What I actually teach

The hurdle is not AI literacy. It is dissecting your own work.

Most AI training in regulated industries fails the same way: a one-day session, abstract examples, a team that goes back to its own work on Tuesday and finds nothing has changed. I work differently — and what I teach is something more fundamental than prompts.

What I teach is how to dissect a regulatory use case. The principle sounds simple: if a task is recurring and can be broken into reproducible steps, it can be built. Filling structured templates, PSUR analyses, customer feedback categorisation, literature screening, equivalence rationale tables — the specific case does not matter. What matters is the pattern: a recurring task, a repeatable structure underneath it, a clear separation of what AI can take on and what stays with the human.

The principle is simple. Doing it is not. In my experience, the actual hurdle is not AI literacy or prompting skill — it is that most people have never learned to dissect their own work. They know how to do it. They cannot easily see how to take it apart. This is not a deficiency of intelligence; it is a skill that no curriculum teaches. And it cannot be lectured into a team. It has to be discovered together, on real cases, week by week — playfully, but seriously.

What changes over the course of an engagement is not that the team has memorised techniques. It is that they have started to see their own work differently. Once that shift happens — and it happens reliably, given enough real cases worked through together — the team no longer needs me. They have a way of thinking they can apply themselves, with whatever tools the future brings.