AI might be helping insurers streamline operations, but the real measure of its success lies in whether clients actually benefit, according to Jonathan Shapiro (pictured), founder at Optalitix.
“It’s really about the client value,” Shapiro said. “If the efficiency gains are going to benefit the client, then… obviously there’ll be cost savings for the insurer.” But for AI to genuinely improve outcomes, it has to drive better underwriting, fairer pricing and faster decisions - not just margin gains.
He pointed to a common concern among clients: that AI will make opaque decisions or penalize individuals unfairly. “I’d imagine for clients, that’s probably one of the biggest fears - is AI going to select against me?” he said. “Is it going to be unfair?”
The answer lies in transparency. “That’s one of the big challenges with AI,” Shapiro said. “Is there transparency?” With some of Optalitix’s insurer clients, including Vitality and Pool Re, auditability was built into the workflow. “They can go into our system, and they can see what data’s come from there. How has it been used? How has it been priced?” he said.
As risk landscapes shift, the pressure is growing on insurers to modernize the models they use. For Shapiro, legacy frameworks can’t keep up with emerging threats like cyber and climate.
“Using old models is just not going to cut it in the new types of risks that are out there,” he said. Data remains the fuel that powers this evolution. “You don’t have the correct data, you don’t have it in the right format, you’re not able to deal with large amounts.”
Whereas insurers a decade ago had limited access to real-time data, today the situation has changed dramatically. “There are satellite systems these days that will… give you hundreds of historical data points and predictive data points,” he said. For underwriters and actuaries to use that data meaningfully, the AI tools handling it must be trustworthy. “You need to trust AI has summarized it correctly,” he said.
AI should not be used to replace experienced decision-makers. “You don’t want it to make conclusions,” he said. “You want it to assist someone with lots of experience in making a better decision.”
He also predicted broader use of simulation tools and next-generation stress testing. “Cyber is a great example,” he said. “The risks today are completely different to what they’re going to be … never mind next year, next week, next month.”
Despite the potential of AI, legacy systems continue to drag insurers down - and ripping them out overnight is rarely an option. “That is the bigger paradox in insurance,” Shapiro said. “For every technology company selling into insurers, that’s the biggest challenge.”
Convincing teams to abandon 20-year-old models is rarely successful. “They like what they understand,” he said. “To tell them to rip it up and rewrite Excel spreadsheets in Python and take a two-year change program is a very difficult message.”
Optalitix’s approach focused on staged upgrades rather than full replacement. “We’re all about phasing,” Shapiro said. “How can we get some value soon, quickly and incrementally?”
Compliance pressures have helped accelerate change. “The regulators coming in and saying, we need auditability, we need traceability - that’s really helped,” he said. “If you’ve got 4,000 spreadsheets saved on a shared drive somewhere, it’s almost impossible to say who changed it.”
Instead of eliminating spreadsheets, the firm reimagined them as a programming layer. “The spreadsheet just becomes the programming language for the actuary - then it’s OK,” he said.
For some insurers, there’s a temptation to ditch everything at once and rebuild with digital tools - but Shapiro warned that’s not always practical. “You’ve got to pick your battles,” he said. “It’s not technology for technology’s sake. It’s technology to make the process better.”
He cited use cases where a flashy front end actually hurt usability. “If you’ve got to paste in… 500 rows of a list of properties that you’re insuring… to move to typing that in on a web page is madness,” he said.
This kind of overreach often breeds mistrust between actuarial and engineering teams. “The guy that understands the pricing doesn’t understand the programming,” he said. “So you get this mistrust.”
Shapiro said many insurers who went all-in on a two-year digital overhaul had little to show for it. “They come to the end of it and go, we’ve actually achieved nothing. We’ve spent a lot of money,” he said.
A better approach? Build in phases. “Start with the West Wing,” he said. “But put it in the West. Don’t put it in the middle so that you’ve got no space for the rest. But you don’t have to build the whole house.”
