Thor Olavsrud
Senior Writer

Swiss Re streamlines insurers’ natural disaster response with AI

Case Study
Jul 31, 20236 mins
Artificial IntelligenceCIOData Management

With natural catastrophes on the rise, the global reinsurer is leveraging predictive analytics and AI to help its insurer clients better anticipate claims and process them quicker in a disaster’s wake.

Hurricane Matthew
Credit: NASA/Handout via REUTERS

Natural disasters have been increasing in frequency, severity, and diversity in recent years, pressuring insurers to be more efficient and to anticipate event and claim fallout. The same goes for reinsurance firms, which provide insurance for insurers, reducing their likelihood of large payouts—a significant factor in the insurance industry’s response to natural disasters.  

According to the US National Oceanic and Atmospheric Administration (NOAA), the US has experienced 360 weather and climate disasters since 1980 in which overall costs reached or exceeded $1 billion for each one, totaling more than $2.57 trillion. In 2023, through July 11, NOAA confirmed 12 such events, including floods and severe storms. To keep up with the unsettling pace, Swiss Re, one of the world’s largest reinsurers, now leverages predictive analytics, machine learning (ML), and artificial intelligence (AI) to help its clients anticipate disasters and mitigate costs.

“If you look at the severity of losses over the past 10 years, one of the trends is that losses in the last five years are almost twice as much as those in the first five,” says Anil Vasagiri, SVP and head of property solutions at Swiss Re. “That’s an unmistakable trend and one which insurers, and reinsurers like Swiss Re, are having to understand and prepare for in order to operate in this sort of environment.”

Along with the overall cost of claims stemming from natural catastrophes (NatCats), insurers face operational challenges from the sudden and voluminous influx of claims that result. Immediately after such disasters, affected areas are frequently difficult or impossible to access, delaying response time to claims.

In his role, Vasagiri is responsible for the data and software assets deployed to Swiss Re’s clients, as well as the company’s overall data strategy. Late last year, Vasagiri’s property solutions division released a significant new tool to help its clients. Dubbed Rapid Damage Assessment (RDA), the platform blends advances in computer vision with other modeling techniques to help insurance clients better understand, plan, and analyze their portfolios before disasters strike, monitor their portfolios as events unfold, and then leverage technology to find and mitigate claims in the wake of an event. This can help identify claims of which policyholders are unaware, says Vasagiri, perhaps because they have holiday homes that have been affected and they live elsewhere, or they were forced to evacuate and have been unable to return. So RDA helps its clients inform those policyholders about how they could file claims.

“If you identify certain claims sooner and put them in remediation, it reduces the severity of losses,” Vasagiri explains. “For instance, putting up a temporary tarp covering a roof that had been blown off minimizes your loss,” he adds as a metaphor for how RDA can help mitigate insurers’ losses as a result of a NatCat.

Bringing AI to bear

Melding a range of data sources with AI capabilities, RDA helps claims managers and loss adjustors make faster and smarter claims decisions during three key phases of a NatCat event cycle.

First there’s pre-NatCat planning for an effective response strategy. To this end, RDA leverages proprietary Swiss Re NatCat models to automatically monitor the probable impact of a NatCat on client insurers’ portfolios so they can plan their response strategy and team mobilization.

Second, RDA addresses post-NatCat planning to help insurers’ prioritize property inspections. In the immediate aftermath of a disaster, Swiss Re coordinates with satellite and aerial imagery partners to capture footage that RDA’s AI models use to assess the damage severity for every property. This can even identify damage insurers weren’t aware of if no loss notice was filed.

Finally, RDA helps with remote claims triaging and assessment to reduce expenses and leakage by analyzing the impact to individual properties using multiple filters to create detailed loss reports for faster and more accurate claims settlement outcomes. This helps prevent losses compounding.

“These individual components existed before, but RDA is the way they’re all put together to make insights really actionable,” Vasagiri says. “We apply advanced computer vision techniques and AI techniques, to identify pockets where there’s visible damage and flag that to our insurers. This allows them to focus on pockets that are most damaged so they can dedicate appropriate resources and align their claim strategies accordingly.”

The platform is currently in full production for hurricanes and tornadoes, with hail and flood in the initial design phase.

RDA in action

RDA proved its worth when Hurricane Ian ravaged Florida and the Carolinas in September 2022 after devastating Cuba. By the time Ian made landfall in Florida, it was a Category 5 storm and the third-costliest US weather disaster on record, as well as the deadliest hurricane to strike Florida since 1935. Swiss Re estimates Ian cost between $50 billion and $65 billion in insured losses.

Once Ian ended, insurers struggled to deploy adjusters because it left large areas inaccessible, and stranded insurance adjusters and residents. But RDA’s NatCat model-driven loss predictions and actual damage assessment from high-resolution imagery helped Swiss Re’s clients optimize deployment of their adjustors, remotely triage claims, inspect properties, and proactively reach out to homeowners to extend support.

For example, as soon it was clear Ian would make landfall in Florida, RDA began delivering insights into thousands of individual policies, held by client Security First Insurance, which could be triggered because of resulting property damage. As the storm moved inland, Security First Insurance was already in the process of reserving resources, planning loss adjustments, and allocating claims resources. Within six days of the event, RDA had helped clients like Security First assess 35,000 properties. That grew to more than 88,000 after two weeks.

Delivering on the promise of RDA, says Vasagiri, required pulling together a cross-functional team with actuarial experience, modeling expertise, and advanced data science capabilities. It also required demonstrating the viability, efficacy, and value of the solutions to Swiss Re leadership and customers.

“When it comes to application of new technology, it’s all about demonstrating confidence to our end users and different stakeholders within the end-user ecosystem and doing it on a consistent basis,” he says. “There are some aspects you can demonstrate by testing, but there are others that only become real when they start using it. So when we put these sorts of platforms together, one of the capabilities we have on the back end of RDA is our own control room, monitoring the event and making sure our clients are actually able to access and understand the imagery and the insights we feed them in real time as the event unfolds.”