Some are standing in the station. Others are already on board. Property/casualty insurance carriers are mapping out itineraries as the AI train speeds across this industry and the ones they insure.

A year after ChatGPT exploded onto the scene garnering 1 million users in just five days, and a month after the NAIC’s adoption of a “Model Bulletin on the Use of Algorithms, Predictive Models, and Artificial Intelligence (AI) Systems by Insurers,” CM Guest Editor Mike Fitzgerald asked industry participants and observers whose opinions he values what all this means for property/casualty insurers.

How will generative AI and advanced analytics impact the insurance industry?

Pina Albo, chief executive officer of Hamilton Insurance Group, the newest public insurance company in the P/C insurance industry, revealed that her company is already using generative AI to streamline underwriting workflows. “But you don’t have to invent every mousetrap yourself,” advised Albo, who confirmed that embracing technology and analytics has been a core tenet of Hamilton’s strategy since inception.

“Everything we do, we think about how can technology and data make us better, make us stronger, make us faster. And over the years we have both invested in proprietary technology tools to enable our business, specific to our business, but we’ve also partnered with some vendors,” she said.

In particular, she noted that Hamilton teamed up with an AI vendor a few years ago to provide a tool specific to the property insurance area. Hamilton insures Fortune 500 and Fortune 1000 companies that have very complex buildings at multiple locations. “We use AI there to read the very extensive and complicated engineering reports and provide us scores. We humans could not do that at the speed and actually at the accuracy that this tool does for us. So, there we find it a real competitive advantage,” she said.

“Every single function will be touched by AI, generative AI and advanced analytics. Every single function,” said Jonathan Kalman, a partner with Eos Venture Partners, a venture capital firm exclusively focused on the insurance industry.

“There is technology right now that has the capability to essentially watch a workflow, reflect on a workflow, understand a workflow and optimize the workflow in ways that we can’t even imagine.”

Jonathan Kalman, Eos Venture Partners

“On a simple level, it could just be using those analytics to develop customized workflows that better serve segmented populations.” AI has “an ability to create micro-optimization of processes in ways that are infinitely scalable,” he said. Contrasting the days of pasting stickers on whiteboards and rearranging them to simplify main core workflows with the new world that’s opened up through AI, Kalman said, “There is technology right now that has the capability to essentially watch a workflow, reflect on a workflow, understand a workflow and optimize the workflow in ways that we can’t even imagine.” Insurers can leverage this ability “to surface and extract data, organize it and make meaning out of it, and when directed through appropriate prompts, to create ways to synthesize improved outcomes,” he said.

This will reverberate through “every aspect of the business”—even the finance function, he said. Giving a heads-up to finance executives, he said, “It will impact reserving. It will impact treasury. It’ll impact investment management. It will impact your balance sheet allocations. It’ll impact your understanding of trading patterns if you’re a public market.”

“When you look across the entire business, it will be impacted. It won’t be impacted this week or next week or at the end of 2024, but we are in the midst of the most profound shift of the last 50 years,” he said.

Albo, a Canadian lawyer by training, who went on to work as a casualty underwriter and leader of businesses around the world for Munich Re before signing on to head up Hamilton in 2018, recalls that one of her first hires was Chief Technology Officer Venkat Krishnamoorthy. “He’s always thinking ahead” and observing what’s on the horizon, she said, revealing that she moved ahead to greenlight a pilot of an internally designed tool derived from Open AI after he pulled her aside to demonstrate it last year.

“We do it bite-sized. Let’s take a pilot project. Let’s see if it works. If it doesn’t work, let’s pivot. Let’s move and fix it. And then once it does work, how do we leverage that more broadly across our organization?”

Pina Albo, Hamilton Insurance

Rather than a vendor partnership, “this is our own AI project,” she said, explaining that the tool essentially reads emails to populate Hamilton’s workflow system.

“Insurance submissions come largely in the form of email. They don’t all come looking the same way. Normally you have humans who will extract that data and populate it into your workflow system,” she said. The in-house tool does that automatically. “We use AI to read the emails, read the submissions, populate [our] workbench and then take it from there. We did it in a very specific area just to pilot it, to test it, to see if it works. And now that we’ve proven that proposition, we will expand it more broadly across our operations,” she said.

