By Lewis Nibbelin, Research Writer, Triple-I
Given the growing ubiquity of artificial intelligence, its practical applications may seem self-evident. But for actuaries – whose work hinges on rigorous modeling and explainable risk assessment – translating AI-driven insights into analysis may pose as many challenges as solutions. A well-defined balance between technological capability and ongoing actuarial judgement is essential to navigating this shift.
“The challenge is not that there’s too much data – it’s having an awareness of what you’re looking for and then finding it,” said Dr. Michel Léonard, Triple-I chief economist and data scientist, in a recent interview for the Casualty Actuarial Society (CAS) Institute’s Almost Nowhere podcast. “If you look at all the data and it’s not focused and translated, the signal is not going to be what you need.”
Noting that many AI models train on varied language sources, Léonard stressed that data understanding and preparation are crucial to confronting the “black box,” or opacity surrounding the training and internal decision-making processes of complex algorithms. To integrate AI into risk assessment, carriers will need to demonstrate the mechanisms and actuarial record behind the models they deploy, especially for regulators and the broader public.
Though dynamic wildfire models, for instance, “very clearly show that the risk is more frequent and severe,” ongoing transparency around how these models work will be key to building “a bridge between regulators and the industry,” Léonard said.
While such models have facilitated greater access to granular, real-time data, critical information gaps continue to impede effective risk forecasting, especially following the 2025 federal government shutdown. Beyond being the longest federal closure in U.S. history, the shutdown also delayed or left permanent gaps in crucial survey data on employment, inflation, and other economic indicators, fueling more uncertainty for decision makers heading into 2026.
“Because of this uncertainty, we’re forecasting on the trend, which means that we cannot stress test or include validation for those stress tests,” Léonard said. “The lack of data on the U.S. economy is the main challenge for us right now.”
Current tariff policies – especially those targeting materials used in repairing and replacing property after insured events – add to the ambiguity. Though insurers appeared to avoid “the worst-case scenario” of COVID-19 levels of market instability last year, strategic stockpiling of imported goods to circumvent later post-tariff prices may have obscured their full impact, Léonard explained.
A pending Supreme Court ruling will determine the future of these policies, leaving global markets and consumers braced for potentially rising costs. Yet Léonard emphasized the insurance industry’s resilience in managing such “extreme, black swan-type events,” pointing out “that’s why we have a reasonable and adequate policyholder surplus” and other assets to ensure consumers remain protected.
Listen to Podcast: Spotify, Apple, YouTube
Learn More:
Tariffs, Shutdown Cloud 2026 Insurance Outlook
Triple-I Brief Explains Benefits of Risk-Based Pricing of Insurance
Tech — Especially A.I. — Is Top of Mind for Global Insurance Executives
JIF 2025 “Risk Takes”: Data Solutions for Today’s Challenges
L.A. Homeowners’ Suits Misread California’s Insurance Troubles
Data Granularity Key to Finding Less Risky Parcels in Wildfire Areas