Microsoft on February 26 unveiled Copilot Tasks, a feature that lets its AI assistant execute multi-step jobs on a user's behalf. Described on the Copilot blog as "a to-do list that does itself," the tool accepts natural-language instructions, builds a plan, and carries it out using a cloud-based virtual computer and browser. No code, no manual workflow setup.
The pitch: tell Copilot what you want done and walk away. Microsoft's examples range from tracking apartment listings every Friday and booking showings, to pulling urgent emails each evening with draft replies attached. It can also generate documents (turning a syllabus into a full study schedule, for instance) or handle logistics like rebooking a hotel when the price drops. Tasks can run once, on a schedule, or as recurring routines.
Microsoft AI CEO Mustafa Suleyman posted on X that the feature requires "no complicated setup or coding skills." That framing puts Copilot Tasks squarely against a growing field: Anthropic's Claude Cowork, OpenAI's ChatGPT Agent Mode, and Google's Gemini-powered browsing in Chrome all occupy similar territory. Whether Microsoft's version works more reliably remains untested outside the company.
Copilot Tasks asks for explicit consent before spending money or sending messages, and users can pause or cancel anytime. Pricing and licensing details were not disclosed. The feature launched as a limited research preview with a public waitlist; broader access is expected in the coming weeks.
Bottom Line
Copilot Tasks is in a limited research preview now, with no pricing announced and a public waitlist open for early access.
Quick Facts
- Launched: February 26, 2026, as a research preview
- Availability: limited tester group; public waitlist open
- Runs on: cloud-based virtual computer and browser
- Pricing: not disclosed
- Competitors: Claude Cowork, ChatGPT Agent Mode, Gemini auto-browse