“We’re building a global diversified specialty insurance and reinsurance company that’s enhanced by data and technology and focused on producing sustainable underwriting profitability,” she told Fitzgerald.

What should executives of P/C carriers start thinking about today as they move ahead to embrace advanced analytics and AI in their operations? What should be on the boardroom agenda?

According to Bruce Baty, a partner with the law firm Norton Rose Fulbright U.S. LLP, regulators have sent a clear message, both with the adoption of the NAIC’s AI Principles in 2020, and now through the adoption of a model bulletin on carriers’ use of Artificial Intelligence Systems, that carriers need to address AI in their Corporate Governance Annual Disclosure reports, which have been required to be filed by all carriers to state regulators since 2015.

The board of directors needs to understand the processes for control over the use of AI within the company. “This is now going to call for a separate AI program for executives to develop out and present to the board that addresses, ‘How are we mitigating risk to the consumers? How are we complying with unfair trade practices acts? [Are we] rooting out implicit bias in the system so that we’re not going back to the days of redlining? And how are we complying with unfair claims practices?'”

“This is a new level of inquiry for the board: ‘Have we developed out our AI program? What does it look like?’ And we want to talk to the executives at the board level to find out, ‘Are we complying with the laws that are already on the books?'”

“The mid-market carriers and the smaller carriers that feel they have got to jump on this fast-moving AI train if they’re going to remain competitive don’t have the infrastructure and the processes built out to develop and to write up a robust AI program. They don’t know what that looks like.”

Bruce Baty, Norton Rose Fulbright US LLP

Said Albo, “I think that board members should be asking us—and they are, quite frankly—’How are you embracing this technology?’ Clearly, there are risks involved, and you have to be conscious of this risk and mitigate the risk. But there is an immense amount of opportunity now, particularly with OpenAI.”

Explaining why that is the case, she said that vendor models trained on specific datasets cost money to build. “With ChatGPT, the barrier to entry has just been lowered significantly.”

“So, our board members are asking us, ‘What opportunities do you see here for using this OpenAI? How can you improve what you’re doing or service your clients and brokers better by embracing it?’ We’re getting questions on both sides: ‘What are the risks and how are you managing those risks? And what bets are you placing?'”

Are P/C insurance company boards and executives well-positioned to take this on? What areas would you encourage them to work on?

Baty offered one example of a client of his firm that is in better shape than most insurers. A team of lawyers, he said, will be meeting with senior U.S. executives of a very large international group to present ideas on best practices—on what a robust AI program should look like. “That group is then going to be making a presentation to their board of directors. But we’re not starting out from scratch with this group. They already have a framework that they’ve worked at. We’re just tweaking that to make sure that we’re hitting all of the marks in terms of the model laws that apply. They’re far ahead of the game. They’ve got their written procedures in place. They’re already in communication with the board…”

“That is not the norm,” Baty said.

“The mid-market carriers and the smaller carriers that feel they have got to jump on this fast-moving AI train if they’re going to remain competitive,…don’t have the infrastructure and the processes built out to develop and to write up a robust AI program. They don’t know what that looks like. They’re going to be catching up and trying to figure this out [at] the executive and the board level.”

“A $300 million carrier is going to need to put millions in, especially at the beginning, to get up to speed if they’re going to be around in 25, 30 years.”

Michael Fitzgerald, SAS Institute

Do P/C insurance and reinsurance company boards need to appoint technologists to be directors? When discussions turn to finding bias in a machine learning algorithm that looks at huge datasets, do boards understand that?

“They do not understand that, and there is a gap,” said Baty, who called out the boards of nonpublic PE-backed companies, in particular.

“As you think back over the activity in the industry in the last 10 years or so, with private equity coming in and buying up insurance companies, primarily life insurance companies to take advantage of the large reserves on the books of those carriers,…you don’t find a lot of insurance expertise on these boards. You don’t find technology people on the board. These are all business and finance people that are driving the companies that they purchased.

What about traditional players, not the newer PE- and VC-backed entities? Who sits on those boards? Does board composition need to change? Is there a technology skills gap at the board level?

