On September 25, Microsoft announced a significant redesign of Copilot, moving it away from a single chat interface and toward what the company is positioning as a genuine agent platform. The new app is organized around three tabs — Home, Code, and Autopilot — and the shift says as much about where Microsoft expects knowledge work to go as it does about the product itself.
Home is the new default screen, merging the familiar Copilot Chat experience with Microsoft's "Cowork" agent, which handles longer-running, multi-step tasks rather than single back-and-forth prompts. Code is an app-builder layer built on GitHub Copilot technology, aimed at letting employees — not just developers — assemble trackers, dashboards, and small internal tools that run as cloud-hosted apps inside their organization's Microsoft tenant. Autopilot is the most agentic of the three: a persistent cloud agent that can monitor Teams channels, run recurring tasks on a schedule, and handle follow-up work even while the user is offline. It replaces an earlier Microsoft agent called Scout, introduced at the company's Build conference in June.
None of this is fully available yet. Home and Code are beginning to roll out through Microsoft's Frontier early-access program over the coming weeks, and Autopilot is only entering private preview at the end of the month. Microsoft has also confirmed that Cowork, Code, and Autopilot will run on consumption-based billing layered on top of standard Copilot subscriptions, meaning organizations that lean heavily on long-running agent tasks should expect variable costs rather than a flat per-seat price.
The Code tab is the part most likely to touch non-technical roles directly. If building a working internal tracker or dashboard becomes something an operations or HR employee can do through a prompt-driven interface rather than filing a ticket with IT, that changes what "building software" means inside a company — and it changes which employees are seen as capable of it. It's a continuation of a pattern already visible in "citizen developer" tools, just with a more capable AI layer underneath.
Autopilot is the piece worth watching for anyone whose job involves a lot of recurring, process-driven work — status chasing, report compilation, routine follow-ups. An agent that can run reliably in the background on a schedule is a more direct substitute for that kind of task than a chatbot ever was, since it doesn't require someone to remember to ask it.
None of this is reason for panic, but it is reason to get ahead of the curve rather than behind it. If your organization uses Microsoft 365, expect Copilot's agent features to show up in your environment over the next two to three quarters, not years. Spending a few hours now getting comfortable with agent-style tools — understanding what they're reliable at and where they still need a human check — is a far better use of time than waiting to be told it's mandatory.
For more detail on the rollout, see TechRepublic's coverage of the Copilot redesign.