In October 2024, more than 9,000 participants running the gamut of insurance industry players gathered for the annual ITC Vegas insurance conference. The hot topic of the conference, unsurprisingly, was the use of generative AI (gen AI) and its potential implications for the insurance industry. In this episode of the McKinsey on Insurance podcast, Matt Cooke, global director of marketing and communications for McKinsey’s Financial Services Practice, sat down at ITC Vegas with McKinsey senior partner Tanguy Catlin for a wide-ranging discussion on the industry’s future trajectory, remaining relevant in a changing marketplace, and creating value with gen AI. The following transcript has been edited for length and clarity.
Matt Cooke: The world’s biggest insurance conference felt like the perfect opportunity to talk about some of the industry’s biggest issues and trends and what we are likely to see in the months and years ahead. Let’s start here at ITC 2024. What are you hearing? What’s new this year? What’s likely to come?
Tanguy Catlin: Many insurtechs are focused on very, very narrowly defined, specific use cases. This creates a challenge for insurance companies that need to integrate multiple use cases to truly transform an end-to-end customer journey or a domain. There’s still a lot of work required for insurance carriers to capture the value that insurtechs are providing. But we are slowly seeing the emergence of platforms that make integrating those insurtechs much simpler from a technology standpoint.
The big topic this year is gen AI. Many of the insurtech companies here provide insurance carriers with gen AI to transform aspects of customer care in sales, claims, or underwriting. There is more momentum toward the emergence of ecosystems, where multiple participants can collaborate to solve very complex problems such as the future of mobility or truly transforming the claims experience.
Unfortunately, many insurers start with an incomplete definition of gen AI that limits their ability to think about the proper use and value capture they can get.
Matt Cooke: When we think about personal, corporate, and life insurance lines, what does gen AI mean there? What are the greatest opportunities? How are we going to see that evolve over the coming year?
Tanguy Catlin: Unfortunately, many insurers start with an incomplete definition of gen AI that limits their ability to think about the proper use and value capture they can get. Gen AI is different from traditional AI and uniquely positioned to solve certain types of problems—but not all problems. There are many issues that insurance companies face that require a mathematical solution. For those problems, traditional AI is the way to go. What gen AI does is make sense of context in a way that opens the door to new use cases.
Even when you use large language models with traditional AI, you need to build a model to solve a specific problem. So there is very little scalability, and each model mostly requires the use of structured data. You have to identify the problem, build a model, and ingest structured data to get an outcome. With gen AI, you flip the problem on its head. You take structured and unstructured data, and you create a very large model that can be used for hundreds, if not thousands, of different applications. You start from the model and ask, “What problem do I want to solve?,” versus starting from a problem and asking, “What model do I need?” That completely changes the notion of scalability. That’s what’s going to be transformative. But that means you need to have a very different approach to your architecture to be able to capture the value of scalability.
Many people still talk about using gen AI to automate what they’re doing, such as automating a report for a finance function. That is the wrong approach. Gen AI changes the way you engage with your data. Rather than asking for a report, think about the dialogue you can have. If you’re the head of a call center, for example, you can ask, “What happened in my call center today?” And gen AI will give you a synthesis. Then you could double-click and ask, “What was the reason that absence was higher?” When you have an answer, then you could ask, “Based on the data we have, what are the levels I should pull?” It’s about engagement and dialogue. While many people talk about gen AI, I think there needs to be a higher level of maturity and understanding of the technology to be able to frame the right approach to capture the value.
There are four primary applications of gen AI across industries. Because its language model is very good, the first applications are in the field of technology. You can use gen AI to create code or take code that is written in one language, such as COBOL, and translate it to another language, such as Java. Most insurers are doing that.
The second is creativity. Gen AI creates vectors that have context that it can then use when you provide further context to create quality content that can be hyperpersonalized. The areas of knowledge management and synthesis are very relevant for insurance. In claims and underwriting, we need to access information that, because we are a regulated industry, is hosted in many different domains and applications. Gen AI is incredibly good at finding the right information, making sense of it, and providing it in a synthesized manner that can be consumed by people. This accelerates the speed at which you can onboard new people in a call center, underwrite a policy, and improve the quality of your services for claimants.
Last is customer care. Coaching or replacing a human in engaging with customers is important. Gen AI has the ability to be multimodal, which means it can emulate empathy, and it is on the verge of transforming the way we approach customer interactions.
Many insurance carriers will have gen AI provided to them by the platforms they’re using for traditional functions such as HR and finance. Where insurance carriers will differentiate themselves will be in underwriting and claims because they’re going to need to create those applications themselves. I’ve seen some amazing ways to improve workflows in underwriting. On the claims front, gen AI can be used for everything from coverage determination to ensuring that BI [bodily injury] treatment is fully compliant with regulations. Many emerging applications require insurers to develop the solutions rather than source them from platforms.
There is a real conversation to have about the relevance of our industry moving forward and the ability to react to changes.
