An industry consortium or other form of coopetition may be the best way for carriers to jump into the fast-moving world of AI, a tech entrepreneur who has held executive roles at major carriers suggested recently.

During an interview with CM Guest Editor Michael (Fitz) Fitzgerald, Henna Karna suggested that building basic AI models is not a competitive advantage in the property/casualty insurance industry. Instead, understanding the nuances of different types of existing and emerging risks and how to manage them are the skills that set the industry apart—and set carriers apart from one another, she said, recommending that insurers focus their energies on innovations that align with those core skills.

“I would love to have a consortium in our industry where we’re producing AI models that are base layer. They’re not the IP layer. They’re just the base that everybody needs to have—because we shouldn’t be inventing ZIP code 500 times,” she said, referring to the use of geocoded data. That “ZIP code invention” is something that Karna, who engaged with hundreds of insurers around the world in a prior position at Google, has heard executives inaccurately call out as technology that sets their firms apart.

“Things like that, those sorts of fundamental AI models could be very useful for everybody across the insurance value chain. And they’re not the secret sauce. Even though we might think so right now, they’re not,” Karna said.

“How do we now change that mindset?” she asked. “I’m looking forward to that change.”

Karna’s work as general manager of Global Industry Solutions in Insurance at Google, creating an open-source offering of a layer of net-new data and analytics for the insurance industry, gave her a broad view of the P/C industry’s progress to date and the opportunities to come. Prior to Google, she spent more than a dozen years leading digital services, technology and data businesses at Verisk Analytics, AIG and AXA XL, igniting personal passions centered on dual goals of helping the industry become more customer-centric by leaning into data and digital experiences and closing protection gaps around the world.

(Read more about Karna’s career and her previous position at Google in prior CM articles, “CM Exclusive: Google’s Insurance Strategy Could Be a Game Changer” and “Google’s Champion of Holistic Insurance: Karna’s Journey Continues“)

“It’s going to be super-exciting to be able to take all [available] information at a point of decision, at the point of underwriting—[to have] a 360 degree view of risk,” she said, referring to industry opportunities ahead that are being fueled by the “multimodality of everything.” In other words, models that pull together data from text, video, images and sensors are increasingly giving carriers a better understanding of what their customers are experiencing, whether commercial enterprises or consumer.

“One of the learnings I had going from AXA to Google was [that] companies that are not AI bread-and-butter companies should not be AI bread-and-butter companies…Our expertise in our industry is risk mitigation, risk management…For us to spend energy and time and focus on that is key.”

Henna Karna

Still, Karna has questions about how everything will come together.

Outside the insurance industry, “we talk about Open AI all the time now. We talk about these organizations that are looking at generating content.” The content generators are focused on what information can be used to train AI models going forward, “and our industry is not able to catch up with that yet,” Karna said. “We have to find a way to think about how do we as an industry [move on to] generation of information that is new, and not lean onto every other area.”

At that point, Karna, who holds a master’s and a Ph.D. in applied mathematics from the University of Massachusetts and a master’s in business administration from MIT, offered her vision of an industry consortium.

“Consortium may be the wrong word. It’s collaboration or coopetition,” she said, prefacing her explanation of a joint effort for the P/C insurance industry that might allow individual companies to move past “lights-on” activities and forward to breakthrough risk management innovations using data, advanced analytics and AI.

Karna described the technology debt that the insurance industry has already incurred that has not moved the needle on innovation. “Sometimes continuous improvement is a contrast to innovation,” she said. “We have spent hundreds of millions of dollars in technology recently. We have spent it in AI. We have spent it in RPA [robotic process automation] and a lot of things that are sort of looking at automation…”

“If we continuously improve piece by piece, we end up not having enough energy to innovate, to think completely differently because we’re trying to get the ball rolling and we’re trying to keep the lights on,” she said.

“One of the learnings I had going from AXA to Google was [that] companies that are not AI bread-and-butter companies should not be AI bread-and-butter companies…Our expertise in our industry, the business we know is risk mitigation and risk management…For us to spend energy and time and focus on that is key.”

“For us to reinvent a [data] cloud or reinvent an AI model that’s going to fix my autogeneration of clauses—that isn’t the game-changing world. It’s a very short-term [win]. It can help us go faster today, but it won’t scale and it won’t sustain us going forward,” she said.

