On the Insurance Unplugged podcast, Lisa Wardlaw is talking to professionals from across the insurance ecosystem every week to understand how they are navigating the insurance world through AI. She has talked to the experts at large carriers like Swiss Re, Chief Data Officers and industry thought leaders.
In a recent episode, she sat down with Paul Bessire, the CTO of Coterie Insurance. Coterie is a Managing General Agent (MGA) and insurtech company that aims to add speed, simplicity and service to small commercial insurance from the perspective of making things as easy on the agent as possible.
Here are some highlights from their conversation.
Start with Why
Find the ‘so what.’ Nothing in data matters without context or application. Leveraging data and technology to solve the problem and understanding the impact that you’re having is more important than just having data or having technology, even if you’re trying to be innovative with it.
How to Think About AI in Insurance
There are two ways that I think of AI. One of them is helping to get data and information. With large language models (LLMs), we already have these existing data sets, for lack of a better term. And so, our ability to leverage the existing large language models initially to help bring that data in-house is kind of step one, right? It is trying to figure out what can we learn about this risk that we don’t need, nor want to get from the agent or the policy holder. I’d rather be able to get that as close to the source of truth as possible, but not from the source of truth, if that makes sense, not from the individual, because I want it to come through with a very clean lens.
We use AI and LLMs to help us with that, and we’ve gotten really good at writing prompts and querying against the models that exist to plug directly in so that we’re gathering that information.
The other thing with AI that I think is really valuable is this kind of fundamental assumption that everything is not linear and not independent.
With AI, it’s like having an all-star team on your side: the expert underwriter with 40 years of experience, the people who are best in their field, and having the ability to replicate that expertise over and over, and faster and faster.
And that’s what AI is. It speeds it up, removes some of the bias and allows us to really understand when every piece comes together. Rather than saying, ‘this piece means this thing and this piece means this thing,’ which is what we had to do before.
The Impact of InsurTechs
I think everybody to some degree should be rooting for InsurTechs because we get to make it up as we go along. When we make a mistake or if we fail, the broader industry gets to learn from our mistakes.
I see InsurTechs as helping pave the path for what things could look like in the future.
Start with the Fundamentals
We could use AI to better understand the age of a building or the age of a roof, but because we’ve always needed to know those things. But does that matter?
Let’s start with the fundamental, what’s the real problem we’re trying to solve? Then, based off the impact we’re trying to have, what’s the real problem we’re solving? Next, where is the data related to this? And finally, what is the approach that we can use?
Most people tend to start the other way around. They say, what’s the data or what’s my day to day, as opposed to what’s the real impact or the real problem we’re trying to solve?
To me, you’re not even just solving loss ratio or combined ratio. I want to know LTV, I want to know inclusive of the likelihood to cancel, inclusive of how long this policy is going to exist on my books, inclusive of my servicing costs, inclusive of claims costs, the claims themselves, our costs to underwrite, which are relatively low.
We automate everything. It’s a hundred percent digital underwriting. What does this policy mean to me? What’s my expectation at the very beginning when I’m quoting this? What do I really think this is going to be worth to me? And solving that problem is very different than, saying, okay, now a claim has come in, what do I do to mitigate it? Or, okay, bad risk. I’m supposed to get bad risk off my books.
When people ask, what is your plan for using AI? I say, it’s not a claim solution. It’s not servicing, it’s not underwriting. We want to know the most about a risk we possibly can as quickly as we can because speed does matter. When you’re talking about agents’ time, and we’re interacting with agents first and foremost for all of this, as much as we can, as quickly as we can. And we need to leverage that to inform us about all of those things so that we have the same breadth of information to aid our servicing, to aid our underwriting and to aid our claims as though they’re all interconnected.
To me, they are all linked.
Advice for Insurers
Three main pieces of advice:
1) Find the so what, think about the impact and the problem you’re solving and the real change or the real impact you’re having first and work backwards from that.
Nothing matters without, so what and context is king when it comes to data or technology or anything. Just implementing something doesn’t mean anything. And if you don’t know what you’re trying to achieve in doing it, or at least what you’re trying to achieve, the execution is obviously ultimately where you see that.
2) Automation is the key to scale. If you’re not thinking about different processes or different places in which technology, and it doesn’t have to be the most sophisticated thing, we’ve got a whole lot of low-hanging fruit automation exercises that no one would characterize as AI that are going to help us out a lot, or even just standardization of practices helps us out a lot.
3) Finally, in terms of thinking about how you can improve, as an organization or even as an individual, don’t compartmentalize. Just because things have been compartmentalized in the past doesn’t mean that they aren’t linked. Anything that ultimately feels linked or correlated needs to be thought about as an enterprise-level solution.