Before CGAD, “there was not a lot of deep thought going into the question, ‘Why is this person sitting on the board of directors of this insurance company?'” Baty reported. “It was because ‘I’ve known them for years’ or ‘I went to school with them,’ ‘I played golf with this person…And CGAD now told us that we need to stop and sit back and think about, ‘Why do we think this person is capable of overseeing our insurance operations? And here are her credentials…'”

“Now we need people on the board who are not dinosaurs like me but rather who are well-versed in this kind of technology that can really contribute to the organization.”

At Hamilton, Albo is proud to talk about the recent appointment of Dr. Henna Karna to the board of directors. Karna, who earned both a master’s and a Ph.D. in applied mathematics from the University of Massachusetts, and a master’s in business administration from MIT, has also designed and developed patent-pending technology and applications in the fields of AI, genetic algorithms, behavioral analytics, deep neural nets and advanced data technologies.

“When I was introduced to Henna, the first call I made was to my board chair and said we’ve got to get her on our board…We were so thrilled not just [with] her particular expertise in this area but coupled with insurance. It just was too good to be true,” Albo said, referring to the fact that Karna has led digital services, technology and data businesses at Verisk Analytics, AIG and AXA XL. Her more than a dozen years of experience working in the P/C insurance industry culminated in a position as managing director, Global Insurance & Risk Management Solutions at Google Cloud from 2020-2023. (Read more about Karna in “CM Exclusive: Google’s Insurance Strategy Could Be a Game Changer”)

Albo said Hamilton’s nominating committee is “always very deliberate about looking for diversity of experience, of backgrounds, of knowledge that’s specific to our business and will help us…We have a very well-rounded and diverse board, and several board members do have an affinity to technology or knowledge of technology,” she said. (Editor’s Note: Other Hamilton board members include Russ Fradin, a partner with the PE firm Clayton, Dubilier & Rice, who served as president and CEO of software and IT provider SunGard Data Systems until the company’s 2015 acquisition by FIS in November 2015, and Marvin Pestcoe, CEO of Langhorne Re, who previously served as chief risk and actuarial officer at PartnerRe, where he was responsible for overseeing risk, capital modeling and reserving functions.)

“Putting a board together is a conscious, deliberate process, and if you go about it the right way, you will be able to have all the skills that you need to run a business as complicated as ours…Technology and AI, any emerging technology is core to our business and should be represented around the table at the board level,” Albo said.

While Albo helped push Karna’s appointment forward, unlike Baty, she believes boards without technologists are capable of asking the right questions. Agreeing with Fitzgerald, who hypothesized that boards have technology gaps, on the other side of the coin, she said, whether as a board member for other entities or just as an executive and a board member of Hamilton, there’s not one continuing education seminar that I attend, not one newsletter I open up that hasn’t got AI and technology top of mind. So, I think boards are turning their attention increasingly to this topic,” she said.

“What opened that door was the whole area of cyber attacks. That in and of itself has opened boards’ minds” to start paying more attention to technology, she believes.

At Hamilton, we very specifically thought about this, and we have that expertise both around the executive table with our CTO and chief data officer but also on the board.”

Who’s responsible for AI governance at the board level? Should the compliance, the risk and the nominating committee, for example, all be raising questions about AI? Or should there be a separate AI committee? What about the management team—a committee of people from underwriting, claims, product development, etc.? Or chief AI officer and his group?

Boards have really struggled with these questions, Baty said. “If it’s the entire board, then that’s not a robust oversight of your AI activity. So, boards have been looking at either creating subcommittees or assigning the audit committee to be in charge of this to embrace their oversight of AI.”

Baty encourages insurance company boards to take a cue from how the NAIC settled on a regulatory team responsible for crafting the recently adopted AI model bulletin. (See p. xx for more about the bulletin.) Regulators identified the issue that sat at various committees within NAIC until they ultimately settled on the H Committee on Innovation, Cybersecurity and Technology. “Just as the NAIC came around to [deciding] we need a dedicated committee to address these issues with its own specialized working groups, that’s how I think [carrier] boards of directors need to evolve. [From] the various established committees we have on the board, I think we need to establish a new innovation, cybersecurity committee of the board that is focused on these issues.”

