Claude Cowork just went to mobile and web — and Anthropic's own data shows most users weren't coding anyway
Anthropic expanded Claude Cowork from desktop-only to mobile and web, and usage data across 1.2 million sessions shows only 8.7% were software development — the rest were business process work, reporting, and content creation, a strong signal that AI-assisted development tools are becoming general business infrastructure.
9 July 2026
Anthropic has expanded Claude Cowork — its autonomous task agent — from a desktop-only tool to mobile and web, rolling out first to Max subscribers. The interesting part isn’t the platform expansion itself, it’s what Anthropic revealed to justify it: across a sample of 1.2 million Cowork sessions from more than 600,000 organisations, software development accounted for just 8.7% of use. The largest category, at 33.4%, was business process work — pulling scattered updates into a report, reconciling spreadsheets — followed by content creation and copywriting at 16.4%.
That’s a meaningful data point for anyone tracking where “AI-assisted development” tooling is actually headed. The coding-agent wars of the past 18 months — Claude Code, Cursor, Copilot, Codex — have been fought over developer productivity. Cowork’s numbers suggest the same underlying agent technology is migrating fastest into the much larger market of knowledge workers who never open a terminal: ops managers, marketers, finance teams, founders running lean.
Tasks can now start on a laptop, continue running autonomously in the background, and get reviewed from a phone — even after the user has closed the app. That cross-device, fire-and-forget pattern is a meaningful step beyond the “chat with an assistant” model most teams still associate with AI tools.
So what
For founders and product leads, this reinforces something we’re seeing across every AI vendor right now: the tooling is consolidating around autonomous agents that operate across a whole workflow, not single-purpose chatbots. If you’re evaluating how AI fits into your product — whether that’s an internal agent for your own team or an AI-native feature for customers — the bar has moved past “does it answer questions” to “can it run a multi-step task unsupervised and be trusted with the output.” That’s the kind of build we do at AI-assisted development — get in touch via our contact page if you’re scoping something similar.