Posts on what we actually think.
Essays on conformal prediction, social inflation, and where generative AI helps and where it doesn't. Written by the team.
Showing 13 posts · Tal Knafo
Clear filtersPolicy Limits in the Comparable Set
Policy limits artificially cap historical settlement data, creating a massive blind spot for predictive models that fail to track collectability.
Module Boundaries: Verdict-Heavy vs Settlement-Anchored Data
Jury verdicts are useful for tail risk, trial severity, and what a jury actually does. They are the wrong target for a settlement value. Mixing the two distributions in one model corrupts the forecast.
Building Forecasts When Comparables Report Verdicts
Legal analytics tools often fail because they treat public jury awards as ground-truth prices. Building a reliable claims forecasting system requires isolating the selection bias of trials from the reality of private settlements.
From Demo to Deployment: Onboarding a Carrier
A sales demo runs on pristine data, but real claims environments run on massive files and legacy mainframes. Transitioning a carrier from pilot to full deployment exposes the gap between software that looks good and software that works.
Want to talk to an executive?
Press, partners, investors, candidates — the inbox is monitored. Tell us who you are and we'll route it to the right person within two business days.