Technology

Anthropic's New AI Assistant Cowork Can Now Work Across Your Devices

Martin HollowayPublished 7d ago4 min readBased on 1 source
Reading level
Anthropic's New AI Assistant Cowork Can Now Work Across Your Devices

Anthropic's New AI Assistant Cowork Can Now Work Across Your Devices

Anthropic released Claude Cowork on web and mobile for Max subscribers on July 7, 2026, making the AI agent available beyond just desktop computers. The tool originally launched as a desktop-only app in January TechCrunch.

This matters because it changes how the tool works. You can now start a task on your computer, check on its progress from your phone later, and get the results without leaving your original machine running. Think of it like asking someone to run errands for you — they no longer need to stay connected to you the whole time. They can work in the background and report back when you check in.

Before this update, Cowork only lived on your desktop. Now that it runs across multiple devices through the cloud, it can keep working even after you close your laptop or turn off your phone.

How People Are Actually Using It

Anthropic published data on how 600,000 organizations used Cowork during the final two weeks of May 2026 TechCrunch. The results show where this type of AI is actually delivering value.

Business process work came out on top. More than one-third of all sessions — 33.4% — involved tasks like pulling together scattered status updates into a single report, creating onboarding checklists, or matching expense reports against receipts. These are the kinds of repetitive administrative jobs that nobody really wants to do by hand.

Content creation came second at 16.4%. That covers things like drafting documents, building slide decks, writing social media posts, and preparing proposals.

Software development — the task Cowork was originally designed for — accounted for only 8.7% of sessions. That is surprising because the underlying technology was built by engineers for engineers over the past two years.

The bigger picture here is worth considering. Anthropic and other companies spent years teaching AI agents to handle complex coding tasks — reading many files, planning changes, running tests, fixing errors when things go wrong. That same underlying structure turns out to work just as well for many other kinds of work. Pulling five people's updates into one document has the same basic shape as updating scattered code files: gather information from multiple sources, make decisions about it, and produce one organized result.

That said, this data came from a relatively small window and a particular group of early adopters — people willing to download a new desktop app in January. The usage patterns may shift now that the tool is reachable from any web browser or phone and used by people beyond the early enthusiasts.

What This Means Practically

Anthropic also launched Claude Tag in June 2026, an always-on AI that sits inside Slack and acts like a teammate rather than a tool you call on when you need it. These two products show how Anthropic sees AI agents fitting into work: some as a constant presence in the background, others as a tool you hand specific tasks to and check back on later.

The data Anthropic released — showing that most use today happens in operations and administrative work rather than software development — gives competitors like OpenAI and Google something concrete to think about as they build their own AI agents.

For anyone in IT or operations deciding where to use these AI agents inside their organization, the lesson is straightforward. If the biggest share of sessions on this tool is going toward business process work and content creation rather than software development, then budgets for AI agents should probably come from operations and administrative departments, not just engineering. Many organizations have not yet thought about it that way.


A brief note on the data: The usage patterns Anthropic released are based on what people reported they were doing during a two-week window in May. Early adopters and Max subscribers may use the tool differently than a broader group would. The cross-device version that just launched may reveal different usage patterns once more people have access from phones and browsers.