What to know about AI, Climate Change, and the Insurance Industry

An estimated 40% of work hours could be disrupted by generative AI in the coming years, according to Accenture. Insurance nearly tops the list of industries most affected, just behind banking. It’s a worrying outlook for an industry already grappling with the uncertainties of climate change. Extreme weather events are on the rise, critical infrastructure is more vulnerable than ever, and actuarial strategies are being revamped for a new risk landscape. Fortunately, those changes coincide with the AI revolution, which promises unprecedented predictive powers for risk modeling and climate resilience.

Here are five things to know about AI, climate change, and the insurance industry:

“Big data is creating a new generation of risk modeling and data strategies, a new generation of fraud detection and countermeasures, and a deeper understanding of the customer and the world. But as you digitize you need machines to help you scale. AI algorithms are the only feasible way to handle all the processing needed to “understand” context and customer. But successful AI is also heavily dependent on data. We have to leverage human judgment in the delivery of these solutions to build the right training data sets and knowledge bases over time, and we need to make sure AI algorithms are subject to responsible AI criteria.”

— USAMA FAYYAD, EXECUTIVE DIRECTOR

“Insurance is a highly regulated, very complicated product to be selling. We needed to find a way to dip our toe in so the experts could understand what's possible. So we had this very large all-around project that aimed to solve four or five different things that we knew were problems. [Our collaboration with EAI] was a ground-up effort to create a list of problems—what do we see out there that's an opportunity already?—see how that process is done, and then pick an opportunity and have data scientists work on it. It was also a chance for our actuaries to collaborate with real data scientists and students. Speaking of the results, on one hand we have a solution. We have a leverageable piece of technology that allows us to automate steps of an actuarial investigation. On the other hand, we also have a group of people who now know how to interact with data scientists.”

— DAVID MESSINGER, DIRECTOR AND ASSOCIATE ACTUARY, FULLSCOPERMS

“Responsible AI in insurance is key for underwriting and pricing models that may be adopted by insurance industry regulators, and it is crucial for developing models that do not discriminate. In the insurance industry, a common problem is in how to make decisions where there is a lot of noise. This is especially true when it comes to decisions about pricing or whether to continue a policy. One good thing about algorithms is that they may be biased, but they don’t have noise.”

— RICARDO BAEZA-YATES, DIRECTOR OF RESEARCH

“Weather extremes are hard to ignore. Bloomberg recently reported a doubling of insured losses owing to exacerbation of extreme weather-related damages. No wonder, then, that the private sector has started to develop AI and data science-driven solutions. Statistical downscaling can fill a critical gap in climate and earth system models, where aggregate low-resolution model simulations must be brought to stakeholder-relevant scales. Our physics-informed, uncertainty-aware AI methods have shown early promise in addressing this challenge and have since been adopted and cited by groups worldwide. This is incredibly important for the insurance sector given the scales of property and infrastructure damage.”

— AUROOP GANGULY, DIRECTOR OF AI FOR CLIMATE & SUSTAINABILITY (AI4CaS)

“In my lab, we study coastal processes using field observations and modeling. I have a firm belief that numerical modeling in the coastal region is only as good as the field data you have to validate and understand the models. Geophysics in particular is interesting because we don't get the same volume of high-fidelity data as other research areas currently using AI/ML, since field work in the coastal ocean tends to be expensive. We have to be a lot more creative about how we use machine learning. There's a lot of promise in physics-informed neural networks or physics-informed machine learning, which can help us fill in data gaps, or machine learning to help model the uncertainty in how coastlines and climate change will evolve together.”

— JULIA HOPKINS, ASSISTANT PROFESSOR, COLLEGE OF ENGINEERING

Learn about our AI Solutions Hub, our new Membership Program, our Responsible AI Services, and our on-demand AI Ethics Advisory Board. Stay in the loop and discover business opportunities.

‍