Sub-100 ms pricing is a governance problem

Fast ML-augmented rating only works when actuaries own the bounds.

Ando Kivilaid

Builds things with AI

2 August 20262 min

#pricing#governance

Making pricing fast is the easy part. Making fast pricing safe is the actual work.

Actuaries own the bounds

ML-augmented rating only works when actuaries set the limits the model is allowed to move within. In practice that means:

▹Shadow mode before any model prices real business.
▹Circuit breakers that fall back to approved rating when something looks wrong.
▹Explainable mappings from model outputs to approved rating grids.

Speed with and without governance

Speed without governance just produces more referral noise — more quotes a person has to look at anyway. With governance in place, auto-binding at Insly went from 22% to 68% while loss ratios stayed within bounds.

See Nora, Insly's AI for underwriting teams

Keep reading

insurance

Why underwriting AI fails without provenance

Underwriters don't need a clever model. They need to see where every extracted value came from.

2 August 2026•2 min

strategy

Is your company visible to ChatGPT? A practical AI visibility check

More buyers ask an AI assistant instead of searching. Here is how I check in twenty minutes whether the assistants know you exist.

23 September 2026•2 min

strategy

AI audit or discovery sprint: which one do you actually need?

Most companies buy the wrong first AI engagement. The real question is what kind of evidence you need next.

2 August 2026•2 min