Insurance companies are leveraging AI to address losses connected to climate

Insurance companies are leveraging AI to address losses connected to climate change.
Not a valid attachment ID.
Insurers are increasingly adopting a range of AI-driven strategies to enhance their ability to forecast rising losses caused by climate-related disasters, including unprecedented wildfires, hurricanes, and floods. In the early months of the year, natural disasters worldwide have already inflicted significant economic damage, with recent events including wildfires in Los Angeles, a cyclone in Australia that impacted the economy, flooding in Jakarta, and a catastrophic storm that resulted in numerous fatalities in the U.S. A recent report from broker Gallagher Re indicates that annual insured losses from such catastrophes have reached a staggering $150 billion—what they've termed "the new normal.Not a valid attachment ID." Traditional loss estimation models often involve intricate physics and complex computer simulations, but their predictions can be inconsistent.

For example, flood models sometimes yield conflicting results, and those focused on wildfires may struggle to account for a multitude of variables, from human impact to the unpredictable trajectory of embers carried by the wind. Some investors in catastrophe bonds actively avoid securities linked to these risks due to a lack of confidence in existing models. "Every model is an imperfect representation of a very complex phenomenon," noted Firas Saleh, director of product management at Moody's Corp. This is where artificial intelligence comes into play. Proponents argue that AI can deliver more precise assessments of property-level risks associated with weather events. Once an AI model is trained on the relevant data, it can process vast amounts of information sourced from aerial and satellite imagery to generate a comprehensive assessment of individual properties, leading to risk reports. Key factors may include the building materials used (brick or wood) and whether debris like pine needles may increase flammability.

Human analysts would struggle to produce such detailed assessments, especially in portfolios containing thousands or even millions of homes. Consequently, AI has become essential for insurers looking to accurately price risks, establish premiums, and maintain adequate resources to cover claims arising from severe disaster losses. "Losses from weather and catastrophes are currently outpacing our ability to manage them, leading many insurers to struggle financially as they fail to secure appropriate rates," said Jay Guin, chief research officer of the extreme event solutions team at Verisk, a catastrophe modeling firm. "AI is a game-changer." Zurich Insurance Group AG, one of Europe's largest insurers, utilizes AI-powered risk modeling software to evaluate catastrophe risk, sometimes tailoring it to meet its specific needs. "If there are fire hazards like overhanging vegetation or debris in your backyard, we can alert you to reduce the risk; otherwise, underwriting you may prove difficult," explained Ericson Chan, chief information and digital officer of the Swiss firm. The need for detailed data is pushing risk modelers to invest more heavily in AI. Verisk's wildfire model incorporates traditional elements such as wind speed and vegetation growth but also provides an additional layer of analysis through AI, utilizing satellite and low-flying aircraft imagery to assess individual homes. Moody's Insurance Solutions is taking a similar approach, recently acquiring CAPE Analytics, which employs AI techniques to deliver immediate risk assessments at the address level. Risk experts are increasingly relying on AI to address a particularly difficult challenge: modeling severe convective storms (SCS), which include hailstorms, thunderstorms, and tornadoes. In 2024, insured losses from SCS reached $61 billion, marking the second-highest level on record, according to Aon Plc. Texas, which sees more hail claims than any other state, recently approved an AI-driven SCS model developed by ZestyAI, a San Francisco-based company. Modeling SCS losses is complex, as insurance claims often stem from multiple, minor hail events that cumulatively damage roofs and homes, rather than a single, clearly defined incident such as a hurricane or wildfire. "You don't get a cavity from consuming one large candy bar," remarked Kumar Dhuvur, ZestyAI's co-founder and chief product officer.

"It's the result of consistently eating a lot of candy over time." ZestyAI's SCS model employs machine learning to integrate a variety of factors, including local geography, climate impacts, a three-dimensional analysis of buildings and roofs, and historical storm damage. The model produces "hail scores" on a scale from one to ten, which clients can use when setting their premiums. "Insurers require more precise tools," Dhuvur stated. In the U.S., where around 100 million properties exist, "manual calculations are impractical—you need AI." Amica Mutual Insurance Co. utilizes AI models to gain a competitive advantage in Texas's Dallas-Fort Worth area, known for high hail losses.

The company employs a ZestyAI model to pinpoint local homes with lower risk profiles based on individual characteristics, such as the age, pitch, and material of roofs, enabling them to competitively price their coverage. "With this approach, we can maintain more stable premiums, allowing us to grow in regions where other insurers struggle to achieve profitability," stated William Pitts, an Amica managing vice president.

[attachment deleted by admin]