Claude Cowork debuts in research preview

This is less a chatbot feature than a workflow engine for white-collar busywork, and that framing is probably the right one. It extends the Claude Code playbook from software tasks to general office work, which is a clearer product story than generic “agentic AI.” The strongest use cases are the boring, repeatable jobs that usually get postponed: sorting files, assembling drafts, and turning messy inputs into structured outputs. Anthropic is leaning on trust and oversight rather than full autonomy, which should help adoption in higher-stakes knowledge work. The main constraint is still obvious: this is only valuable if it reliably handles context, app switching, and file operations without becoming another brittle agent demo.