I help CEOs build the workflow that replaces the one rejecting them when their organization won't become what it must: the budget, the technology, and the mandate are there, and still nothing changes. I've spent a career learning to see what systems hide.
I trained as an intelligence analyst. The discipline isn't predicting the future. It's reducing the range of surprise: decomposing vague questions into testable ones, weighing competing explanations, naming the gaps, and stating conclusions with calibrated confidence. It's the exact discipline AI transformation now demands of executives, and almost none have been trained in it.
I spent seven years at a Fortune 500 company building 0→1 AI products for geopolitical risk forecasting. I navigated the corporate immune system from the inside: watched it kill promising initiatives, suffocate acquisitions, and turn six-month projects into three-year odysseys. I learned that it can't be beaten with better technology, and rarely by fighting it at all: the change that survived was built outside the core, given real work, and let in only once it had won.
I served thirteen years as a Swiss reconnaissance NCO. The map is never the territory: you brief one plan, the terrain surprises you, and you adapt or you fail. It taught me discipline, risk management, and leadership under pressure, and a permanent habit of looking at what the roadmap can't show you.
I am not just an advisor but someone who builds and deploys. I ran a web agency serving EPFL projects (Solar Impulse, Alinghi, SwissCube), and spent eight years as software architect for the European Respiratory Society building large-scale scientific publishing and recommendation systems. Building is how I find out whether an idea survives contact with reality.
My master's thesis (Université de Lausanne, 2010) was on Henry James and the stories that refuse to resolve; its title is Title: yet to be chosen. Years later, at the EPFL Innovation Park, I designed a programming language that makes legal contracts computable, and the hard part was the same problem: a contract, like a story, is full of questions that are not yet true or false. The language gives undefined a seat at the table and returns what is missing instead of forcing an answer. In 2019, Stanford's CodeX group published the Legal Specification Protocol, a three-part design for computable law: an interface for non-lawyers, a data structure, an interpreter. We had been running all three since 2018. I took the system from prototype to production inside a Fortune 500 company and stayed to see it through: two million claims a year, 3,500 call-centre agents, millions of API calls a month on fifty thousand serverless functions, a full year at 100% availability on the group's own report, at a thousandth of the previous infrastructure cost. An external audit had said the new architecture was unproven and would never hold our cost estimate, because serverless always goes over budget. We had costed it properly, network and compute included: $15 a month. The first months came in at $8. It ran beside the old architecture for a year, and the old one was retired. I go all the way to production. The part I'm rare at is the part before the answer exists.
I work with digital leaders across Europe through the Boye & Company community, and I write on specification, organizational intent, and decision-making in the age of AI agents. The trilogy maps the journey: Book 1 is how to think; Book 2 is why and what; Book 3 is how.
“I first met Samuel during his work on computable contracts. Right from the launch of Eiffel Broker, I wanted to draw on his expertise in ontology and in integrating AI agents into our business processes. His experience inside a large international group brings us decisive strategic value. He has tested the subject in the field and knows it inside out, with its limits, constraints and opportunities. His role as an advisor is a real asset for us.”
In October 2018, before assistants could hold a conversation, I wrote How AI Assistants will impact businesses and consumers for the Boye & Company blog: assistant interfaces would leave no room for ads, trust would be the only moat, a handful of platforms would become the new middlemen, and consumers would delegate their purchases to agents. The voice-era numbers missed, but the mechanisms are on every 2026 board agenda. That gap (right on the forces, wrong on the channel) is why every claim needs a check, and checking is a discipline.
Awake gives you the map, the written core-or-edge decision, and the first workflow with a go or no-go. It costs CHF 7,500, fixed, and if I cannot give you all three, you get your money back.
Map What's Blocking YouField 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.