Insurance runs on estimating risk and processing claims, and AI is reshaping both sides. Tractable ($185M raised) pioneered computer-vision damage appraisal — photograph a crashed car and get a repair estimate — while ClaimSorted ($13M) rebuilds the third-party claims administrator around AI plus expert human adjusters. On the underwriting side, whole carriers are being built AI-native: Nirvana Insurance ($260M) prices commercial trucking risk from telematics data, Honeycomb ($76M) underwrites multifamily and commercial property, Stand Insurance ($65M) uses physics-driven models to insure high-value homes in catastrophe-prone markets, and Coterie ($27M) automates small-business underwriting.
The users are carriers modernizing legacy processes, MGAs building products incumbents will not touch, TPAs and adjusters handling claims volume, and brokers who need faster quotes. The technology pattern is consistent: replace slow, sampled human assessment with models that consume richer data — images, telematics, IoT sensors, property records — and price or settle continuously.
Leaders prove loss-ratio improvement, not just workflow speed. Anyone can automate a form; the durable advantage is underwriting risk more accurately than the incumbent's actuarial tables, and that shows up in results over years, not demos. Regulatory fluency matters equally, since insurance departments scrutinize models used in pricing and claims decisions.
Buyers evaluating this category should ask for documented loss-ratio or cycle-time results, model explainability sufficient for regulators, integration with existing policy administration systems, and data requirements up front. NeuronFeed tracks 35 AI insurance companies with a combined $1.5 billion raised — and notably, Armilla AI ($31M) now insures AI itself.