Karen Clark on AI, Frequency Perils, and Why Reinsurers Stopped Trusting the Models

Artificial intelligence is transforming catastrophe modeling by enhancing physical dynamical models, specifically for frequency perils like severe convective storms. Unlike traditional statistical methods that struggle with the complexity of these amorphous weather events, AI-informed models leverage vast amounts of high-resolution atmospheric data to improve forecasts for phenomena such as derecho durations and tornado activity. Karen Clark, founder of Karen Clark & Company, emphasizes that model accuracy depends on continuous testing against real-world events, citing the 2021 Arctic air outbreak as a benchmark for validation. While residential property data has improved significantly, commercial risk modeling remains hindered by inconsistent replacement cost information. Ultimately, the industry must move toward frequent, transparent model updates to maintain relevance in a changing climate, as better data and refined modeling techniques provide critical arbitrage opportunities for insurers and reinsurers navigating evolving catastrophe risks.