Matt Cooke: Now, in terms of maturity of usage, where we are today across the industry? Obviously, the evolution is constant, but how much progress do you think the industry is likely to make? Where is it now and how is it likely to advance in the coming year?
Tanguy Catlin: The pace of developing and improving gen AI is unlike anything I’ve ever seen. The emergence of what we call AI agents, analytical entities that have the ability to apply human judgment and orchestrate tasks by leveraging other models to solve complex customer problems, is transformative. So, on one hand, progress is very fast. On the other hand, it’s very slow, because progress is only as good as the adoption we can drive. The technology is making enormous progress; we can solve problems we never considered just two or three years ago, but driving change in human behavior is much harder. In our experience, gen AI adoption and behavioral change require an investment of three to four times more time and resources, including capital, than the technical side.
Matt Cooke: We’re going to move to personal lines. Recently, US insurance carriers and reinsurers were looking at two major hurricanes in Florida in a two-week span. The insurance industry is still working hard on the postdisaster situation there. Will some markets in the world become uninsurable? How is climate change affecting premiums and coverage?
Tanguy Catlin: I first would like to express empathy and compassion to all those who are affected. The reason I love this industry so much is because I believe we are purpose-driven and our purpose is to get people back on their feet, and I’m hoping that collectively we’ll be able to do that.
Having said that, I’ll start with a sobering point. In the US, the percentage of loss from a catastrophic event that is covered by insurance versus not covered has been in steep decline for 80 years. If you extend it beyond catastrophic events, there are many risks that are emerging now—cybersecurity, political risk, and IP, for example—for which our industry is not yet able to provide solutions in line with what society needs. There is a real conversation to have about the relevance of our industry moving forward and the ability to react to changes. Climate change is one of them. Regarding the solutions that we need to tackle, we need to think about driving protection, trying to have people ready for some of those events better than before. It means working with regulators. Rates, pricing, and coverage are not always adequate, and regulators have a role to play. Insurers will need to continue to optimize concentration risk, product coverage, and a number of other elements.
There is a huge challenge for the industry to educate drivers about how the actions they are taking are significantly increasing their risks.
Matt Cooke: I want to turn to mobility. The automotive sector is coming ever closer to autonomous driving. What’s the latest? What does it mean for insurers?
Tanguy Catlin: Sensors and software are making cars somewhat able to control themselves. And the question is whether to buy insurance through insurance companies or buy it embedded in the car provided at the point of sale by the car manufacturer. Many car manufacturers have brands as strong as, if not stronger than, those of insurance companies. They have access to data that, when married with claims data, allows them to have or build a pricing advantage. They have much better control of the repair value chain when there is a claim and could potentially drive the cost down. At the same time, car manufacturers are not ready to put a lot of their capital toward providing insurance. And insurance companies need [access to vehicle and behavioral data as well as claims data] to be able to underwrite and price risk better.
The second point is that somewhere between 94 and 96 percent of car accidents are caused by human error, and technology should be able to overcome most of those. While accident frequency has come down, accident severity has come up. Part of it is because the price of a car has increased dramatically, and there are no cheap repairs. If you repair a bumper or a windshield, it costs a fortune now because it’s embedded with sensors. We’ve not really been able to address the cost curve of insurance.
But accident frequency has not come down nearly as much as it should have, based on all of the technology embedded in cars. That’s driven by the fact that human behaviors are now affected by distraction. There is a huge challenge for the industry to educate drivers about how the actions they are taking are significantly increasing their risks. More collaborations are required across the ecosystem, but the key will be the access to data so that we can convey it to consumers in a way that allows them to understand.
What’s happening right now is that insurers are looking for ways to get more efficiency from the capital they’re deploying.
Matt Cooke: Let’s turn to the life space now. Populations are aging in so many countries, including in large countries. What can life insurers realistically offer people in their 50s, 60s, and 70s?
Tanguy Catlin: Start with understanding what life insurance was designed for, what society looked like back then, what society looks like now, what the trends are, and how the industry needs to evolve. Just after World War II in the United States, people did not need to worry that much about retirement, and life insurance was really about death protection for the family. Today, fewer individuals get married, and they marry much later. Typically, both partners have a job, and they have fewer kids much later in life than in the past. The need for death coverage is much lower and because people are living much longer, and healthcare costs have risen significantly. If you start with those basics, the insurance industry will need to find ways to be relevant in the retirement, health, and wealth management spaces.
Matt Cooke: We’re seeing leading insurance companies in the US and Europe implementing a flywheel approach with three elements: distribution at scale, superior investing, and new flexibility in sourcing capital. Can you talk a little bit about the future of that trend as you see it?
Tanguy Catlin: What’s happening right now is that insurers are looking for ways to get more efficiency from the capital they’re deploying. The trend of creating value across liabilities and assets management is here to stay. For the longest time, life insurers have been able to create value by looking at diversification within the liability business. Now that the industry is realizing there are synergies across liabilities and asset management, that trend is going to accelerate and, I suspect, differentiate the winners from the losers in the future.