look more like an employee you can hand real tasks to, with real oversight.

I don't think that makes AI magic. I think it makes it a lot more useful for people who've been AI-curious but never found the time to build a workflow around it. The workflow is starting to build itself.

Where I'd start, if I were you

You don't need to hand your entire inbox to an AI tool tomorrow. A reasonable first step:

• Pick one recurring, low-stakes task: meeting notes, a weekly internal update, or a first draft of a follow-up email.
• Let an AI tool draft it.
• Review before it goes anywhere.
• After a couple weeks, decide if you trust it enough to skip the review step for that specific task.