Insurers have to “be really conscious of how many dollars do we spend on the proactive now” activities vs. how many are spent on building responsible insurers of the future. “Is it 50 cents and 50 cents? Or is it 20 cents and 80 cents?” The question of how dollars should be allocated, she believes, should then evolve into a different question: “Can we lean on each other to solve some of the basic things?”

“We didn’t invent APIs as an industry,” she continued, referring to application programming interfaces (that allow two or more computer programs or applications to talk with each other). “We’re using them. But if there was a best practice for APIs, we would just take that and go. We don’t have to create a new version of that.”

“If there was a best practice for telematics, we would take that and go. And then our competitive advantage would be how do we do more beyond the base layer,” she said.

Karna continued: “It still surprises me when I go into an executive meeting and the discussion is about how different we are because we do location monitoring differently. But that actually just means [using] location-specific data, geocode or something like that. It’s not actually location monitoring” via sensors or real-time data.

“Transparency about what we actually are good at, and then what we think is not our competitive advantage, that I think is going to be pretty important.”

“Fundamentally, what is our business? Let’s get very good in that business and use the ecosystem to do everything else with us—lean on the ebbs and flows of the ecosystem to get us all the ingredients that we need, and then we focus on what we think is going to be our biggest game-changing, value-add differentiation for the market that we’re in,” she said.

“We tend to do all of the above, and we lose a lot of momentum that way.”

Describing Karna’s vision as a collection of AI microservices sitting above the technology infrastructure level, Fitzgerald drew an analogy to the work of the industry standards organization ACORD. “Nobody has their own workers compensation application anymore, right? They just pick up the ACORD data and format” and add to those.

Been There, Done That

Building on Karna’s idea, Fitzgerald suggested that the P/C insurance industry might take a cue from the telecommunications industry. Twenty-five years ago, pre-3G, telecoms were “building their own protocols,” he said. Eventually, “they standardized on a common approach. So, when they went to 3G, 4G and now 5G, they don’t build their own network protocols anymore.”

Would it make sense for nominating committees of insurers to consider former telecom executives—or executives from other industries “who have seen this before from outside”—on their slates of board director candidates, Fitzgerald asked Karna, who was recently appointed to serve on the board of Hamilton Insurance Group.

Karna agreed that the idea makes sense. “Not only have they [telecom executives] understood what normalization is [but also] the fear of it. There’s a concern, right? We always spend so much energy thinking that we’re different that we perhaps have to step back and say, ‘OK. Actually, this isn’t what differentiates us because that’s just base core.’ And then leaning into what that difference is.”

“The CPG [consumer packaged goods] industry is similar. At some point, they had to normalize also,” she said, referring to the process to reorganizing data in consistent formats from disparate sources.

“Executives that have been there, seen that and have seen technology take a different course of action to create relevancy, and [seen that] that relevancy is good for the end customer, can bring a little bit of grounding to [insurance] boards” by sharing their journeys. They might also offer perspective for gathering good questions. “At the board level, I think that’s the most important thing—[that] we are thoughtful about the questions we ask and sincere about what we hope is a good outcome,” Karna said.

Leadership and Boardroom Gaps

Fitzgerald also asked Karna to reference her experience in dealing with P/C insurance carriers to describe any gaps in technical understanding preventing executives and boards from moving their organizations away from the incremental small steps to make quantum in using AI and advanced analytics to fulfill their missions of risk mitigation and management.

“A very common theme was what does cloud actually do?” she said, noting that information about potential workflow improvements and risks of implementing very large databases on the cloud come from internal sources. “That assumes the internal folks have the expertise already and have seen it better somewhere else. Generally, that’s not the case. [Carriers] may have hired one or two, or maybe even 15 people that have a sense of what is better elsewhere, but those are not the people that are making it to the top to make the discussion happen.”

Karna reported hearing executives make statements such as, “We’re going to move to the cloud and we’re going to be able to use AI more broadly because we [will] have the new technology in place.”

“Those are distinctly different comments, but they’re put into a [single] sentence. And boards, similarly uneducated in the concepts, will say, ‘That sounds awesome.’ But implementing AI, going on the cloud, incurring extra cost, reducing the expected costs—those are all different. They take their own time, and they have different prerequisites.”

