Newsletters
Risk Central
Digital EditionHow AI is Transforming Insurance: Real Results, ROI, and the Road Ahead
The insurance industry stands at a technological inflection point. After decades of relying on legacy systems, manual workflows, and tightly coupled platforms, carriers and brokers are increasingly turning to artificial intelligence to solve long-standing operational challenges. Yet while the technology offers significant promise, implementing it successfully requires more than just adopting new tools. It also demands a fundamental rethinking of how insurance organizations approach technology itself.
“This is a pretty technologically abused industry,” said Ryan Cantor, Chief Product & Technology Officer at Origami Risk. “It’s pretty old tech. A lot of offline, a lot of on-prem, off-prem. If you want an update, it’s going to cost you six figures to test regression.”
That reality is beginning to shift as AI creates new opportunities for insurers to modernize core processes, unlock efficiencies, and make smarter decisions at scale.
Where AI is Delivering Tangible Results
According to Cantor, two areas stand out as producing the clearest, most measurable impact from AI in insurance today: submission intake and claims processing.
On the front end of the business, submission intake represents one of the most immediate opportunities for AI to drive revenue growth. “The single largest way to increase revenue is to increase your at-bats. That is, to weed through submissions based on your risk tolerance, your profiles, your scope of business, your niche, whatever it may be,” Cantor said. Historically, this process has relied heavily on human labor, but AI excels at converting unstructured data into structured data.
Ryan Cantor, Chief Product & Technology Officer, Origami Risk
“I joke about it: you could give me an application on a napkin, and I could probably turn it into a structured submission and allow you to make computer-smart, intelligent decisions at this point,” Cantor said. That capability enables carriers and brokers to manage submission data proactively, model different risk scenarios, and expand their available universe of business without exposing themselves to unknown risk.
On the other end of the spectrum, claims processing is being transformed by AI as well—and not a moment too soon. “The workforce is aging. I have two children, and neither one of them is growing up dying to be a claims adjuster,” Cantor said.
AI is well-suited to enforcing complex claims guidelines, identifying fraud indicators, and triangulating disparate data sources. That last capability is increasingly critical, as AI is being used by bad actors to manufacture damaged vehicles or embellish hurricane damage on homes.
“Not only from a detection standpoint, but triangulating sources of data—what was in the written statement, what was in the photograph, the direction of the house or the direction of the car that didn’t quite align with the other car,” Cantor said. “It’s surpassing individual human capability.”
The alternative to AI, he noted, is adding more human checks and balances, which will drive up costs without necessarily improving quality. “AI is great at needles in haystacks. If you have a 600-page internal guide or training material, AI will follow it if you train it properly. It doesn’t talk back, doesn’t take vacation, doesn’t want to retire.”
The ROI Question: A Payback Period, Not a Question of Value
For insurance leaders wondering whether the investment in AI is worth the cost, Cantor’s answer is direct: the ROI is there, but organizations need to approach it as a strategic investment rather than an immediate payoff.
“The ones who are doing it and are committed, like anything, we’re a core system. People make an investment in the core system for a future payoff,” Cantor said. “When we talk about AI, it’s somehow, now they’re gun-shy. Well, there is a future. You do have to train it. You have to dive in. For a little while, you’re gonna carry your old cost. Because with AI, first you try, then you trust.”
Organizations that have committed to the investment are already seeing results. “The ones that do, or that were early movers, are absolutely seeing measurable ROI. It’s a non-question that the ROI far surpasses the investment.”
The benefits extend beyond insurance-specific use cases as well. Data migrations and system conversions, which are historically expensive, time-consuming projects, can now be dramatically accelerated.
“If I showed you the source data and it knows Origami Risk or whatever system you’re going to, and I know what date formats it needs to be, and I know what the schema is—that would historically have been millions potentially for some customers. Mapping the data, humans looking at it, transforming it, testing it, refining it. And AI can do it in a couple days.”
The Talent and Mindset Challenge
Despite the clear potential, implementation obstacles are real and they often have less to do with technology than with people and process. Cantor points to a shortage of insurance-minded professionals who are open to new ways of doing things.
“Our organizational challenge isn’t necessarily finding insurance-minded individuals. It’s finding insurance-minded individuals who aren’t so ingrained that there was only one way to do it,” Cantor said. “Being open to the idea that there is AI or different kinds of cloud technologies or serverless environments or queuing or more modular API-first approaches.”
The result, when the wrong people are brought in, is predictable. “You hire the same company or you hire the same person, and you kind of get what you’ve always got. And you might get a new system or you might get a new implementation, but it’s still suboptimal because it didn’t really leverage all the technical capabilities that are now available.”
The problem is compounded by internal resistance to change. When a CEO sets a clear direction toward technology transformation, the people closest to the existing processes often push back. “You have an agency problem,” Cantor said. “You’re a CEO in an insurance company, you’re hearing from your internal team: they can’t do this, and they can’t do that, and you’re getting all the scare tactics.”
Cantor also warns against a superficial approach to AI adoption, i.e. simply layering it on top of existing broken systems.
“The times where you’re just connecting all your crappy systems together so that you could now pay a tax of AI to ask it a question to pull all that data together…that’s an IT tax. That is: You have really bad technology, and now you’re paying more money on top of your money just so you can get them to talk to each other with AI in the middle of it.”
Failed implementations, in his view, often don’t have to fail. “They failed because they were putting a square peg in a round hole as if the hole had to be round. You could change the hole to be square. It’s your company—do anything you want.”
Keys to Success: Focus on Outcomes and Embrace Modularity
For insurance leaders looking to succeed with AI and modern technology, Cantor offered two key takeaways.
First, focus on the outcome, not the journey. “When you’re evaluating any technology, focus on the output. Don’t focus on the journey. Every software is going to be different and just be open to that ride,” Cantor said. “What people get in trouble is they say, here’s the output I want, and then here’s what every single step of that workflow needs to look like. Then what you get is going to look a lot like what you had.”
Second, and perhaps most importantly, embrace a modular approach to the insurance tech stack. That’s the bet Origami Risk is making with its own platform, prioritizing drag-and-drop workflows, an API-first mindset, and the ability to plug in and connect to virtually anything.
“I’ve been at Origami Risk for two and a half years, and I haven’t seen even within the same line of business, two people want to do the same thing,” Cantor said. “Tightly coupled policy administration systems or rating systems or form generation or rules engines inevitably leads to constraints.”
That modularity extends to AI itself. “Today, you might use Origami Risk’s out of the box AI, but you and I see the evolution. In five or ten years, every company may have their own AI models,” Cantor said. “If you wanna be able to bring your own AI model, you should have the flexibility and the modularity to be able to plug and play.”
Looking ahead, Cantor sees the insurance industry moving decisively away from the closed, tightly coupled ecosystems of the past. “The world now is a little bit more modular, a little best in class. It doesn’t mean you’re being asked to just switch to a bunch of systems, but you are bringing data and insights and AI and specialty tools from a variety of places. And being able to kind of orchestrate that centrally is going to be really, really important.” &
