As California’s summer fire season approaches, Zesty.ai is using machine learning to help insurance companies and homeowners make realistic risk assessments — and reduce damage to homes and property.
sAttila Toth is located in the suburbs of Oakland, California, in the nearby wooded hills. The CEO keeps an eye on what locals call “the city,” and in the distance, San Francisco, or “the city.” Nearby, Toth sees a tangle of redwoods, eucalyptus, and oaks—and the wildfire dangers they pose.
This “wild-urban frontage” is not far from the site of the 1991 Oakland Hills fire, which suddenly broke out in a densely populated area. Over the course of four days, 3,000 thousand homes were destroyed in one of the city’s richest neighborhoods, causing an estimated $1.5 billion in damage ($3.2 billion in today’s dollars). Twenty-five people were killed. Toth says this area will surely burn again.
Uncertainty is when and what other areas are at risk. “The bottom line is the lack of data-driven understanding that every homeowner and business owner faces,” says Toth, 49.
That’s where seven-year-old Toth’s startup Zesty.ai comes in. His company was collecting and using data He. She To train machine learning models to better assess risks from climate change, such as wildfires, on behalf of their customers, most of whom are insurance companies. “We take satellite imagery, we take building permit data, we take local weather station data, and we use artificial intelligence to explain the impact of climate risks on each property,” he says.
There is no shortage of needs. In the Golden State alone, eight of the state’s 10 most destructive fires have occurred in the past five years, according to the California Department of Forestry and Fire Protection, or CalFire. These fires have caused more than $25 billion in insured losses associated with wildfires. But not all property is insured: CRC Group, an Alabama-based insurance wholesaler, estimates that there are $9 billion in uninsured losses from 2018 Camp Fire alone.
The average cost of home insurance in California is $1,177 a year, according to the Insurance Information Institute, up 25% over the past decade, despite the state’s highly regulated market. Industry Association Notes That of the 10 largest wildfires causing the highest average insured losses in California, 8 occurred in 2017 or later, and only 1 occurred in the 20th century. data From reinsurer Munich Re that insured losses from California wildfires in 2017 and 2018 exceeded losses for the entire previous decade. Third notes that “A large part of this loss trend is due to people moving into high-risk areas,” including areas at risk of wildfires. And in recent years, hundreds of thousands of California customers annually—many in rural areas—have Projection entirely by their insurance companies.
Zesty.ai CEO Atilla Toth at a wild and urban frontage near the site of the 1991 Auckland Hills fire.
Cyrus Frivar Forbes
Zesty.ai’s show for insurance companies is: Don’t rely on overly public and outdated maps to decide which buildings and homes to insure. With its data, the company creates a single score, such as a credit score, that assesses wildfire risk on a property-by-property basis. This “Z-Fire” score includes all kinds of information about the home—including its age, materials, roof type, amount of vegetation nearby, and the slope of the adjacent land—that isn’t always captured when insurance companies assess risk. The company is also developing results to assess risks from hurricanes, floods, and other natural disasters.
Zesty.ai operates on relatively small amounts of capital compared to other AI startups. In 2018, the company raised $12.8 million in project financing worth $47.4 million, according to Pitchbook. Earlier this year, the company took out $10 million in venture debt from fintech firm Brex. The company has a long list of popular insurance clients including Farmers Insurance, Aon, MetLife and Berkshire Hathaway. Forbes estimates revenue at about $25 million last year (the company declined to comment on its financial statements, other than claiming that revenue “trebled” in 2021).
The business has been helped by an explosion in the amount of readily available images, thanks to hundreds of new satellites in orbit that collect data as well as the use of drones for aerial photography. Toth takes those photos and combines them with more data: property records, building permits, weather and fire history. Toth says his company’s software is able to quickly render a 3D model of a ceiling based solely on a two-dimensional image with an accuracy of one-tenth of a degree. With all this input, Toth says, his company’s findings can elicit complex questions such as: How likely is this characteristic to be within a disaster zone? If so, how bad was this disaster?
“Artificial intelligence seems like voodoo to some of our customers, so I say, ‘Think of me like a chef,’ he says. ”What goes into the soup? We look at the density of vegetation, we look at the slope, the wind patterns are very important, the distance to previous fires – unfortunately wildfires tend to repeat themselves.”
