AI isn’t being challenged enough in Belgium — and that’s a problem

One in three employees now uses artificial intelligence at work — slightly above the European average. Yet Belgium is still falling behind. Not in adoption, but in how we use it. And that distinction matters more than we think.

What do I mean by “how”? Most people rely on AI for quick, superficial tasks: rewriting an email, generating a summary, looking something up. Useful, certainly. But far removed from what the technology is actually capable of. As a result, the public conversation about AI remains shallow, and we fail to notice that the foundations of society are quietly shifting beneath our feet. We quickly conclude that the results are disappointing, that AI does not really know anything, that ChatGPT’s answer is no more reliable than a fortune teller’s prediction. And then we move on. What’s for dinner tonight?

OpenAI measures how much “thinking capacity” users demand from its models. One surprisingly mundane metric captures this: reasoning tokens. Requests involving analysis, planning or coding force the model to engage in deeper reasoning and therefore generate more of these tokens. On this measure, Belgium scores below the European average, while countries such as Poland, Latvia, Lithuania, Greece and Spain submit more demanding tasks. That finding comes from OpenAI’s EU Economic Blueprint. Not a neutral source, perhaps — but the figures remain uncomfortable.

The handy assistant trap

The difference in reasoning tokens has consequences that are more concrete than they appear. It means we are currently extracting far less value from these tools than they could actually deliver. Worse still, limited use leads to underestimation — and that becomes self-reinforcing. If you see AI only as a convenient assistant, you draw small conclusions. You see no reason to change how you work, how you educate, or how you invest.

Belgium’s High Council for Employment recently found that 39 per cent of workers believe they need to strengthen their AI skills. Yet only 14 per cent followed any form of AI training in the past year. That gap — between the urgency people feel and the action they take — is part of the problem.

For anyone using AI today, ask yourself one concrete question: what would you do differently if tomorrow you had fifty intelligent assistants working for you? What problems would you assign them? What results would you expect? What safeguards would you build in?

Then try it. Not to polish a sentence, but to tackle a real problem. The difference in output is striking once you allow the model to actually work.

Measuring the wrong thing

For policymakers, counting how many people use AI is no longer enough. We must measure what they use it for. Are these tasks that genuinely demand time and thought? Or is usage largely cosmetic — a veneer of modernisation that leaves underlying practices untouched? Invest less in awareness campaigns and more in genuine skills development.

Belgium is a knowledge economy. Our prosperity depends on what people do with their minds: analysing, advising, writing, making decisions based on information. These are precisely the activities AI is now transforming fastest. Anyone who underestimates this is not misjudging a technological trend, but the economic foundation of this region.

The problem is not that Belgians fail to use AI. The problem is that we do not yet understand what AI truly can do — or what it will fundamentally change. That understanding will determine whether our companies, schools and governments remain competitive over the next decade. A knowledge economy that sees the nature of knowledge itself changing, yet looks away, is taking a serious gamble.

And then, once again, we return to discussing dinner.