Independent Analysis of AI in Insurance

Large language models are often criticized for being non-deterministic, but human insurance decisions have never been deterministic either. Research found a 55% median difference between underwriters pricing the same cases. AI changes the problem by making variance more measurable and consistent, but shared models can also turn independent errors into correlated ones. For insurers, the real question is not whether AI is deterministic, but what shape of error they want on the book.

AI is moving closer to the messy reality of insurance operations. This week: agentic underwriting workflows, OpenAI slowing model development over cyber concerns, Munich Re’s At-Bay acquisition, and the growing push to define AI-related coverage.

If AI makes carriers better at selecting and pricing risk, should they need less reinsurance?

Probably not. Reinsurance demand is shaped by more than underwriting confidence, including capital, liquidity, volatility, concentration, and price. And if carriers increasingly rely on similar models and decision systems, better individual risk selection could actually make diversification harder.

Part 3 of The Economics of AI in Insurance looks at what AI may change for reinsurance, and what it probably won’t.

AI is moving from promise to operating reality. This week: regulators sharpen examination tools, Allianz prices transformation, data centers strain insurance capacity, and reinsurers confront the data problem.

Everyone expects AI to cut underwriting costs. Almost no one is asking where the resulting capacity goes. This thought experiment follows it, and finds the more important question hiding underneath: not what AI does to expenses, but what it does to uncertainty.

This week: autonomous AI liability, Moody’s warning for retail P&C distribution, AI-driven underwriting changes, the EU AI Act, AI-enabled cyber losses, and the growing transparency gap in AI research.

Jevons Paradox explains why cheaper technology often leads to greater demand. But insurance isn’t coal. This analysis examines why AI may expand the insurance industry through a different economic mechanism—and why that distinction matters for insurance leaders.

This week’s AI Insights focuses on quality over quantity. Inside: Germany expands AI oversight of insurers, Starbucks provides a real-world lesson on AI deployment, Anthropic releases Claude Opus 5, Europe moves from AI rules to AI enforcement, and the AI data center boom creates new insurance exposures. Plus, links to every primary source discussed.

The management frameworks that shaped a generation of technology leaders still influence insurance today. AI isn’t proving them wrong, but it is revealing assumptions many executives no longer realize they’re making.

Deloitte’s 2026 State of AI in the Enterprise survey found financial services carries the highest share of surface-level AI use of any industry surveyed, even as worker access to AI tools has roughly doubled. InsuranceIndustry.ai reads the report’s findings on sovereign AI, agentic governance, and workforce redesign specifically for insurance leadership, and finds three distinct problems hiding inside one survey.”