Computer use is now in public preview inside Microsoft Copilot Studio, letting agents built on the p...
The AMW Read
Incremental capability expansion (computer-use, UI-level automation) for an already-established enterprise agent platform, not a new player or structural shift.
Computer use is now in public preview inside Microsoft Copilot Studio, letting agents built on the platform operate websites and desktop applications that lack APIs. Instead of scripted selectors, builders describe a task in natural language and the agent uses built-in reasoning to read screens and adapt when a layout changes. The release adds a hosted browser powered by Windows 365 so teams can start without provisioning a machine, plus starter templates, credential storage for logins, and an allow-list that halts a run automatically if the agent tries to leave the approved sites or apps. It also connects to machines already running Power Automate desktop flows, letting existing RPA setups extend into computer-use tasks. Access is limited to US-based Copilot Studio environments at launch.
The move pushes Copilot Studio past chat-and-API orchestration into direct UI control, the harder and more general form of automation that legacy dashboards, supplier portals, and API-less internal tools have always resisted. By packaging vision-based reasoning with guardrails like credential vaulting and allow-lists, Microsoft is framing computer use as a governed enterprise capability rather than a raw automation trick, and bundling it into a platform many customers already use for Power Platform workflows rather than selling it as a standalone product.
For builders and RPA vendors, the practical question is procurement: teams already inside Power Automate get computer-use access without a new vendor evaluation, which raises the bar for point-solution automation tools competing on UI-navigation alone. Per the AI Market Watch index, coverage of Microsoft Copilot Studio rose to 5 items in the last 90 days versus 1 in the prior 90 (name-matched over pipeline-ingested sources only), consistent with a faster feature-release cadence on the product worth tracking for enterprise-agent procurement decisions.
