Flyreel founder talks present, future of AI in the insurance industry  

Editor’s Note: Cole Winans is the vice president and general manager of Property Solutions at LexisNexis Risk Solutions, and was founder and previously CEO of Flyreel, an advanced AI solution for residential and commercial property insurance carriers that was acquired by LexisNexis in 2022. Prior to Flyreel, Winans developed and deployed over 100 successful  software products and applications.  Winans recently sat down with InsuranceNewsNet Editor-in-Chief John Forcucci to talk about the use of AI in the insurance industry.

John Forcucci:

AI has been utilized by the insurance industry for a while now in a variety of applications. The company you founded, Flyreel, is a great example of that. There seems to be a new product or application for AI and generative AI nearly daily. How is AI going to impact the industry at large?

Cole Winans:

It's definitely at the forefront of people's minds and I think some aren't quite sure how to react to it. There are a few schools of thought. Some are avoiding it all together. We've seen some businesses reject and prevent their employees from using generative AI products. And then some are saying, "My gosh, it's going to disrupt everything." And the reality for us is we think we're trying to take a very calculated and cautious but optimistic approach because the technology is really amazing. So, it's different than what we've seen in the past. This isn't like blockchain where, in our opinion, it got hyped and there isn't necessarily a fit in the market for products around that yet. This is an interesting situation where if I were to summarize, I'd say technology has never been further ahead of the market.

And that's an interesting thing to think about. What that means is the valuable use cases haven't really been discovered yet.

Forcucci:

There's a lot of legacy data, dark data, misused data, unused data, and then you've got sort of the Flyreel situation where there's so much that might be automated and made more accurate potentially through some of the tools out there. So, what does that picture look like and what's coming into focus now?

AI in insurance: 'What can we actually do today?"

Winans:

Two concepts come to mind. One is: what can we actually do today? The other is: what could the market look like now that this technology is here?

Forcucci:

Let’s start with what might be done today? What do you see?

Winans:

There are a couple of key areas. One is just distilling the data that is available to carriers to help make for more informed decisions with less friction. If you take the vast amount of data that's available from analytics providers or within the walls of a carrier, they're leaning on data science teams to construct projects that query data sets. Then they format that data, and they present it back to the executives. And the cycle time is usually months to then make for a more informed decision. But this technology shortened that cycle time into a natural language transaction that is instantaneous.

That's really where this technology shines. So, what does that mean and what are we seeing now? It's a force multiplier, and it enables better decisions on lots of data in a short period of time. You can think of it almost as an underwriting co-pilot. You can imagine, as they pull up a record, there could be suggestions; it could evaluate data against an underwriting policy and say, "Hey, watch out for this." It could enrich data. So, I think that's the concept that we're seeing: You're able to operate on larger amounts of data in smaller periods of time.

Forcucci:

So, being able to analyze larger data sets can provide new insights that wouldn't have been clear previously because the amount of data you're looking at versus that particular policy wouldn't have been available or couldn't have been crunched immediately? Is that the idea?

Winans:

Yes. And this isn't even limited to generative AI . We can look at history as an example here with Flyreel and firsthand experience. When we first brought Flyreel to the market, carriers didn't necessarily believe in AI. And also at the same time, they didn't look at the interiors and exteriors of homes, they were only looking at the exterior because it was too expensive to send people inside homes at scale. So what happened? Well, we allowed them to see both and we gave them a lot more data. So, then their next complaint was, "This is too much data."

So how did we solve it? We developed AI that literally looks at all the photos and the videos for them through the lens of an inspector and only calls out the scary stuff, highlighting it for their underwriters. When they open up a big inspection report, they only have to look at these two items. That's very powerful. And what this does, generative AI and these large language models, they can do that for text. That's their specialty. You can look at historical patterns in the market. You can see correlations to aerial imagery as well, analyzing roofs at scale: "Show us the ones with tarps and missing shingles." You don't have to look at all those images. It can ingest all of that and direct you appropriately. We've now achieved the equivalent of that in text and data.

Looking closely at customers' pain

Forcucci:

What's next for LexisNexis? How are you using this? What will you be doing with this next?

Winans:

We are looking closely at the pain that our customers are experiencing in the market. It's a very public pain with carriers having to pull out of some stakes. It's on full display.

We've spent some time looking at the needs in underwriting. You've got to select the right risks; you've got to get data on those. The thing is, so much of the pain is on the broader book renewals and re-underwriting, versus new business. Traditionally, carriers have been really inspecting and looking at their new business, but they've got massive liability, dormant premium and risk that's sleeping across their existing book. So how do we enable them to look at all of that, analyze all of that data, and then act on it? And that's at a scale that they've never had to deal with before.

So that's how we're applying AI. We're helping them select the right risks. We've basically been building a platform for this, called Total Property Understanding.  We enable them to select the right risks, capture data through an AI assistant that guides the insured through scanning the actual property using Flyreel, and then acting on it while actually configuring and training the system based on their underwriting policies.

