The technological shift is cutting deep into the world of work, and artificial intelligence is no longer playing a supporting role – it has become the driving force, catalyst, and guide all at once. While many are still speculating about its risks and opportunities, Microsoft’s latest study, Working with AI: Measuring the Occupational Implications of Generative AI, provides a much-needed counterweight to all the theory: it shows how people are actually using AI, what they are doing with it – and what they’re not. It’s not about promises, scare tactics or visions, but a sober look at the day-to-day reality for hundreds of thousands of users in the US. 200,000 anonymised conversations with Bing Copilot were analysed, with a clear focus on facts: what tasks can be accomplished using generative AI, to what extent, with what goal – and with what result?
The potential becomes clearest in those tasks that have always relied on thinking, writing, analysing, and structuring. Professions such as interpreters, historians, journalists, salespeople or writers rank highest among those whose tasks can be most effectively supported by AI. This is made measurable by the so-called AI Applicability Score. But it’s not about replacing the human – it’s about relieving them, through suggestions, speed, and clarity. In practice, that means users get help writing texts, researching, explaining complex subjects or organising information. The AI becomes a co-author, a researcher, a translator, a mental aid – but it remains embedded within the human workflow.
The study also debunks a common misconception: that AI is strongest where it can simply replicate human labour. In reality, this is rarely the case. Instead, its power unfolds where it complements rather than replaces. And precisely in areas where physical experience, craftsmanship or emotional intelligence are key, it quickly hits its limits. Digger operators, foundry workers, machine operators or carers are found at the lower end of the AI applicability scale – not because their jobs are less important, but because they simply can’t be digitally simulated. They are based on direct interaction with the world, on instinct, experience, presence.
Still, it would be wrong to read the study as a roadmap for future job losses. The authors stress that the findings offer no predictions about automation or salary trends. The interplay within modern companies is too complex, the new skills emerging from AI too unpredictable. Many roles change not because they’re replaced but because they evolve. New tasks arise, jobs are upgraded, their appeal may grow – not despite, but because of digital support.
The study gains particular depth from its distinction between the user’s intent and the model’s actual output. What is the person trying to achieve – and what does the AI do? In over 40 per cent of cases, the aim and the outcome don’t match exactly. This shows that AI doesn’t merely carry out commands, but often acts as an independent idea generator. It doesn’t always deliver what was expected, but often something useful – a text structure instead of a direct answer, a fresh perspective instead of a standard solution. It opens up space to think, invites questions, shapes thought.
Educational background also plays a measurable role. The higher the qualification – particularly at university level – the more intensively and successfully AI is used. Yet, unlike what many assume, there is no clear link between income and AI use. This suggests it’s not just about how much someone earns, but about the kind of work they do – and how willing they are to engage with machines creatively.
What remains is a realistic and yet encouraging picture. Artificial intelligence is neither saviour nor threat, but a tool whose potential unfolds through human use. Its greatest value lies not in automation, but in collaboration. It changes work – not through elimination, but through transformation, rethinking, and augmentation. The Microsoft study makes this tangible and shows that the real revolution doesn’t lie in what machines can do, but in what people choose to do with them. Those who use AI wisely won’t be overtaken – they’ll be the ones shaping what comes next.
