Microsoft just admitted that its Copilot button — the one it plastered across Windows, Office, and Edge — was annoying enough to warrant burying. If you've been designing AI features lately, this should feel uncomfortably familiar.

What Happened

Microsoft has rolled out a significant redesign of Copilot, its AI assistant. The headline changes include visual tweaks and deeper integration across the productivity suite, but the most telling detail is this: they've relocated the Copilot button. The one that appeared uninvited on taskbars, in browser toolbars, and across Office apps. The one users complained about incessantly. It's been moved somewhere less conspicuous.

Microsoft is claiming early wins — usage increases of 27–43% following the redesign. But here's the asterisk they're hoping you'll gloss over: that data covers just one week. Microsoft themselves concede it "may not be indicative of long-term usage trends."

One week. That's not a trend. That's novelty.

The Bigger Picture

Here's what makes this story interesting beyond Microsoft drama. The Copilot button saga is a near-perfect case study in the hardest UX problem in AI right now: integration.

We've moved past the phase where simply having AI in your product was enough. The question now isn't "does your product have AI?" — it's "does the AI feel like it belongs there?"

Microsoft's original approach was brute force. Put the button everywhere. Make it unmissable. The logic was straightforward: if users can see it, they'll use it. But that's not how people adopt tools. Nobody wants to be poked constantly by a feature they didn't ask for. It's the digital equivalent of a shop assistant following you around asking if you need help every thirty seconds.

What's revealing is that Microsoft didn't remove Copilot — they just made it less aggressive. And that's the tension every designer working on AI features has to navigate. Make the AI too prominent and it feels pushy. Make it too subtle and nobody finds it.

There's a sweet spot. Most teams haven't found it yet.

I think the mistake many teams make — Microsoft included — is treating AI as a feature that needs its own dedicated entry point. A button. A sidebar. A separate mode. But the best AI integrations I've seen don't announce themselves at all. They're woven into existing workflows so naturally that you barely notice the boundary between your work and the AI's contribution.

Think about autocomplete in your search bar. You don't click a "suggest" button. It just happens, in context, at the moment you need it. The best AI features should work more like that and less like a toolbar button demanding attention.

The one-week usage spike is worth dwelling on, too. It's tempting for product teams to redesign, see a bump in metrics, and declare victory. But initial curiosity isn't the same as sustained adoption. People click on new things because they're new. The real question is whether they're still clicking in three months — and whether those clicks are actually productive.

If you're measuring AI adoption in your own products, be honest about the difference between novelty and genuine usefulness.

Tool Spotlight

On a completely different note, there's a quiet CSS development worth your attention: contrast-color(). It's a native CSS function that automatically selects a legible foreground colour based on whatever background you give it. No JavaScript. No accessibility plugins. The browser just handles it.

Why mention this on an AI blog? Because it represents the kind of algorithmic design automation that actually works — it's baked into the system at the right level, it doesn't require a dedicated button or workflow change, and it solves a real problem. A huge number of websites still fail basic contrast checks, and this addresses it invisibly.

It's automation done right: useful, respectful of the designer's intent, and completely out of the way.

If you're working on design systems or theming engines, contrast-color() is worth exploring. It won't replace your judgement on colour palettes, but it'll catch the cases where your carefully chosen brand colour becomes illegible on a dark background at 2am when nobody's checking.

Takeaway

Next time you're designing an AI feature, ask yourself: does this need a button? Or does it need to be woven into something the user is already doing?

The best AI features don't announce themselves. They don't need a dedicated entry point or a special icon. They show up in context, do something useful, and get out of the way.

Microsoft spent two years learning this lesson. You don't have to.