techcrunch.com web signal

Anthropic Opens Claude Cowork to Web and Mobile for Max Users

TL;DR

  • Anthropic extended Claude Cowork from a desktop-only January launch to web and mobile starting Tuesday, available to Max subscribers.
  • In a 1.2 million-session sample across 600,000+ organizations from the last two weeks of May, software development was just 8.7% of Cowork usage.
  • Business process operating led at 33.4% and content creation and copywriting followed at 16.4%, per Anthropic's own study.

Anthropic's tell on where the agent market is actually going came bundled with a rollout this Tuesday, and it is worth pausing on. The company opened Claude Cowork to web and mobile for its Max subscribers, TechCrunch reported, extending a product that launched as a desktop app in January. That part is a modest logistics update. The part that matters is the usage data Anthropic published alongside it.

According to the same reporting, Anthropic sampled 1.2 million anonymized and aggregated Cowork sessions from more than 600,000 organizations over the last two weeks of May. The largest category at 33.4% was business process operating, followed by content creation and copywriting at 16.4%. Software development, by comparison, only accounted for 8.7% of Cowork usage. For a product line that emerged out of Anthropic's coding-agent narrative, that is a striking self-report.

Why this matters if you are not tracking agent roadmaps closely: the last year of framing treated coding as the beachhead for agents because developers were an obvious audience willing to pay. What Anthropic is now signaling, and what OpenAI has been signaling with Codex per the same TechCrunch write-up, is that the beachhead was never only coding. TechCrunch notes Codex began as a software development tool but is increasingly being used by non-developers for reports, spreadsheets, presentations, research and data analysis. The buyer profile shifts from a VP of Engineering to a COO, and the seat-count math changes with it.

The honest caveat is that Anthropic sampled and framed the data itself, so the 33.4% headline should be read as directional, not settled. The reporting does not say how many of these sessions completed cleanly, what the organization-size skew looked like, or whether business process operating is being propped up by a handful of heavy users. Max is also a paid tier, so the mix inside Max may not represent Anthropic's broader user base.

Still, the forward-looking read is straightforward. If a coding-tools vendor keeps publishing numbers that say coding is under a tenth of usage, the ops, finance and comms teams inside enterprise buyers should probably start budgeting agent seats the way engineering already does. That is the seat-expansion story worth watching next.