In the lecture hall since 2017. On my own account since 2019.

These are not the questions I think get asked — they are the ones that actually come up every time. The answers are the same ones I give in the course.

May I put student grades or coursework into an AI tool?

No. No personal data about students and no confidential documents go into public AI tools. In the classroom I work with a traffic-light rule: red is names, grade lists, exam submissions and confidential school documents. Amber is your own teaching material — because of possible third-party rights — and internal concepts. Green is public content, your own texts and synthetic examples. The rule of thumb: once something is in someone else's system, you will not get it back. Data protection starts in the prompt.

Do I need prior knowledge to keep up in an AI course?

No. The programme I direct at WBZ Kanton Luzern is explicitly billed as hands-on continuing education for beginners and professionals alike — the German course title says so outright, and both groups sit in the same room. From the anonymous course evaluations, translated from German: “The instructor did an outstanding job of engaging even the participants with little prior knowledge.” What you should bring is not prior knowledge but a task of your own that you want to work on.

Does AI understand what it is saying?

No. A language model has no understanding of the world — it is good at the language game, but it does not know what a residual-current circuit breaker is; it only knows the words that occur around that term. That is not hair-splitting, it is the practical core: fluent language feels like knowledge, and the mistakes happen precisely in that gap.

How do I tell whether an AI answer is made up?

Honestly: often not directly — and that is the point. I run the same demo in class regularly: I ask a model for three studies on AI adoption in Swiss SMEs, with sources. It delivers three plausible studies with authors and year. None of them exists. The conclusion is not avoidance but a duty to check: wherever standards, figures or facts are involved, you verify. Always. The AI did not lie — it continued. The responsibility lies with whoever publishes without checking.

Will AI replace the teacher?

AI will not replace you. But someone who uses AI well will replace someone who does not — which is why I teach this. The same holds for teaching as for any other profession: the model takes over the routine, you supply the context it does not have. Your professional experience is not a disadvantage against the technology, it is the precondition for using it sensibly.