Two weeks after the AI Open House in Basel, I stood in front of a very different audience: around thirty first-semester students at the University of St. Gallen, building their first ventures in ten teams as part of the START Sprint by START Global. My workshop was called: AI in the workflow.
Two skills instead of ten tools
Tools change every month, so I did not teach tools. I taught two skills that survive every model update. First: delegate, do not substitute. AI does not replace your thinking; it takes over work packages that you define cleanly, with a clear definition of done. Second: integrate, do not bolt on. AI creates value when it sits inside the workflow, not in a separate window you open when you happen to remember it.Live, on their own projects
No slide deck doing the heavy lifting here either. I picked real cases from the room and built the solutions live: for one team, delegating a scoped work package end to end; for another, analysing official requirements and turning them into a question engine you can test against real profiles. Then every team applied the same two skills to their own project while I went from table to table.The contrast
Professionals ask me whether this works with their processes, their documents, their rules. These students did not ask whether. They opened their laptops and built. That is the generation entering the workforce with AI as a given, and it is the clearest argument I have seen for why companies should stop treating AI as a side project.Teaching is the part of my work that gives the most back, whether at CAAIE, on a stage in Basel or in a seminar room in St. Gallen. Thanks to START Global for the invitation, and to the teams for their energy. I left with more ideas than I brought.