InsurTech Ohio Spotlight with Hal Northrop
Hal Northrop is the Vice President of Client Development at ZestyAI, an AI-powered predictive property and climate risk platform for insurers, powering best-in-class risk decisions for leading property and casualty carriers in the United States. Hal was interviewed by Andrew Daniels, Founder and Managing Director at InsurTech Ohio.
Hal, what are some of the profound impacts that artificial intelligence (AI) is having on property underwriting?
“This is a great starting point. Climate risk is having a massive impact on the insurance industry in the US. Traditional risk models or stochastic models simply are not able to accurately predict risk with these perils. Over the last decade, the damage from secondary perils (think wildfire, hail and wind) have actually exceeded that of hurricane damage. This is a significant shift in how to think about risk, so artificial intelligence (AI) is an important tool for forward-thinking insurers to attack the risk and help adapt today's underwriting to tomorrow's climate risk.
Let me break it down. When you think about stochastic models, they will generally analyze risk over a portfolio, region or zip code. Artificial intelligence allows for address-specific, property-specific, policy-by-policy risk-splitting.
What this means is that AI is valuable not only to the insurance company but also to the insured. A tangible example would be a homeowner who has a flawless roof, no debris in the yard and no overhanging vegetation. They’re going to get an attractively-priced premium on that property while a neighbor who may have neglected their property is going to be paying a higher price to get coverage.
That leads me to the final message here, which is transparency. AI is truly enabling an impressive capability of transparency to the policyholder, which is unique in the world of getting insurance coverage. Historically, somebody would type in an address, go into this black box, receive a score and then a premium. Now, with AI, you get that precise information about the property, but it'll also say, ‘Hey, here are the top three reasons why this property is getting a good rating and premium.’ And then equally, ‘Here are the top three reasons why you're paying a higher price and what mitigation efforts you can undertake to improve that premium.” What that does is it provides an opportunity to engage with the client and provide them a better understanding of the reasons behind the pricing, allowing them to digest that and take mitigation action, if necessary.”
Can you share some results that back up the industry's optimism around AI?
“ZestyAI gives clients the ability to make faster and better decisions by augmenting traditional data with hyper-local data. We use aerial imagery, permit, transaction, weather and IoT data, combined with artificial intelligence to turn more than 200 billion data points into comprehensive digital records and property-specific risk scores. AI allows for precision but at scale.
Let me give you some practical examples. One of our national carrier clients wanted to analyze their book to understand the risk levels of about a million properties. In the world of underwriting, realistically, you might be able to inspect one to two percent of your book in any given year. In contrast, we were able to analyze a million properties, in a relatively short period of time, without any IT integration.
In collaborating with the carrier, we identified the items in their book that they felt were most important to understand. They wanted to understand things like the presence of a swimming pool, a trampoline, a secondary structure, a poor roof or a tarp on a roof. Around 12-15% of homes in the United States added a secondary structure during COVID. We were able to detect all of this, and of the million properties, we were able to take it down to about two percent that needed a closer look. Of the two percent identified, more than 60% resulted in some type of action, meaning the carrier contacted the policyholder, and they were able to get appropriate coverage for their property. That was all completed inside a 90-day window without any IT integration.
The other example I'll give you is in a wildfire-prone state. One of the carriers had effectively stopped writing business in California because traditional models didn’t give a complete view of wildfire risk. Through our AI-enabled property risk assessment, and our Z-FIRE model that was trained on more than 1,500 wildfire events across 20 years of historical loss data, we were able to provide a level of detail that allowed the carrier to write roughly 30,000 new policies in the state. It's particularly important in California because it has its own challenges as a state. The carrier was able to then generate a significant amount of premium, and the homeowners were able to get coverage where they historically have been deselected.”
What should those investigating AI and modeling solutions understand about how one provider might differentiate from another?
“When it comes to AI modeling, all of the providers are not the same. Insurers need to partner with providers who can provide data on the greatest number of policies possible. ZestyAI uses multiple sources of imagery because insurers need that coverage, not just in the metro areas but also rural areas.
Insurers also want to make sure they not only have hyper-local information and intelligence for any given property, but that needs to be complemented with valuable climate knowledge. Insurance Institute for Business & Home Safety (IBHS) has been an incredible partner for us. From a meteorology perspective, it's important to understand how a given property is going to act in an event like a wildfire or a severe convective storm. Having that combination of high-tech AI and specialists who understand the potential impacts that a catastrophic event could have on a given property is something unique to ZestyAI.
One final thing: a strong regulatory presence is important in the underwriting world. It's not only about being able to accurately assess risk, but being able to differentiate risk, split it and then explain it to the various departments of insurance. They need to understand why you're asking for a certain rate.”
Do you have any predictions on what AI will accomplish in the years ahead?
“One of the things that’s going to occur more readily is something I mentioned earlier, which is transparency. AI is going to allow insurance companies to communicate with homeowners on what specific actions they can take to lower their property’s risk. I'll reference a study that we did jointly with IBHS: We analyzed over 70,000 wildfire-exposed properties, and we found that the homeowners who cleared vegetation and had defensible space around their homes were able to nearly double the likelihood of that property surviving a wildfire. This goes back to this ability to provide hyper-local data to the carrier, which then can be communicated back to the policyholder to explain the property-specific actions they can take in order to mitigate risk. Either the uninsurable become insurable, or the insured reduce their current premium.
The insurance industry is unique in that there's motivation for both the insurance companies and the policyholders to reduce risk and loss. People want to protect their homes. AI can be a significant enabler to not only providing and improving customer experience, but ultimately, it can reduce the actual dollar value of losses experienced through these secondary perils.”