A few days ago, I was talking to R.B. in a crowded coffee shop at Zurich’s Central railway station. The background noise would have killed any romantic conversation, but our exchange was so intense that I felt transported to an entirely different world — the universe of Agentic AI.
R.B. is a young physicist who spent a few years working as an actuary in leading reinsurance companies. Today he is an independent consultant and a prophet of a new faith: “Stop writing IT programs — befriend Agents.”
I had watched a video in which he explained risk rating in remarkably simple terms, and I was impressed by his approach. A common friend brought us together, and there I was, trying to understand how his method could be shared with the students at our online insurance school.
As the philosopher George Steiner once said, our minds are constantly distracted by wandering thoughts. And so, while listening to R.B., my memory jumped back to my time at MIT, when Professor Arnoldo Hax unveiled the mysteries of Monte Carlo simulation. I immediately saw how useful it could be for calculating the price of reinsurance covers. With Arnoldo’s help, I developed a simple way to apply the method.
Back in Zurich — this was in 1975 — our chief actuary told me he would “carefully consider” my initiative. I never heard of it again. To be fair, Monte Carlo simulations were only just reaching the insurance market… for a younger generation of actuaries.
Earlier in my career, I had another revealing experience with risk rating. A young colleague had just returned from an internship at our U.S. subsidiary and brought back a new reinsurance method: excess of loss. One day, I thought we could use it to protect our clients in Chile against earthquakes and mentioned it to my boss. He was interested.
A few days later, he called me to his office, where the Head of the Technical Department was already waiting. They were discussing the Chilean case. René, my boss, said that the number he had calculated was 2. The Head Technician smiled and replied, “Great, René, I arrived at the same number!”
Timidly, I concluded that we could quote our client 2 per mille.
“Yes,” said my boss. “That is the price.”
But the Head Technician’s expression told me something was wrong. I later learned that he would have quoted 2 percent. In the end, we wrote to our client that the cost would be 1.1%.
And that, my friends, is how prices were calculated by Agentic AI’s grandparents.
