Generative AI is helping figure out who is riskier for insurers

Sixfold, which is using generative AI to speed up the insurance underwriting process, has raised $6.5 million in seed funding led by Bessemer Venture Partners and including Crystal Venture Partners, the startup exclusively told Semafor.

Sixfold’s customers, including Builders & Tradesmen’s Insurance Services (BTIS), can upload their own underwriting manuals and proprietary data to its software system, which the company then uses to make recommendations as underwriters review new applications.

If a roofing contractor applies for insurance from BTIS, for example, Sixfold could cite the company’s guidelines indicating roofers are riskier to insure than painters, and suggest the underwriter quote them a higher monthly premium.

“We make recommendations back to the underwriter based on that underwriter’s own rules,” said Sixfold founder Alex Schmelkin.

Sixfold also incorporates data from outside partners like BuildZoom, which tracks millions of building permits across the U.S. Sixfold can use that information to flag, for instance, that a roofing contractor has also frequently received permits for electrical work.

Sixfold is part of a wave of startups racing to apply the technology behind ChatGPT to regulated industries, like healthcare and insurance. The last time Silicon Valley tried doing that, it didn’t exactly go well.

Over the last decade, journalists and researchers have repeatedly shown that automated tools used in those sectors were biased against poor people, racial minorities, and other groups. But Schmelkin argues Sixfold won’t face the same problems because its models can cite their sources and don’t operate like a black box.

“That’s precisely where the prior attempts at this have failed. People would show up and they would say, ‘I know better than the underwriters,’” Schmelkin said. Earlier competitors, he explained, often designed proprietary tools that would analyze a dataset and spit out a recommendation without explaining how they determined it.

Sixfold isn’t trying to replace the evaluation processes used by giants like Prudential or State Farm; it just wants to make it easier and quicker for underwriters to parse disparate sources of information.

Startups that promised to transform the insurance industry have largely failed, The Wall Street Journal reported earlier this week. The article centers on new insurance providers like Lemonade and Hippo, but the kinds of challenges they faced could also impact Sixfold and other software firms. For example, the need to navigate strict regulations that vary from state to state.

The Colorado Division of Insurance proposed an expansive set of rules earlier this year that would force life insurance companies to disclose the AI models they use and submit them to strict oversight measures. The effort is being closely watched, and could influence how other states choose to regulate artificial intelligence more broadly.

• 91% of life insurers reported having fully or partially-automated underwriting programs in place by 2021, according to data from the insurance industry research firm LIMRA as shared with Bloomberg.
• A lawsuit filed late last year accuses State Farm of using AI to cloak “old racial discrimination tactics in a veil of technology,” The New York Times reported.