The AI Leapfrog Coming for Insurance Underwriting

In the late 1680s, a man named Edward Lloyd opened a coffeehouse on Tower Street in London. The place filled up with sailors, merchants and ship owners swapping news about which boats had gone down in a storm and which had come back loaded with sugar. Somewhere in that smoky room, over bad coffee, the modern insurance industry was born. People figured out how to put a price on catastrophe. They learned to look at a rotting hull bound for the West Indies and say, with a straight face, “I’ll take that bet for three percent.”

Three hundred and forty years later, the bet is still on. The coffee is better. The paperwork, somehow, is worse.

This is the strange thing Jonathan Crystal keeps bumping into. Crystal runs Crystal Venture Partners, a firm that backs startups working inside the guts of the insurance business, and he spent two decades in brokerage before that. He recently sat down with Karen Webster at PYMNTS to talk about what’s broken, what’s fixable, and the possibility that the oldest risk-pricing industry on earth might be about to leapfrog everybody.

60% of the Mail Never Gets Read

Crystal told Webster that for 20-plus years and untold billions of dollars, insurance has been trying to go digital. Paper became PDF. The fax became an email attachment. The filing cabinet became a server. And yet: the average underwriter, the human being whose job it is to look at a risk and decide whether to insure it, only gets to about 40% of the submissions that land on their desk, he said.

The other 60% of the work sits there in a big stack waiting to be read, and eventually expires or gets declined by default. An entire industry has been busy digitizing its inbox without actually reading the mail.

Crystal calls this “the industrial model of production,” which is a fancy way of saying: we put the old assembly line inside a computer and called it a transformation. It wasn’t. The paper stacks just got thinner.

The Leapfrog

And this is where the story takes a turn, because the thing about being late is that sometimes you get to skip a step.

Think of the countries that never built landline networks and jumped straight to mobile phones. Places that never had a banking branch on every corner and went straight to payments on a screen. They didn’t have to un-build a legacy system. They just started fresh.

Crystal sees something similar happening now in insurance, but with artificial intelligence. The industry that couldn’t quite get its act together to become digital might accidentally become smarter instead.

His thesis is this: AI can read. A lot. Fast. You hand it a thousand-page litigation file or a patient’s medical history going back 20 years, and in roughly the time it takes to refill your coffee cup, the technology has pulled out what matters. The signal inside the noise. The detail an underwriter would have flagged on page 847 if they’d ever gotten to page 847.

Right Pricing, Not Fast Pricing

Crystal is careful not to oversell this. He doesn’t think AI is going to replace the pricing models insurers have spent centuries refining. What he thinks it will do is execute those models properly for the first time. All the risk has been there. Most of it just wasn’t being priced correctly because nobody had time to look.

He calls this “right pricing.” Not faster pricing, not cheaper pricing. Just pricing that actually reflects what’s in the file.

Then he goes further. Imagine, he says, a world where risk is priced not once a year at renewal, but continuously, every time a new piece of information becomes available. A customer adds a warehouse. Their rate adjusts. A region’s wildfire data updates. Their rate adjusts. The model is always on. “Fully autonomous,” in his words.

A Conversation, Not a Letter

That’s a big swing. It would reorganize the way customers relate to their insurance. And let’s be honest: that relationship is currently defined by confusion, irritation and the suspicion that you’re being overcharged for reasons nobody will explain.

Crystal thinks can AI change that too. Instead of a letter that reads like it was drafted by a committee of attorneys in 1978, you’ll be able to ask your insurer, out loud, in plain English: “Why is my coverage priced this way?” and “What could I do to pay less?” And get an answer that makes sense. On the spot.

Where the Humans Still Win

Crystal is not a techno-utopian. Ask him about the limits and he gets specific. There’s a whole category of risk where the data is thin, the history is anecdotal, and the only real tool is human judgment. The weird stuff. The long-tail stuff. The one in-100-years event. Earthquakes in places that don’t usually have earthquakes. Novel cyberattacks. New drug classes. The rare, the unprecedented, the genuinely strange. Those still need a person in the room.

The coffeehouse, in other words, isn’t going anywhere. It’s just getting an AI assistant.

Where Crystal Ventures Is Putting the Chips

The firm’s portfolio traces the thesis. Sixfold AI helps commercial underwriters get through their pile. With Coverage is a brokerage platform. Roe is trying to make the customer-facing side of insurance feel less like a DMV visit. And Charter Space is pointed at financial infrastructure in newer industries, like space.

Ask customers what they hate about insurance, and they’ll rarely lead with price, Crystal said. They’ll lead with saying it’s slow, the service is bad, and nobody explains anything. Three hundred and forty years of practice, and the industry still hasn’t figured out how to be pleasant and understandable.

That, Crystal said, maybe more than any actuarial model, is what AI has a real shot at fixing.

The Twist

The business that was born pricing catastrophe (fires, shipwrecks, plagues) spent two decades failing to manage a much smaller catastrophe — its own paperwork. Now, just as it was finally getting around to the job, something bigger arrived.

And the companies that were behind on email might end up ahead on intelligence.

“When the world changes,” Crystal told Webster, “a lot of things change together.”

Who knows, maybe somewhere Edward Lloyd is nodding into his coffee.