Agents are becoming core to how people interact with technology, and I identified the opportunity, innovated, pitched, formed a consensus, and led the design of foundational patterns for agentic collaboration on Windows. From early blue-sky exploration to shipping core patterns for agent presence, interaction, and collaboration, this work defined a new paradigm for human-AI interaction at the OS level.
As agents become core to people's workflow - creating a predictable and habit forming starting point makes it easy. We designed the taskbar to be a launching pad for agentic activities. Just type @ and Windows will surface relevant agents.
A persistent entry point to start any agentic task
A seamless integration for GitHub Copilot
3rd party agents participate seamlessly
As agents start working, it's important to keep the user in the loop of what the agent is doing and why. We designed a clear presence for agents in the taskbar, where they can show you what they're doing, and their reasoning process.
Hover preview as agent works
Agent hand raise for user input
Clear visibility into the agent's output, and ability to continue iterating
A crucial part for agentic collaboration is the ability for users to step in and work with the agent when needed. We designed for lightweight collaboration where users can jump in, provide feedback, correct the agent, or take over control when needed without disrupting their flow.
Keeping user in the flow while keeping them aware of agent's state
Notifications keep users informed about agent needs
Quick actions in notifications lets users unblock agents with a single click
When complete - user can review and act on the agent's output
When complete - user can review and act on the agent's output