“And they will draw [input] from all elements of the company—accounting, underwriting, legal, compliance,” he said.

(See also, Fitzgerald’s view in “Organizing for Action: The Board of Directors in the AI-Powered Insurer.“)

How should small and midsize carriers be preparing for the new world of AI?

Baty asked Fitzgerald, insurance industry advisor for SAS Institute, for his own answer to the question.

“The most fundamental thing I would say is you have to find resources to put at this rather than giving them some specific project,” Fitzgerald said. “They need to look at their balance sheet and [come to the conclusion that] if we’re going to survive, we have to invest in an area that we traditionally never have before and then see where that goes. I’m talking about a material impact—a $300 million carrier is going to need to put millions in, especially at the beginning to get up to speed if they’re going to be around in 25, 30 years.”

Baty honed in on the governance and risk management tasks ahead. “This is not a knock on the people who currently head up IT, but that’s not their area of expertise. So, you’ve got the head of IT, you’ve got the CISO, but you don’t have anybody that’s really grounded in AI principles and really understands that technology. And if you’re going to play in this market—and you have to as this train keeps speeding down the track—you’re going to have to get on board with this because this is how product is going to be sold. This is how product is going to be developed. This is how claims are going to be managed in the future. And you have got to embrace this technology.”

Albo framed her advice for leaders of AI-powered insurers by reflecting on Hamilton’s approach. “First and foremost, we look at what is necessary to execute the strategy for our business and what we are trying to achieve.”

“This involves thinking not only of the here and now, but more importantly, thinking about the future and what it’s going to take to be great,” she said, drawing inspiration from hockey legend Wayne Gretzky, who attributed his success in the game to an unwavering focus on where the puck was going.

“We use that same kind of mentality—always trying to think, ‘What is out there? What should we be thinking about? What new technologies can we embed in our business?'”

In addition, Albo advises, “We’re not trying to boil the ocean. We’re not playing buzzword bingo here. We’re looking very specifically at what do our underwriters, our operations staff—what do they need to make their lives more efficient, more effective, [and] to make us more responsive.”

“And then we do it bite-sized. Let’s take a pilot project. Let’s see if it works. If it doesn’t work, let’s pivot. Let’s move and fix it. And then once it does work, how do we leverage that more broadly across our organization?” she said.

What are carrier leaders most worried about as they grapple with how to get on board the AI train? What missteps have carriers made?

Kalman said: “The carriers that are at the forefront are the carriers who started investing in a data infrastructure a decade ago, hands down. Those are the ones that listened when the advisers from McKinsey and Accenture and other firms said, ‘Look your data today. Not only is it siloed, but some of that data is embedded in departmental solutions or even embedded within an application. And what you need to do is really develop a sophisticated technology architecture specific to the use of information, separate from the transaction processing systems underneath it. And you need to save the data and leverage the fact that the cloud economics are fractional from what it would cost historically. You need to start creating that environment, especially so that you can focus on the issues around cleansing the data.'”

For carriers that have already taken those steps, “it’s easy to step into a proprietary large language model. It’s been easy for them to say we already know where [the data] is. We already have governance policies. We already have invested in teams to do it. They will accelerate ahead of other carriers that do not have the economies of scale to make those investments.”

Kalman answered no with a head shake when asked if he sensed that most of the industry had done that data and governance work. Then, he offered that “most everybody has done something.”

“It’s not like it’s new, but they haven’t necessarily made it a priority that’s risen to the board level, where you’re having people say, like, ‘Are we doing this fast enough?'” Over the past 20 years, the P/C industry has spent the most money retrofitting core policy admin systems—and they spent a fraction of that on analytics,” he stated. “But optimizing the core processing platform isn’t competitive differentiation. It’s competitive parity. You need the analytics to get differentiation because every insurance company has a different underwriting box, different expectations and a different approach,” he said.

Carriers that don’t have the resources to move in the right direction will see their financial performance impacted over time, Kalman believes. “The rate of change in this industry is going up by order of magnitude every year. Every part of the industry is being touched,” he said.

All hope is not lost for lagging carriers, however, Kalman said. He went on to offer some sage advice about winning the AI talent war. See related article, “‘Critical Thinkers Needed for AI-Powered Insurers.”