Insurance is an information problem

I work in a company that has replaced an entire advisory model with AI.

Not as a supplement, but as the primary adviser.

This gives me a fairly specific perspective on what AI is actually good at - and what it is not. Yet.

AI is not good at relationships, gut feelings or trust. But, among other things, at handling large amounts of information, it is more systematic than any human can be. Without parallel.

Insurance is an information problem. Thousands of data points per company. Terms that change. A company that grows, while the policy stands still.

Most brokers start by looking at the policies. At Leaf we start by analysing the company. Only when we understand the actual risk profile do we look at the mismatch - what is the company exposed to that the policy does not address?

The traditional broker looks at their portfolio once a year. Our AI agent checks 24/7.

We are not smarter, but we are systematic in a way that humans can never match alone.

Yes, we are replacing the broker function, and we are replacing a model that was never built to solve the task adequately.