Clippy is back, and this time you can't turn him off

Do you remember Clippy? The cheerful paperclip that popped up in the corner of Word, waggling its eyebrows and asking if you were maybe writing a letter. The whole world said no, Microsoft finally listened, and in 2007 it was put down in what felt like a public celebration.

Now he’s back. Bigger, and bolted in place.

A whole key of its own

Today’s successor is called Copilot, and it no longer lives in the corner of a single program. It’s built into Windows itself, and new keyboards come with a dedicated Copilot key, the first new key in decades. Clippy at least had the decency to be turned off with one click. Copilot is part of the furniture.

And it won’t stop there. Rumours about the next version of Windows describe AI as the system’s central nervous system, not just a feature in the computer but the computer’s way of being. At the same time, hundreds of millions of users stubbornly refuse to leave old Windows 10, even though support has ended. It’s worth pausing here: people are choosing a system without security updates over one that has decided to help them.

It’s not AI people are saying no to

It’s easy to draw the wrong conclusion here. The resistance isn’t about people disliking AI. The same people who refuse to upgrade Windows happily paste text into an AI chat over lunch and let it polish an email. AI use at work is growing fast, every survey shows that.

What people are saying no to is something else: not being asked. An assistant you open yourself is a service. An assistant already sitting in the wall when you arrive at work is a surveillance issue, a cost issue, and a who-decided-this issue, no matter how good it happens to be.

Clippy at least asked if he could help. That was the only thing he did right.

AI you choose

When we built EuroWork, that difference sat in the room with us from day one, like a warning paperclip in the corner.

So our AI waits until someone asks. The desktop you see in the morning shows what’s happened since yesterday, in a way you can grasp in a second, and then it’s quiet: nothing pops up, nothing demands attention, no one has decided you need help right now. When someone does use the AI, the cost appears next to the answer while the work is happening, down to the last penny. Every AI answer is labelled as an AI answer, as required by the EU AI Act since August. And when an AI agent is about to do something for a colleague, it first shows its plan and waits for approval, because what gets sent should have a human’s name and responsibility behind it, not a paperclip’s.

It’s the same technology as in the built-in assistants of operating systems, often the same underlying models. The difference is who holds the choice: you choose to open it, you see what it costs, and you can answer for what it does with your information, because it’s processed within the EU and nothing is saved in secret.

The lesson from the paperclip

Clippy didn’t die because he was bad technology. He died because he couldn’t tell the difference between offering help and intruding. Now his successors are being built into the very floors and walls of our workplaces, and the question for every leadership team isn’t whether employees should have AI, that question has already been settled by the employees themselves.

The question is whether your AI is something you’ve chosen, with a receipt and accountability, or something that came with the building.

Your employees are already using AI. Do you know what they’re sharing with it?