An AI assistant can look at the record and the data that comes back and gets captured and act on it just like an underwriter would. It's an end-to-end system that enables them to analyze their risks at scale. Because that's what they’ve really got to do now.

Forcucci:

How is AI in insurance industry being used, with all the additional data that is available now? Obviously, you've got the Flyreel example, but what else are you seeing? Who is doing this well?

Winans:

I would say there are a couple other examples. One is adding structure to unstructured data. There's so much data and information that's out there that hasn't necessarily been accessible because you need humans to structure it. These [AI] agents take great direction, and they can read tasks at large scale and basically comb through data and structure it into a format that's digestible or usable.

Forcucci:

The agents are sort of the second level of the AI that will independently take your request and go figure out a way to do it, essentially.

Winans:

Right.

Forcucci:

Who's using AI agents well? Where are you seeing anybody beginning to implement that type of AI in the insurance industry?

Winans:

I'm starting to see folks in the data science divisions and the IT divisions begin to run experiments. They're also proactively engaging in conversation so that we can share knowledge around best practices with us and our customers. So, we're having multiple engagements and conversations with carriers at this time.

Forcucci:

What do you see as the biggest roadblock to AI in insurance industry being implemented well and quickly?

Winans:

I think that it's always a trade-off of precision and recall and the accuracy that's necessary for this to become valuable in the market or to a carrier. The need for accuracy is probably the biggest challenge because when you require high accuracy at scale, it imposes deep, difficult requirements on the system. So what are the trade-offs of good enough versus great? And most of the difficulty getting something to an enterprise ready state is really in that last 10% of performance. You can get to 90% accuracy pretty quickly, but it's going to take you five, 10 times as long to go from 90% to 99%.

Also, depending on the use case, there is the need to implement and instrument these systems so that they're logging and traceable so you can explain how they arrived at a decision and be able to defend it. I would say wherever you need to potentially defend a decision legally, that's going to be a much more difficult system to develop. Not because you can't, but because these systems out of the gate don't have that traceability. You have to build it.

Innovation 'happening on weekly cycles'

Forcucci:

What kind of runway is needed to get to that level? Are we talking about something that we'll see two to three years from now?

Winans:

There's innovation and it's happening on weekly cycles, not monthly, or annually. I would say, it's really less the pace of the technology, though, and the pace of this market. I would say most likely a couple of years. One of the other challenges here is the availability of talent.

You have carriers and businesses just like ours that are going to be competing with Microsoft, Open AI, Google for the talent to build these systems. So, it's very interesting. At the same time, you are seeing the cloud providers take a lead and equip their customers with tooling to democratize a lot of this as well.

“Now these systems can provide great customer experiences...”

Forcucci:

We see the use of AI, obviously, in underwriting and claims. What other areas of insurance either are beginning to use this or will be using this?

Winans:

I think the groups that will be using this will also be in acquisition retention. You can imagine if you have behavioral indicators, a system taking that in and potentially developing profiles or alerting as to whether there's a potential for churn. You have support like, does this claim qualify? A lot of those questions are truly better for machines anyways, but the delivery has been suboptimal where the consumer is like, "Hey, I don't want to read an article, just give me the answer."

And now these systems can provide great customer experiences and give them the answer. So, that next piece is that customer experience layer. That's where things are really going to evolve, I think. It's so much a question of what's in my agreement with this carrier? What is covered, what's not? What do I have, what do I not? But also, for shopping for insurance. That’s another really interesting topic.  The consumer’s got to find the right product. I think that the friction and the barrier to entry to find bespoke products will go down as well.

Forcucci:

The insurance industry on many fronts has been seen as slow to move, slow to make progress. Is AI the shot in the arm that is necessary to get major carriers and others really to move forward? Do they have any choice?

Winans:

I think the need for regulation and to be very, very cautious and thoughtful with data certainly impacts the speed. But I do think every carrier will ultimately need to adopt this. What we're seeing is an interesting convergence of events, where you have this massive technological advancement, and record underwriting losses in the market. And the market is reeling with pain, which is creating real pressure on carriers to drive profitability and reduce expenses. And, so, you have one of the largest innovations in automation converging with a need to reduce expenses and drive profitability. It will be interesting to see how that takes shape. I feel confident that every major carrier will be adopting this technology if they haven’t already.

Forcucci:

How will the adoption of AI change the insurance industry?

Winans:

I think that AI will be the new base, essentially. It's going to be at the core fabric of every one of our businesses. It’s a new way of creating solutions and products – and of thinking about things. AI will gradually just become table stakes, like the internet. It's  going to be at the very core fabric of every business in the industry.

John Forcucci is InsuranceNewsNet editor-in-chief. He has had a long career in daily and weekly journalism. Contact him at [email protected]. Follow him on Twitter @INNJohnF.

© Entire contents copyright 2023 by InsuranceNewsNet.com Inc. All rights reserved. No part of this article may be reprinted without the expressed written consent from InsuranceNewsNet.com.

John Forcucci is InsuranceNewsNet editor-in-chief. He has had a long career in daily and weekly journalism. Contact him at johnf@innemail.