“I’ve seen a good amount of coverage on implementing AI. But what I haven’t seen a lot about is leading” AI programs.

That’s how Carrier Management’s Guest Editor Michael (Fitz) Fitzgerald started a conversation brainstorming the focus of this section of our magazine—”Leading the AI-Powered Insurer”—a theme that fits squarely within CM’s overall mission of providing content about management and leadership of property/casualty insurance carriers.

The idea for the theme “came out of work that I was doing with boards of directors around controlling and governance,” said Fitzgerald, referring to his work as an Insurance Industry Advisor for SAS Institute, Inc., an analytics and software solutions provider. The theme evolved as Fitzgerald and CM’s content team interviewed executives like Pina Albo, the chief executive of Hamilton Insurance Group, Matt O’Malley and Steve Stabilito, two underwriting leaders at AXA XL, as well as Henna Karna and Ursuline Foley, who both serve on the boards of directors of P/C insurance and reinsurance organizations.

“As we talked to more people and we went through the interviews, this idea that I’d been coached on, that I’d heard of in the past—that leadership exists at all levels of the organization—really came up,” Fitzgerald said. “I hope what this [magazine] issue does is to draw out exactly that fact—that the leadership of the AI-powered insurer is at the board level, it’s at the executive level, but it’s also [at] the middle management level” where people like the AXA XL executives “are working really hard on data and incorporating that into the business. That’s very, very important. And how do you support those efforts?” he asked, offering a question for P/C insurers reading the article, “How Underwriters Win Business with Data and ML at AXA XL” to ask themselves.

Fitzgerald continued: “I feel really good about the issue in terms of teasing out what board members and executive teams need to be focusing on. And I think largely they are. I feel very good about the middle management piece…Look for the people who are trying to push these efforts forward, and especially the supporting efforts around data. Recognize those efforts and celebrate those wins,” he advised.

“But there’s a big hole in the majority of most organizations,” he said, noting that he was troubled by the fact that talent leaders at traditional insurers seemed to be out of the loop on AI-focused activities.

Another theme of this edition of Carrier Management is talent management, and when CM Deputy Editor Elizabeth Blosfield tried to blend the “Leading AI” theme together with the talent theme for her article, “Are Carrier Talent Leaders, Workforces Prepared to Wade Into the AI Tidal Wave?” she could only find one representative of an traditional carrier to interview. “Most of the carriers I reached out to thought it was an interesting topic, but said they didn’t have anyone who could speak to it or haven’t invested in AI in that capacity,” Blosfield reported.

“With all of the money and mindshare being invested in AI, with all of the risks and opportunities, why isn’t AI upskilling a Q1 2024 priority for HR?” Fitzgerald asked, highlighting a “troubling gap” that may exist in the leadership of AI programs of P/C insurers.

Fitzgerald said he was surprised in a positive way to learn about the diligence of the nominating committees of boards of directors of insurers in adding members with technology skills to their ranks. “Who would think that people looking for board members would be touched by AI?” They are. “They have recruit with this whole idea of diversity in mind,” Fitzgerald said, referring to a focus on diversity of thought and experience described by Hamilton’s Albo as she spoke about the selection of Dr. Karna, a technology entrepreneur with advanced degrees in applied mathematics and business administration, and broad industry experience at Verisk Analytics, AIG, AXA XL, and Google.

“The leadership of the AI-powered insurer is at the board level, it’s at the executive level, but it’s also at the middle management level.”

Michael (Fitz) Fitzgerald, SAS Institute

The elevation of AI to the board nominating committees, “really drives home” the fact that “there are going to be very few places” within insurance companies that won’t feel the impact of AI.

Everything in Not Advanced AI

As Fitzgerald and CM editors pulled together the content for the edition, he stressed the need to drive home another point. “People are calling everything AI” but he believes the focus of leaders’ risk management efforts should be on the types of advanced AI called out in a definition offered by members of the NAIC Big Data and Artificial Intelligence (H) Working Group when they surveyed auto and home insurers on their use of AI in 2022 and 2023. (See accompanying textbox) Paraphrasing the definition, he said, “It’s the use of technology in a way that doesn’t require the intervention of a human. In other words, a computer that can program itself, a computer that can find insights in data without the human pointing to it.”

“To me, that [definition] gets you away from some of the simpler and less risky issues or discussions of the technology,” he said. “Using AI, even if it’s a natural language processing ChatGPT-like large language model to do the submission intake summarization” doesn’t fit that definition. “You’re not going to get goosebumps thinking about what could go wrong.”

“What the NAIC definition does is it goes to an area that people generally don’t want to think about because it is scary [to think] what if we have this agent, this employee, this thing that’s in there working on its own?”

“The value of that definition to me is that it says these are computers doing things on their own.”

Fitzgerald confirmed that definition in the Model Bulletin on the Use of Algorithms, Predictive Models, and Artificial Intelligence (AI) Systems by Insurers crafted by the NAIC Innovation, Cybersecurity and Technology (H) Committee and adopted by the NAIC in December, encompasses both types of AI. In states that adopt the bulletin, regulators will be expecting insurers’ risk management and governance programs to address predictive algorithms and advanced AI.

Still, insurers need to appreciate the difference. “We’ve lived in the world of predictive modeling since the 90s, and while it is mysterious and magical in its own way, it still is very statistically based. You start at A, and you get to B, and you can tell how you got there,” Fitzgerald said. In the new world, “you start at A, and you end up at C, and it’s not always evident or even explainable to say how the machine got there. That’s the piece that people need to focus on when they think about Leading AI. [It’s] not something that’s deterministic. You’re talking about something that you might not even ever be able to explain. …It might be absolutely the right answer, but you don’t know why.”