Anthropic has launched Claude Cowork on mobile and web, transforming its AI agent from a desktop-only tool into a cross-device platform. The rollout, initially in beta for Max subscribers, allows users to start tasks on a laptop, continue them autonomously in the background, and review them from a phone, even after closing the app. "Your work goes everywhere with you, and keeps going without you," Anthropic wrote in the announcement.
Usage data reveals coding is not the priority
Alongside the mobile launch, Anthropic published anonymized usage data from 1.2 million Claude Cowork sessions sampled between May 11 and May 31, drawn from over 600,000 organizations. The numbers challenge the dominant narrative that enterprise AI is primarily used for coding. Business process and operations tasks — such as aggregating scattered updates into a single report, building onboarding checklists, and reconciling spreadsheets — accounted for 33.4% of all sampled sessions. Content creation and copywriting came in second at 16.4%. Together, these two categories represent roughly half of all Claude Cowork usage. Software development, by contrast, accounted for just 8.7%. DevOps and infrastructure followed at 7%, research and intelligence at 6.4%, data analysis at 5.8%, document processing at 4.1%, and sales at 4%. The remaining 12 categories each remained below 4%, including personal assistance at 3.8%, education at 2.4%, and meeting intelligence at 1.8%.
Sponsored Protocol
Anthropic's strategy: targeting knowledge workers
Anthropic describes these dominant use cases as "the work around the work" — tasks that cross every role but rarely appear in job descriptions. "People use Cowork to assemble and structure information they can use to act on their expertise," the company explains. A lawyer can delegate document formatting, a hiring manager can synthesize interview feedback, and a team lead can produce slides explaining a decision. The aim is to capture the vast market of professionals who do not write code but handle reports, presentations, and coordination daily.
Sponsored Protocol
Three new mobile features: sync, background, and approval
The expansion to mobile and web introduces three concrete capabilities. First, sessions now sync across devices: users can start a task at their desk, check progress from a phone, and retrieve output from any device. Second, and more significantly, Cowork can run tasks in the background with no device online at all. Users schedule work for a specific time and Claude executes autonomously. Anthropic offers the example of Monday morning client prep: "Claude works through email threads, transcripts, and recent news, builds the briefing doc, and leaves the follow-up email drafted but unsent. Review it over coffee." Third, when Claude encounters a decision requiring human judgment, it surfaces the question to the user's phone. "Nothing ships until you have reviewed and approved it," Anthropic states. Desktop remains the most feature-rich surface, with access to local files and the browser, but the web version opens Cowork to users who cannot install software — a meaningful advantage in IT-controlled enterprise environments.
Sponsored Protocol
A competitive landscape in flux
The launch comes during a busy period for Anthropic. Days ago, it released Claude Sonnet 5, a model narrowing the gap with Opus-class models at a lower price ($2 per million input tokens until August 31). Two weeks earlier, it launched Claude Tag, a Slack-native AI agent for team collaboration. While Cowork is single-player, Claude Tag is multiplayer. Together, they form a pincer strategy to penetrate enterprise operations. Meanwhile, the Ramp AI Index from May showed Anthropic pulling ahead of OpenAI in business adoption (34.4% vs 32.3%), driven by Claude Code. But Cowork targets a much larger market: every knowledge worker with a laptop and a slide deck due by Friday.
Challenges remain. Security firm Armadin recently reported a sandbox escape vulnerability on Windows, though Anthropic responded that exploitation requires prior local code execution. As Cowork moves to web and mobile, risks of prompt injection and data exposure grow. Geopolitical tensions also complicate the picture: Alibaba announced it will ban Anthropic's AI tools starting July 10, following Anthropic's letter to the U.S. Senate accusing the Chinese giant of a distillation attack. Yet Anthropic signed a 20-year, $19 billion lease with TeraWulf for a data center in Kentucky, signaling long-term ambitions.
Sponsored Protocol
For more context, read about DeepSeek's plans for proprietary chips to bypass US export controls and Savi Security's AI scam protection app with live call monitoring. For an external overview, see the Wikipedia entry on Anthropic.
