More than half of US states have now adopted the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, requiring carriers to document how AI supports underwriting decisions rather than replaces them. For Mike Maletsky (pictured), vice president of underwriting transformation at Hiscox, that distinction is not a compliance footnote. It is the operating principle behind how the specialty insurer is deploying AI across its underwriting business.
"At Hiscox, when we use AI, we want it to augment our underwriting experience, not replace the underwriting judgment," Maletsky said. "We use AI to eliminate some of the busywork, some of the administrative tasks. We're not automating the decision-making, and the human accountability, an underwriter saying yes or no, is paramount to what we are trying to do."
Maletsky, whose role spans transformation across Hiscox's USA's product lines, described the shift in underwriting over the past five years as a change in visibility rather than authority. Where underwriters once worked with narrow slices of information, AI tools now surface a far broader set of data points for the same risk.
"Five years ago, underwriting in the specialty space was like going around a house with a flashlight, looking for things in a very narrow scope of vision, trying to find pieces of information," he said. "What AI has helped us do is really shine a floodlight on the entire house and let us see more corners, more nooks and crannies, get into all the places."
That expanded visibility is translating into faster quotes and lower servicing costs, he said, without adding headcount, a trend explored further in Insurance Business's report on Cowbell's AI-native underwriting system. "Lower friction means faster quotes, faster responses, and a better overall experience, and probably most importantly, it helps us handle higher volumes of business without sacrificing human judgment, that human-in-the-loop component," Maletsky said. " As we scale, technology can help us manage growing volumes more efficiently and drive productivity gains. "
Unlike personal lines, where underwriting can follow largely templated rules, specialty risk is built on the premise that no two accounts are identical, a structural feature Maletsky said limits how far automation can go, a theme echoed in Insurance Business's look at how K2 Insurance Services is rethinking underwriting. He compared AI's role to an in-car navigation system.
"I think about the old GPS, the handheld ones you'd plop on your dashboard," he said. "It tells you where to go, it tells you traffic conditions, but it's not driving the car for you. We view AI like that: an assistant right there alongside you to make underwriting decisions faster, remove friction, lower the cost to serve, and handle higher volumes."
"Technology is seen as enhancing the process rather than automating end to end." Maletsky said. "There are too many differences between insureds. There's only so much you can glean from data points along the way." He also pointed to unresolved liability questions around AI-generated content, including hallucinations, copyright exposure and privacy concerns, as reasons Hiscox keeps a human decision-maker in place on every account.
For Insurance Business America's audience of brokers, agents and underwriters, Maletsky argued that freeing underwriters from administrative work should strengthen, not weaken, the relationship side of the business, a point underscored in Insurance Business's coverage of AI-powered underwriting from submission to decision. Insurance, he said, remains fundamentally a trust-based trade between underwriters, brokers and insureds.
"AI is letting us do that more, because it's taking off the administrative burden on underwriters, freeing up more time for them to talk to the broker, talk to the insured, and make sure we're giving them the coverage they need," he said.
"You don't go to a restaurant because their reservation system is really good. You go because of the experience, the food, which equates to the insurance policy in this analogy."
Maletsky said the skill set Hiscox looks for in newer underwriters is shifting away from information-gathering and toward critical evaluation of AI-supplied data. Early in his own underwriting career, he said, a large share of his time went toward locating the specific application fields that mattered for a given product's guidelines.
"It's less about that and more about decision-making," he said. "Critical thinking becomes much more important. It's no longer a simple rules-based process, get the information, plug it into a formula, and there's your output. Now it's: Is this information reliable? What's missing? Does it make sense?"
That includes a willingness to challenge the model's conclusions rather than defer to them, he said: "At some point, a human is going to need to say, I don't trust the output of the model, to challenge that output, to be willing to disagree with it. That critical thinking skill will be very important going forward."
Maletsky added that as AI absorbs more administrative work, communication skills, from relationship-building to email correspondence to picking up the phone, will matter more for new entrants to the specialty market, not less.
