Notes on AI, legal technology, the law, and institutional judgement in specialty insurance
March 2026
Designing AI for Underwriters
I thoroughly enjoyed chairing Waswate Ayana and Amil Malik's excellent Designing AI for Underwriters session for the Insurance Institute of London, at Convex. The session explored the practical realities of building AI tools that underwriters actually want to use, with talking points that included: the challenges of adoption; the necessity of respecting tacit expertise; starting small with a 'minimum loveable product'; building trust by surfacing flaws in outputs; the importance of partnering with underwriters from a place of empathy; and designing for genuine human/AI collaboration rather than replacement.
February 2026
When your AI Conversations Aren't Privileged: Lessons from the First US Ruling on Generative AI and Legal Protection.
If anyone in your organisation used a consumer AI tool last week to think through a sensitive commercial problem, nothing they typed is protected. Not by privilege. Not by any other legal doctrine.
That's the effect of United States v. Heppner. In this piece, I look at gaps in the court's reasoning, why an English court might reach a different result, and what organisations should actually be doing about it now, not just legally, but structurally.