1. Without human judgement, you get volume, not quality.

AI can produce non-stop. Text, images, analyses, decision-ready briefings, in any quantity and in ever shorter time. What it can’t do: judge which of it is worth keeping. That sense of judgement, the feel for what’s right, for coherence and the emotions tied to it, has to come from people, at every single step. Delegate it fully to the machine, and what you get is arbitrariness at impressive speed.

2. We don’t know the solution yet. That’s the precondition for finding it.

Anyone arriving with ready-made answers hasn’t yet heard the essential questions. Every organisation is different: different people, different data, different risks, different opportunities. What works for a family business will fail in an NGO, and the other way round.

3. Listening is the most important part of the work.

The real need rarely lies where the first request points. «We need a chatbot» often means: «Our customers wait too long for answers.» Only once the real need is named does the conversation about technology pay off. I take the time until that’s happened, even when it takes longer than planned.

4. Transformation happens in people’s heads, not through tools.

New software without new thinking is just a more expensive version of the past. The question is never only which system gets introduced, but what changes alongside it in how the team works, who’s responsible, and what they’re capable of. That kind of change can’t be bought. It has to be guided, patiently, and with respect for what was already working.

5. Sovereignty is a stance, not a product.

Handing your data, your models, and your decision paths entirely to a small number of non-European vendors means giving up more than technical infrastructure. You give up room to act. In every engagement I check which technology fits the context: the data situation, the risk appetite, the legal requirements. European and open-source alternatives are seriously competitive today in many use cases, and they create fewer dependencies. Where they aren’t, I say so just as plainly.

6. Compliance is the foundation, not a brake.

Data protection, GDPR, governance: these aren’t annoying hurdles to be sidestepped. They’re the conditions under which AI can hold up in an organisation over the long term. Cut corners here and you build on sand, and you’ll notice it two or three years on, when the bill arrives.

7. Start small, measure as you go, stop in time.

A pilot that fails after four weeks and is cleanly wrapped up is worth more than any strategy document still sitting unused in a drawer four months later. I prefer small, concrete steps whose results can be checked to confirm whether the path chosen is the right one.

I don’t arrive with a finished solution. Nobody has one, except you yourself, at some point in the course of a successful project. What I do bring: the craft of asking the right questions, the patience to listen until the real concern becomes visible, the experience to propose a first workable path, and the staying power to see it through.