Claude Cowork acts across your files now. Confidence is not correctness.

TL;DR

• Chat makes you drive every step. Cowork flips it: you describe the outcome, it plans the steps, works across your files and connected apps, and hands back the finished thing.
• Anthropic says more than 90% of Claude Cowork usage is not software development. It is expense memos, contract trackers, client decks, folder cleanup. This week it went to web and mobile, and scheduled tasks run with no device online.
• That is a governance surface, not just a convenience. It acts across whatever you connect, and it is confident when it is wrong. Scope the access, keep the review.
• Start with one task you already resent, give it context first, and check its work like a code review.

This piece is from Code Meet AI, my newsletter where i document how i actually use AI to ship real products, receipts included. New guides land there first.

This Monday Anthropic put a number on something most of us felt but hadn't measured. They expanded Claude Cowork to web and mobile and shared what people actually run on it: more than 90% of usage is not software development. Not code. The biggest categories were business operations and content work. Reconciling the quarter's spend and drafting the memo about it. Turning a folder of contracts into a renewals tracker with the risks flagged. Building tomorrow's client deck from call transcripts.

They call it "the work around work." Good name. It is the layer of the job that is real work, takes real hours, and somehow is nobody's actual job.

If you write code, you have had this delegation engine for a year in the terminal. Cowork is that same engine pointed at everyone else's work. The interesting part for us is not the convenience. It is that the tool now acts, on its own, across whatever you connect it to. That changes what "review" means.

what Cowork is, in one honest paragraph

You already know chat: you ask, it answers, you ask again. You are the project manager of every tiny step.

Cowork is the same Claude with a different job description. You hand it an outcome: "turn this folder of contracts into a renewals tracker, flag anything risky." It makes a plan, breaks it into steps, works across the files and apps you have connected, and comes back with the finished thing. While it runs, you do something else. Since this week it follows you: start a task at your desk, check the status on your phone, pick up the result later with the laptop closed. Scheduled tasks run even with no device online.

Chat: you drive. Cowork: you describe, then review.

my actual Claude home. one toggle flips chat into Cowork, and "send me a daily briefing" is already sitting there as a suggestion.

One tool, two jobs, and the second job is most people's actual day.

what people actually run on it

Not moonshots. Chores. The patterns showing up again and again in the early write-ups:

Folder archaeology. Sort, rename, and deduplicate a cluttered folder. Rename 400 photos to match the product list. The task postponed for a year because it is three hours of mind-numbing clicking.

Pile of documents, should be a list. A folder of contracts becomes a renewals tracker with dates and risk flags. A stack of receipts becomes an expense sheet. Dense PDFs in, structured table out.

The recurring report. The weekly status, the morning brief, the numbers pulled into the same format every Monday. Described once as a routine, in plain language: what to gather, what format, where to deliver it. It runs every morning without you. Cron, described in English, with file and app access.

First drafts from your own material. Tomorrow's client deck from the call transcripts. The variance memo from the spreadsheet. Not "write me something generic," but "turn my raw material into my draft."

The common shape: high effort, repeatable, source material already in your files.

the part that should get your attention

Here is where the senior-dev instinct earns its keep. Two things about this tool are load-bearing, and both are easy to wave past.

It has access to what you give it. That is the feature and the risk in one sentence. Connect the folders and tools the job needs, not everything you own. This is least privilege, the same principle you apply to a service token or an OAuth scope, except the person setting it up may never have thought about blast radius in their life. A general-purpose agent with read and write access to your whole drive is a very different thing from one pointed at a single project folder.

And it is confident, including when it is wrong. A renewals tracker with one silently wrong date is worse than no tracker, because someone will trust it. You already know this shape from AI-generated code that compiles clean and still ships a bug. The difference is that a non-engineer has no compiler, no test suite, no diff habit to catch it. The review step is not optional politeness. It is the entire job that does not get automated away.

So the honest framing is not "AI does your admin now." It is "AI does your admin now, and it needs the same governance you already give your code: scoped access, and a review before you trust the output."

how to start: one resented task

Don't tour the features. Pick the one task you flinched at earlier. Then:

1. Give it context first. Before the task, point Cowork at a short context file: what this project is, what good output looks like, what to leave alone. If you have given an AI a memory folder before, this is the same muscle. Two minutes, and the output quality jump is not subtle.
2. Describe the outcome, not the steps. "Make this folder make sense: group by client, newest first, flag anything duplicated" beats a 12-message back-and-forth.
3. Review like a manager, not a believer. Open the deliverable, spot-check five rows against the source. A code review with different file types.
4. If it worked twice, make it a routine, and scope it tight. Describe it once, give it access to exactly the folder it needs, and stop doing it forever.

the punchline

The AI didn't take the work. It took the work around the work. What's left over is the part that needed you all along: deciding, and checking. For us that is not new. We just called it code review and least privilege. The rest of the world is about to learn why we were so annoying about both.

I am putting together a routines starter pack: three plain-language routine templates (the folder cleanup, the pile-to-tracker, the morning brief) you can adapt in five minutes. It goes out through the newsletter first.

Written by Malik at Code Meet AI. This is episode 03 of Using AI, properly. Episode 02 was give the AI a memory, episode 01 was the art of prompting. New guides land there first.

i'm Malik, i've been a mobile engineer for 9 years, and i run my work and this newsletter on the system these episodes describe. The delegation habit started in the terminal, moved to my second brain, and now it does the work around the work. Next up: where a single sentence stops helping, and context begins.