Climate losses expose limits of traditional risk models, BCG says

Insurers are increasingly using AI-based systems that allow risk assessment at the property level.

Traditional catastrophe models are becoming less effective for insurance underwriting and pricing as climate-related losses increase to over $100b in both 2023 and 2024, according to the Boston Consulting Group (BCG).

In its May report, The Private Capital Opportunity in AI-Enabled Climate and Sustainability Sectors, BCG said the losses come as the number of billion-dollar climate events more than doubled since 2000.

BCG said traditional catastrophe models rely on historical loss data and broad geographic assumptions, which are becoming less effective in reflecting changing climate patterns.

It said insurers are increasingly using AI-based systems that combine satellite imagery, sensor data and atmospheric models to produce asset-level, forward-looking risk assessments.

The report said this allows insurers to assess risk at the property level rather than using broader geographic models.

BCG identified asset-level risk analytics as a key insurance application, with AI used to assess property characteristics and model potential hazard impacts as part of underwriting.

According to BCG, one US insurer improved its combined ratio by 4.4% in the first year after deploying ZestyAI’s platform, although the report did not identify the company.

The report said improved risk visibility could support underwriting for an additional 15 million to 20 million insurance policies in markets where hazard data has been limited.

BCG estimated the value of AI-enabled asset-level risk analytics at around $30b, concentrated in the property and casualty insurance sector.

It said excess and surplus lines insurers and climate-specialty carriers are amongst users of these tools. Parametric insurers were also identified as using AI-driven climate risk models in underwriting.

BCG added that portfolio climate stress testing tools are also being used by insurers and financial institutions to assess long-term climate exposure and support disclosure and capital allocation.

Across all applications, the report said AI-enabled climate risk modelling could generate around $75b in annual value by 2028.

BCG said adoption is being driven by the need to reduce losses, improve pricing accuracy, and meet disclosure requirements.