Most insurers remain stuck at the trial stage of AI, according to figures cited by Hippo's chief product and AI officer. Andrew Ramsay argues four things separate the few firms getting results. They define an ambitious business problem before choosing a tool. They fix fragmented and inconsistent data before exposure to live systems. They connect AI to policy administration, claims and servicing platforms. They design workflows with adjusters and underwriters from day one rather than in an innovation silo. Regulated use also needs explainable systems, human checkpoints, bias analytics and ongoing monitoring. Hippo has put AI into claims intake and routine customer service. The figures come from Capgemini's World Property and Casualty Insurance Report 2026.
What changed
Insurers spent two years running isolated AI pilots and proofs of concept.
- 60% still in pilot stage
- 42% set no KPIs for AI
- 74% cite data quality barriers
- 86% vs 35% integrated data and cloud
Sources