Making AI stews like this inherently comes with trade-offs, says Mike Lyons, managing director at Boston Consulting Group. Although he didn’t comment on Zesty’s model specifically, he said in an industry survey Forbes That “some are really hard to scale. It’s very hard to implement specific recommendations for your company or for this building.” This is because, he explains, the more characteristics a model covers, the more that model relies on rules of experience and general assumptions rather than facts on the ground. “They must have some kind of reasoning.”
Born in Hungary in 1972, Toth came to the United States in 1995 after graduating from Budapest Business School. He arrived in Chicago, where he earned an MBA from Northwestern University in 2003. After graduation, he spent seven years at EY and McKinsey consulting firms, moving between different corporate offices around the country.
By 2008, Toth was Managing Director of SunEdison, overseeing green energy projects, including the large-scale installation of solar panels at the Staples Framingham, Massachusetts headquarters that provide nearly 700 kilowatts of power for the property. At SunEdison, Toth found himself working again with a former associate of McKinsey, Kumar Dhuvur, with whom he started Zesty.ai in 2015 (Dhuvur is currently Zesty’s Head of Products).
The next big challenge is modeling flood damage, which caused $82 billion in global damage last year.
“When we started this business, it wasn’t an insurance business, it was a business of looking at rooftops and we designed 70 million roofs in the United States,” Toth says. But then came the Tubbs Fire in 2017, which devastated sections of Napa and Sonoma counties, about 60 miles northwest of Oakland. By the time the 23-day fire was complete, it had burned more than 36,000 acres and destroyed more than 5,600 buildings, nearly half of which were homes in the city of Santa Rosa.
The score produced by Zesty.ai for the property is not fixed. Like the credit score can be improved. Firefighting experts constantly tell homeowners about “defensible space,” and the critical need to create a minimum burnable radius around the house. But there are other steps that can be taken as well, such as adding a fireproof roof.
After the company incurred huge losses in the wake of the Tubbs Fire, says William Bates, assistant senior vice president at Rhode Island-based Amica Mutual Insurance Company, it began reassessing its wildfire risk models. Amica, which secures 40,000 homes in California, was using a competing product from CoreLogic in Irvine, California. But as they analyzed their losses, they discovered that CoreLogic had identified some characteristics as low risk for wildfires that in fact weren’t. Amica turned to Zesty.ai and its Z-Fire system.
“If Z-Fire had been used to completely reinstate the entire Amica wallet in California, in 2020 Amica would have prevented 95% of wildlife losses in the state,” Bates said by email.
Currently, California is considering a statewide insurance rule change that, for the first time, would require insurers to offer rates “based in part on reduced wildfire risk from property-level wildfire mitigation efforts undertaken with respect to individual properties.” risk assessment “.
If California enacted a wildfire insurance rule, it would be the first such state to do so – And a great boon for enthusiasm. If it is passed, that means insurance companies will have to work with Zesty or one of its competitors, or develop their own models within the company.
Zesty.ai’s biggest competitor is Jersey City-based Verrisk. Other companies in the space include CoreLogic and Cape Analytics, headquartered in Mountain View, California.
Verisk’s core product, FireLine, has been checked for not being adapted fast enough. Earlier, in 2018, United Policyholders, a national consumer advocacy group, told the California Department of Insurance that the spread of FireLine was “partly responsible” for the state’s “market crisis” regarding homeowner insurance in fire-prone areas. But the campaigners of United Insurances haven’t fully embraced Zesti’s approach either.
The 2017 Tubbs Fire destroyed more than 5,600 buildings and caused economic losses of more than $8 billion.
Justin Sullivan / Getty Images
“My instinct is that Zesty is a positive development looming, but it’s not any kind of magic wand by any means,” said Amy Bach, CEO of United Policyholders.
Zesty hopes that its proprietary – and patented – AI technology will give it a long-term boost. Plus, in part because he’s open about the factors that go into his grade (in other words, he’s not some kind of “black box”) he’s got approval from Insurance commissions in six different states: California, Arizona, Montana, Oregon, New Mexico, and Utah. “That’s a very, very big difference,” Toth says. For example, in California, its model is one of only two approved for wildfire risk. The other belongs to Verisk. By the end of the year, Toth says he expects his company’s models to have regulatory approval in 25 states.
The company is also expanding its model to include more types of natural disasters. Toth says the company already has high confidence in its models for wind damage, hail damage and other types of storm damage. The next big challenge is modeling flood damage, which caused $82 billion in global damage last year according to the Swiss Re Institute, but Toth is confident his company can meet the challenge.
“Insurance used to be a necessary kind of evil, right? I’m paying for it, and I hope I never have to use it,” Toth says. .”
More from Forbes