“At the board level, if we don’t have enough perspective there, then nobody asks the right question to go deeper,” she said, again affirming the idea of bringing someone onto the board “that has done it already. It need not be in insurance. It may just be that they have seen something better and can imagine what it could be like in this new situation where we are struggling.”

She offered examples of better board-level questions about cloud migrations: “What is the point of going cloud? What is the journey supposed to be like in a good way? What are the potholes we should be watching out for?” When those questions are answered at the board level, the CEO then has a perspective that is independent of his own direct reports, she noted.

Now, with AI, executives and boards also don’t appreciate that there are different types of AI, generative AI being just one of those.

“Just because OpenAI is fastest growing, we started getting focused on the GPUs,” she said, referring to graphics processing units. (GPUs, initially developed for putting high-end graphics into computer games, have become the building blocks of artificial neural networks.) “What does the GPU have to do with multimodality? What does that have to do with reinforcement learning with human feedback (RLHP)?” she said, referring to a form of machine learning.

“Someone at the board level, having seen the growth of AI as its own independent thing” and all the variations outside of the context of insurance, can bring useful insights to insurers who are new to the world of AI, Karna said.

Internal Consortiums

While Karna might not describe it this way, she has experience in building “internal consortiums” at insurance companies like AXA XL. In her AXA XL role, she designed and implemented an award-winning data and technology ecosystem, called DEEP (Data Ecosystem & Engagement Platform), that delivered analytics capabilities at scale across the enterprise. Asked what advice she would give to companies that want to start bringing people within an insurance company together today to solve data and advanced analytics problems, Karna said not to wait for the CEO.

Business leaders “shouldn’t expect the CEO’s word to be golden. Oftentimes I hear, ‘If it comes from the top, it’ll happen. That is not the case. It depends on the CEO.” In other words, leaders of businesses or functions within insurance organizations “shouldn’t get the message that if we don’t think our CEO’s up for the task, we can’t do that,” she said. “Our responsibility when we’re in a company is to help our CEO think outside the box that they may be in.”

Reflecting on her AXA XL experience, she also suggested transparency with internal stakeholders as another ingredient for success. “It was painful, but we took the time to be super eloquent,” she said. What sounds like internal marketing was an effort to let people know project leaders were “not out to reduce headcounts in IT [or] to reduce the operational underwriting work.”

“We’re actually just focusing on a better business-decisioning capability,” not fixing anything, she said, recalling another message that project leaders were super careful to deliver: “We’re trying to do something differently, but that doesn’t mean the things that were there were wrong.”

Additionally, she said, “The technology has to feed a business use case. It can’t be, ‘I want to go cloud because I want to reduce cost.’ You will actually be years out before you see that reduction in cost. [But] if it’s a business use case and the business is waiting, anticipating and opining with you as to the direction in which you operate to build technology or build the AI use cases, then you’re going to see value.”

“That helps calm the nervous system of the company,” she said, referring to the impact of meeting intermediate milestones of value creation. “As soon as we see value, we have a reassurance, a response or feedback loop that we can then lean on. When we don’t have a feedback loop by the business, it doesn’t go well. And the feedback loop has to be by the business. It cannot be within the IT department or within the Operations department. It really needs to be where it hits the customer, where it hits the top line or bottom line,” she said.

Finally, purpose is the fuel for people who drive game-changing transformations, she said, recalling her time at Google. After someone suggested that Google engineers “all have car insurance,” and therefore understood what insurance was all about, Karna led off the next town hall meeting describing the industry’s true value in creating resiliency with the story of a historic dam failure.

“That is what we are after in our industry. We are actually working on those really complex problems. And, internal to an organization, if we start to become too narrow—we’ve got to reduce this [cost] or we’ve got to change that [process]—then I think people lose hope,” she said.

Similarly, AXA started every town hall meeting with a customer story. “What happened? What was our role? The timeline that it took? All the good and bad? …The government of Fiji had to deal with this typhoon…The responses were these, [and] had we done it better, what would have happened?”

“It made people work the extra hour without me asking, ‘Hey, could you work on a weekend?’ They understood that their use case was just precise enough to make [something] happen,” and they understood the domino effect for something else to happen. “We were leaning into a chain reaction.”

“That is why human beings work extra hard,” she said.