Claude Cowork Goes Mobile: What Leaders Need to Know

Here is what I found most interesting about this release.

It follows you across devices.

Start a task at your desk, check on it from your phone between meetings, and pick up the finished output anywhere. Chat and Cowork now live in one place, so handing off a task is as easy as asking a question.

It keeps working in the background.

Close your laptop and Claude keeps going on Anthropic's servers. Your work no longer stops the moment you step away, and it now runs even when none of your devices are online.

You can schedule tasks to run on their own.

This is the one I am most excited about. Set a briefing to build every morning at 6am, and Claude works through the overnight emails, transcripts, and news, assembles the document, and leaves the follow-up drafted but unsent for you to review over coffee.

The big decisions still come to you.

When Claude reaches a call only you can make, it pauses and asks. You get a notification in the Claude app on your phone, so you can approve or redirect it from wherever you are. Nothing ships until you sign off.

Something interesting was published alongside this feature release. Anthropic analyzed 1.2 million Cowork sessions and found that coding made up only 8.7% of them.

Very interesting to see a tool most people assume is for engineers being used overwhelmingly for everyday knowledge work. The two biggest categories were business process and operations at 33.4% and content creation at 16.4%.

Together that is roughly half of all usage, and none of it is code.

Anthropic calls it "the work around the work," the connective tasks that never make it into anyone's job description but eat a huge share of the week. You can see more numbers like these in our running AI statistics guide.

So what does that actually look like, and how is it different from just prompting Claude?

With prompting, you ask, you get an answer, and you do the next step yourself. With Cowork, you delegate the whole job and it runs several sub-tasks at once across your connected tools. A few real examples:

• A weekly pipeline review kicked off from a single trigger: one sub-task pulls the data from HubSpot, another builds an interactive dashboard, another posts the summary to Slack.
• An expense report created by dropping a folder of receipts and asking Claude to turn them into a clean, formatted sheet.
• A renewals tracker built from a messy folder of contracts, with the risks flagged for you.
• Tomorrow's client deck assembled from call transcripts and recent pipeline data.

There is also a shift in tone worth noticing.

Anthropic, whose leadership has warned that AI could wipe out half of entry-level white-collar jobs, is now framing this around augmenting the daily grind rather than replacing the worker.

That is much closer to how we believe human-centric AI adoption should actually happen.

What the news is: Claude Cowork now runs on web and mobile. It works in the background, it can run scheduled tasks on its own, and the decisions that matter still route back to you for approval.

What is noteworthy: This is not a coding tool. Half of all usage is operations and content, the "work around the work" that quietly consumes your week. That is exactly where a busy leader can get time back.

How to try it: If you are on the Max plan, start a Cowork session from the home screen at claude.ai or from the Cowork tab in the Claude mobile app. Other plans are coming soon. The best first step is to point it at a real folder, thread, or half-finished deck and describe what "done" looks like, then let it run.

My take: The win here is not the tool, it is knowing what to hand off. The leaders who get the most out of Cowork are the ones with strong Workflow Literacy, who use AI to augment their work before automating it, and who treat every scheduled agent as a lift to their Impact per Hour. That judgment is the real skill, and it is exactly what we build in AI Leader Advanced.