The ketchup bottle is shaken. Everything is about to come out

Anyone who’s ever held a glass bottle of ketchup knows how it goes. You shake it, you tap it, you turn it upside down. Nothing. You shake it again. Still nothing. And then, without warning, everything at once.

The AI transition in working life behaves the same way. For years, it’s sounded like something far off in the future, and then one day it’s all on your plate at once. The bottle is shaken now. The red is flowing thick toward the opening.

What the forecasters are actually saying

Microsoft’s AI chief has said that AI will handle large parts of everyday office work within a couple of years, not within a generation. Anthropic’s CEO, the head of one of the companies building the technology, estimates that half of all entry-level jobs for new graduates could disappear within five years. And Handelsbanken expects that nearly one in ten Swedish jobs could be gone within a decade.

You don’t have to believe any single number. The point is that the people closest to the bottle are the ones shouting loudest that it’s about to empty itself.

The wrong waiting room

The natural response would be for society to prepare: large-scale retraining, safety nets built for rapid shifts. A few countries are doing this, Singapore and South Korea are investing heavily in reskilling their workforces. But in most of Europe, it’s strikingly quiet, and there’s a structural reason: politics runs in election cycles. An uncomfortable and expensive decision today, whose benefits won’t be visible for ten years, is rarely rewarded in the next election. It’s not malice, it’s how the clock is built. Four years at a time can’t handle a shift measured in decades but landing in months.

The conclusion for anyone running a business or a public sector organisation is uncomfortable but simple: help won’t arrive in time. The transition will happen where it always actually does, at workplaces, one organisation at a time.

What a workplace can do while others wait

It sounds big, but it breaks down into two manageable parts.

The first is making AI use something shared and visible instead of a hundred private experiments. On this point, no one needs to wait for anyone: in EuroWork, the cost appears next to every response while the work is happening, every AI answer is labelled as an AI answer as required by the EU’s AI regulations, and anything an AI agent does leaves a receipt and waits for a human’s approval. What’s visible can be managed. What’s invisible is just something to hope for.

The second part is learning, and here it needs to be said straight: if you’re waiting for a national retraining programme, you’ll be waiting a long time. But learning doesn’t have to be a programme. In EuroWork, it’s built into the workday: a get started! course that shows what the technology does well and poorly, and an education assistant that’s always available in the menu and answers questions about AI without ever rolling its eyes. If someone has a silly question at nine o’clock on a Tuesday, they get an answer at nine o’clock on a Tuesday, not on a course day in November. That’s how habits are built, and habits are the only thing that helps when the bottle empties itself.

Before Thursday

In the original version of this text, written for politicians, I ended with a call to wake up now, preferably before Thursday next week. The politicians still haven’t gotten back to me.

But you, who lead an organisation, don’t need their alarm clock. The bottle doesn’t care about election cycles. The person who today can answer what the AI is used for, what it costs, and who approved it will also handle the day when everything arrives at once.

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