Two on the machine, three from the books. In your room, in English or French.
I am a keynote speaker based in Lausanne. I speak in English and French, in Switzerland and across Europe, for boards, leadership days and conferences, in formats from twenty to sixty minutes. Everything in these talks is first-hand: the intelligence craft I was trained in, the systems I built, and seven years inside a large company's immune system.
A history of artificial intelligence, told as a story, for people who will use it.
HAL did what he was told. Jarvis knows what is meant. Between those two sentences lie seventy years of history, and this talk walks through them: the dreamers who wrote the specification centuries early, the rule-writers who drowned in their own rules, the idea that was buried twice and won, and the machine that finally learned to talk.
For non-technical rooms: boards, management teams, professional associations.
First given in October 2026 to a non-technical audience at an international insurer.
How to work with AI agents once the first weekend is over.
Describing what you want and letting the machine build it works, for a weekend. Then the project grows, and the machine forgets on Thursday what it decided on Tuesday. What fixes it is not a better prompt. It is the oldest habits of engineering: write the rules down once, agree on the words, let the machine check what a machine can check, and keep a human at the gates that matter. None of these habits is about code.
For teams and owners who already use AI and want it to hold on real work.
First given in August 2026.
How intelligence analysts cut through noise and decide under uncertainty.
Certainty never comes. Intelligence analysts know this, and they decide anyway, because they have a discipline for knowing when they have enough. This talk tells one true story, a family and a last flight out of Kyiv, and takes the room through what the analyst does with it: three states instead of two, one question that ends the waiting, and triggers decided before the crisis. It ends on the part nobody warns you about: when the analysis works, nothing happens.
For owners, executives and anyone who is sitting on a decision.
Scheduled for November 2026 at a data and AI conference in Lausanne.
Why transformations fail, and where to build instead.
Every large organisation has an immune system. It cannot tell a bad decision from a threatening innovation, so it attacks both, and it is very good at its job. AI wakes it up, because AI is a threat to the coordination that justifies many roles, and sabotage is rational when nobody at the top says what happens to the people whose work gets automated. The room scores its own company on twelve antibodies during the talk, then learns what the score is for: not fixing the antibodies, but building where they cannot reach. With one case of a project that beat the audit and died anyway.
For CEOs, owners and the leaders they bring with them.
First given in October 2026 to a peer group of digital leaders in Zurich.
Eight things an organisation needs before it runs agents, and one loop that makes them compound.
For forty years computers were file cabinets: store, open, edit, save. With agents the object is a task: declare, delegate, monitor, conclude. And when a machine runs the cycle, whose judgment it uses becomes the question. Context engineering answers "how". It never answers "should I". This talk names the missing piece, the Foundation, then the full set of eight primitives a well-run company already has, unwritten, and shows the only loop that is allowed to change them.
For executives who have decided to run agents and want to know what has to exist first. Works for the technical executive too.
Talks 01 and 02 form a half-day: where the machine comes from, then how to work with it.
Talks 03 to 05 are the three books, one talk each. They are booked on their own; in the short form each takes twenty minutes and fits a conference slot, a board offsite or a leadership day.
Book the half-dayUse the form below, or write to [email protected]. Say when, where, and who will be in the room. I reply personally within a week, with a yes or a no and a fee.
Four weeks for a talk as it stands; eight for one adapted to your room. Shorter is sometimes possible. Ask.
One fee for the room and travel at cost, no extras. The fee depends on the format and the audience, and you have it in my reply, before anything is agreed.
English or French, the same talk and the same fee. The slides follow the language of the room.
Lausanne, Geneva, anywhere in Switzerland, and across Europe. Online on request for the twenty-minute formats.
Yes. Most of the talks have a short form and a longer one, and I adapt the cases to your sector once I know the room. The idea does not change.
Or write to [email protected].
Field notes on stalled AI transformation, executive judgment, and the specification economy — the weekly letter behind Book 2. About once a week. No spam. Unsubscribe anytime.