In the rapidly evolving world of insurance, technology is a key player in driving innovation and efficiency. In a recent episode of the Insurance Unplugged podcast, Lisa Wardlaw sat down with Anil Vasagiri, SVP and Head of Property Solutions at Swiss Re, to delve into the transformative power of AI in the insurance sector.
From the use of AI in modeling secondary perils to the promising future of Generative AI (GenAI) in underwriting and fraud detection, Anil offers a comprehensive look at how these technologies are reshaping the industry. In this excerpt, we share some of the highlights of their discussion around the opportunities and challenges in adopting AI and GenAI in insurance.
The Transformative Potential of AI
Lisa Wardlaw:
From your vantage point and depth and breadth of experience, what are you seeing as truly transformative opportunities and aspects for AI in insurance, and where do you think we’re getting it wrong?
Anil Vasagiri:
AI is a tech that has truly transformation potential and it has already had quite some material impact on different facets of insurance value chain, and it’ll continue to have an outsized impact going forward. I mean, some of the use cases that are emerging are fascinating. Especially when you start looking into the so-called secondary perils where the pure physics-based modeling approach doesn’t necessarily hold true. That’s where the combination of AI and ML based approaches have proven to be exponentially more impactful in terms of modeling the peril. So, application in extreme and models is a fascinating area that continues to evolve exposure insights, and this could be in the context of feature extraction, computer visions being around, and I think it’s being perfected as we speak. If you look at the quality of the output that comes out today, I think in terms of feature extraction, change detection, I mean just the basic nuts and bolts of information that you need for underwriting and ability to sort of get it at a high degree of precision. It’s just fascinating. So now you’re starting to see the application of AI and behavioral sciences in the field of claims and fraud detection and things like that, which again has such a rich potential if applied in the right manner. And so that’s just in terms of generic AI technology and techniques.
In the territory of large language models and GenAI, you’re now starting to see a whole new set of use cases that are being unlocked. Clearly there is still much to do there and more advances to be had, but even the spectrum of application of these technologies is so fascinating on something more mundane as policy wording and sort of contracts reconciliation and in consistency on one end to essentially taking it all the way, we have this notion of advanced analytics and applied application of analytics.
So, you kind of have this notion of descriptive, predictive, and prescriptive. I think the sheer spectrum of the applications and how you can use this is just fascinating. Now, clearly some areas may have been overhyped, and for me that’s more about our understanding of technology and how we are using it to solve a problem than it is a knock on the fundamental technology in itself. And so, I wouldn’t necessarily view this as an issue. It’s all part of the learning journey. We will get through this, and AI is not the first technology to go through this hype cycle and it won’t be the last and it’ll just course correct. I’m thrilled to see the innovation and the thought leadership emerging insights on how we use some of these technologies, but equally so when you look out and see the innovation markets that is being generated, that’s coming out in the market’s, just fascinating.
The GenAI Opportunity
Lisa:
To me, one of the hurdles of digitization is the accuracy, speed, credibility, reliability, which the underwriters, the claims adjusters, the modelers, the risk managers must be able to trust. Are you seeing an inflection point with Generative AI? In terms of LLMs, we still have work to do to mature it, to train it and all that, but are you seeing an inflection point where it’s becoming more reliable than these lower fidelity tools we had for a specific point?
Anil:
Yeah, I think it’s on a spectrum, but increasingly some. I’ve used the policy wordings, the contract as just as one of the many examples where context matters and understanding the context and is as important as the content itself. For me, when you start thinking about the application of GenAI, any use case that has a human in the loop is a fair game.
If you look across end-to-end within the spectrum of your business processes, whether it’s a question of dealing with high velocity, high volume data streams, or it is computationally intensive, a very limited narrow application of data extraction, I think these are the two bookends of the opportunity spectrum. I think you touched on this point around content being trapped in semi-structured artifacts and such. And increasingly that’s one of the areas of opportunity.
The insights are going to come from being able to process and apply these techniques at scale to get to those levels of insights on a consistent basis. So yeah, the use case as you described, it’s one of those that hasn’t been fully flushed out properly, but I think again, the LLM or the generative AI for me sits on top, is sort of the wrapper.
It’s an orchestrator of many of these base technologies, which extract these components to put together that sort of comprehensive overarching insight. For me, that’s really the power of this new capability that we now have our hands on.
Anil’s Call to Action for Insurers
Lisa:
What is one thing you wish everyone would start doing, stop doing and continue doing?
Anil:
I think it’s sort of willingness, I guess insurance is a regulated industry, and so there is great consideration around using these technologies in a responsible manner. And of course, all of that, no doubt is right on the top.
But in the context of all of that, I think the industry can benefit from using some of these new technologies that are emerging and use it quite responsibly. So, I guess the call to action is don’t be afraid to try and see, whether it’s GenAI, whether it’s LLMs of some sort or whether it’s pure AI on feature extraction. But you should definitely be willing to test it out because at the end of the day, I think the value you extract may be very different from your competitor or somebody else. So, there is an element of what is truly unique for your organization, and that may be very different. And so don’t be afraid to sort of experiment or test out. So that’s one.
And the other is that inputs matter. All data is not the same. The better quality of data you put in, the richer your insights and outcomes are, so quality over quantity, I think that really is what is going to be impactful.
This is just a fascinating time and space we’re in. So truly one should be open to testing them out and willing to sort of explore a bit.
Listen to the episode of Insurance Unplugged with Anil Vasagiri.
