Copilot in WindowsSystem Tray

The Problem

Windows keeps two kinds of things in the system tray:

1. Apps running quietly in the background 2. Apps you need to reach in one click.

Microsoft 365 Copilot was in there but right-clicking it offered exactly one thing, and that thing was to Quit the application.

It was very difficult to justify its presence in a coveted spot like the system tray, especially as Windows was doing a cleanup of Copilot logos and entry points.

This case study is how we built a case to keep it and evolve it.

I found an insight in telemetry

The tray was never going to be the biggest entry point. The useful question was how well it converted against the alternate entry points.

Read down the volume column and the tray looks dismissible.

Read across to conversion and it stops looking that way.

The tray was not underperforming. It was underbuilt.

User Expectations

Nobody has an opinion about tray design. They have habits.

Every other icon in that strip answers at least one of three questions:

1. What is this app doing right now 2. How do I get to it 3. How do I make it stop.

Copilot answered only the last one, which is why the icon read as clutter rather than access.

We wrote the expectation down as two ranked jobs.

P0. I want to access M365 Copilot instantly from my taskbar without needing to permanently pin the app.

P1. I want to quit M365 Copilot in the background without it running in the background.

What the neighbours were doing

We audited what first-party and third-party apps actually put in their tray menus, because those menus are the convention users had already learned.

The apps treated the tray as a status surface first and a launcher second, and it earns the space by telling you something you did not already know.

Designing for three surfaces

Partway through, the programme changed shape. Consumer Copilot and Microsoft 365 Copilot were merging into a single Windows app, and the tray stopped being a standalone feature.

It became one of three shell surfaces that had to behave identically: the taskbar jump list, Start and Windows Search, and the tray.

This means if the app was quit to the background it needed to reflect across these three surfaces in conjunction.

Writing the code

Built out the requirement in our app codebase on a new branch. A series of checks for accessibility and localisation and then we were off to merge to our main branch.

The tray shipped in the unified Copilot app and held its numbers: 25.4% attach, roughly 250,000 chat users a week, on a base of about a million. It survived the rationalisation it had been written to defend against.