Free Preview

This is a preview of some of our exclusive, member only content. If you enjoy this article, please consider becoming a member.

As insurers have become more comfortable using AI, there’s a challenge beyond simply learning the technology: learning how to secure it.

“I think our job right now is to leverage the AI capabilities in a way that is secure and properly governed,” said Tim Hardcastle, co-founder and CEO of InsurTech INSTANDA, speaking with Carrier Management at InsurTech Connect in Las Vegas this year. “That’s the number one priority for us.”

While plenty of opportunity exists for insurers with the use of AI, the same tools that make technology more accessible also create new exposures. AI agents can handle tasks and move information, he said, raising concerns about sensitive data reaching unintended destinations or actions occurring without a user’s knowledge.

“Unless it’s managed, controlled and governed, you could end up with some risk exposure,” he said. “Whether it’s personal information or in a company environment, you could lose customer data.”

Making AI capabilities available to insurers in a secure, governed environment should be a top priority for insurers, Hardcastle said. However, that reflects a challenge for carriers as they work to unlock the productivity potential of AI without giving autonomous systems unrestricted authority.

“How do we take this amazing, incredible power that’s now available to us and put it out there to our clients and prospects in a way that they can trust us and know that it’s well managed?” he said.

Another challenge may be the pace at which insurers can adapt. Although AI capabilities are advancing rapidly, implementing changes across a workforce requires training, communication and leadership support. Employees in established roles may have to learn new ways of working, even as some tasks become candidates for automation, Hardcastle added.

“You’ve got people who have worked in a certain way for many, many years and they’re having to then learn how to work with AI,” he said. “But change management historically within any industry, not just insurance, is not something that happens in hours or days. It takes weeks. You have to do a big job on communication. The leaders have to come out and talk about why we’re changing, and you have to get people on board. Change takes weeks and sometimes months at the human level.”

He gave the example of a high-net-worth customer’s coverage that might encompass multiple properties, vehicles, yachts and valuable artwork. As another example, a painting’s stated value could require photographs, authentication records or a physical appraisal before an underwriter can assess the risk. Those steps cannot necessarily be compressed into a fully automated transaction, he said.

“Every policy is different,” he said. “So it’s interesting opposing forces, really, of the increasing AI capabilities and then the ability for organizations to change as quickly, aside from the governance and security concern that runs in parallel.”

Hardcastle described agentic AI as a continuation of INSTANDA’s founding approach to insurance technology, enabling insurers and managing general agents to build and modify products without depending on IT teams for every change.

“We always took the view that insurance is really quite complicated,” he said. “It’s very diverse. There are lots of different ways to do insurance.”

That complexity led INSTANDA to develop a no-code interface intended to let insurance professionals configure technology more directly.

“It was our view that no individual software company could ever figure out how that would all work and then make it into an application that would then serve everybody in the U.S. and Europe,” Hardcastle said. “So we figured out that to meet the needs of insurance, you had to separate the application hard work, which is the coding, and then create a layer which allowed people to interact with the application and do it as if they were in a Microsoft Office environment where you don’t need any training and you can manipulate things very easily. That’s why we invented the no-code interface.”

With AI, the company sees an opportunity to extend that approach, allowing users to work with prompts and agents rather than needing deep technical expertise.

“Like everybody, we’re all using AI,” he said. “You don’t need any real deep technical skills, do you? You just need to be able to prompt well and then maybe get some agents working for you and be able to actually think about how to orchestrate some of the activities. So it’s a very different world we’re now in.”

This is particularly pronounced in complex commercial and non-admitted insurance, where INSTANDA sees a role for its Clear platform designed to coordinate back-and-forth processes and move work among brokers and underwriters. Hardcastle said INSTANDA is now bringing agentic AI capabilities into that platform, with a division of labor that would reserve final authority for an experienced human underwriter. The objective is not to eliminate judgment from complex risks, but to reduce the manual coordination surrounding it, he said.

“What we’ll be able to do is put in an agent that will behave like a junior underwriter,” he said. “It won’t make the final decision, but it will act as if it were someone as part of the process, and it can interact with the broker.”

For carriers considering AI at scale, Hardcastle’s message was that broader access alone is not enough. Successful deployment will also depend on controlling data, keeping underwriters accountable for decisions, helping employees adapt and making sure the cost of automation is justified by its value. With all of that in place, he said that the company feels positive about the future of AI in insurance.

“I think we feel pretty bullish about the future because having been one of the only companies that figured out democratization was the right way to go, for us using AI capabilities, we’re just accelerating down that pathway and making more of that capability available to the insurers and the MGAs,” he said. “But we’re doing it securely and in a way that’s governed.”