AI Wisdom Ep. 10: Architecting the Underwriting Foundation of the Future with Nick Lamparelli

Digital Transformation - April 29 2020

On this episode of the “AI Wisdom – Talking Innovation in Insurance” podcast, host Ron Glozman speaks with Nick Lamparelli, Chief Underwriting Officer, reThought Insurance. Nick discusses how emerging technologies are impacting the underwriting process and how insurance companies need to develop a framework to support the development of a strong business case and sound technology evaluation criteria for digital transformation. Click the link to listen or read the full transcript below to hear what the future holds for insurance.


Full Transcript

Ron Glozman: Hello, and welcome to “AI Wisdom – Talking Innovation in Insurance.” On this podcast, we talk to business and insurtech leaders about how artificial intelligence is transforming the way we buy and sell insurance. I'm your host Ron Glozman, Founder and CEO of Chisel AI, and a strong believer in the power of AI to help people work smart and enrich their lives. So, let's get into it.

Whether it's raising a barn, building a home, or constructing a commercial office tower, getting the foundation right is critical to the strength and longevity of the structure. The same premise can be applied to insurance underwriting. It's paramount to have a strong foundation and framework to guide the adoption of new technologies, the evolution of underwriting techniques and the architecting of new automated processes. I'm very pleased to have Nick Lamparelli, Chief Underwriting Officer, reThought Insurance and Podcaster at Coverager joining me today to discuss how new technologies can help the insurance industry make underwriting more efficient and customer friendly. Welcome, Nick. Before we jump in, can you please introduce yourself?

Nick: Sure. My name is Nick Lamparelli, I'm the Chief Underwriting Officer at reThought Insurance, a Podcaster at Coverager, and just an overall Insurance nerd and geek.

Ron: Awesome. So, let's jump right into it. I'd love to get your thoughts on...well, first of all, I'd actually like to thank you for participating in our 2020 Commercial Underwriting Priorities eBook. For those of you haven't had a chance to read it, we got about 12 or so thought leaders from across the industry to contribute a handful of thought pieces, and Nick was kind enough to contribute one. So, if you haven't had a chance, go check out his chapter. So, I wanna jump into something you said in the ebook. You mentioned that insurance companies need a framework under which they evaluate technology. I'd love to hear you go a little bit deeper on the ideas of a framework that you have in mind and any recommendations on how that framework should be used?

Nick: Yeah, I think there are a lot of folks in insurance that are sort of running towards technology, you know, so desperate to have solutions that they keep hearing a lot of the fancy buzzwords around technology. And then they internally run around and say, "How can we use this to solve these myriads of problems that we have within our organization?" And it's the wrong framework for problem-solving. It's the hammer and nail problem, right? Technology is the hammer, and everything looks like a nail, so you try to use it on everything. In reality, what I don't see in a lot of insurance organizations, they don't break down their problems. They don't have the proper framework to sort of isolate which technologies and how do we implement those technologies in order to solve their problems. So, my background is in predominantly property. And if you kind of niche it down with even more focuses in natural catastrophes, and there's a lot of different technologies, a lot of different things that you can use between catastrophe models, and high definition imagery, aerial imagery, third party data, and proprietary data that exists out there that can enrich the modeling process.

In reality, what I don't see in a lot of insurance organizations, they don't break down their problems. They don't have the proper framework to sort of isolate which technologies and how do we implement those technologies in order to solve their problems.

And all of those are extremely valuable, but it basically still boils down to what is it that you want? Why do you exist? Or what is it that you're trying to accomplish? Because that will tell you how much data, how much technology you need and which technologies you need in order to solve that problem. You know, if you're just a general property writer, you will require a certain level of data, a certain amount of data, and certain kinds of models will tell you what you need to do. But if you don't know why you exist, and you start to run towards technology, the technology providers will sell you all the technology you would ever want. But at the end of the day, you're gonna be spending a lot of money, a lot of resources, a lot of time trying to implement all of that without necessarily fulfilling your mission. So, the framework that I sort of think about is to sort of back the train up into the station to think about why do we exist? We are insurers or we're in the insurance industry. Okay, so we're dealing with risk. What kind of risk? What kind of risk are we comfortable with as an organization? What kind of claims do we want to pay? If you can't answer those questions in a really detailed and deep way, you can throw all the technology you want at the problem. You're still gonna have additional problems down the road.

So, the framework that I sort of layout is to first start out that Simon Sinek start with why, why are you in business? What is it that you're trying to do? Not all insurers want to write, all underwrite, or all pay the same set of claims. What claims are you comfortable with and to what level are you comfortable paying those claims?

In my business with natural catastrophes, we're dealing with really large claims that happen very infrequently. My job is to figure out how infrequent, how severe, that's my goal and so, we've developed a framework around making sure that we're as accurate as possible when it comes to that. And even then, there's only a certain degree of accuracy I can get to, then it's being comfortable in that area and figuring out a conservative way to make my stakeholders comfortable that we actually know what we're doing.

And so, we've detailed the amount of technology, which technology, how much data past that trying to figure out, "Okay, what don't we know? And how do we make sure that we don't get ourselves into trouble?" That's sort of the framework that I was describing. It doesn't matter if you're an auto insurer, or if you write product liability, or in your workers compensation or even life insurance. I think that's an inadequate framework in order to connect technology to the insurance practice that you're trying to fulfill, whatever that mission is.

Ron: It's so funny that you say that. When I was a young boy, my parents sent me to...after school, they wanted me to study math and computer science, and they thought that would really made a big difference. And so, they sent me into like, extra classes above my grade. And one of the techniques that they taught that I would say, has been probably the most profound across my life, irrelevant of what I'm doing, is to break down complex problems. And that was, I think the core of this framework is to take a problem that might seem too big or too difficult, or like you need too much data to solve it, and getting comfortable to, using your words, getting comfortable with understanding what parts you have data for, and what decisions you can make based off that data. And then really diving deep into it. And I think it works great for math, it works great in logic and computer science, but it also works in insurance. And I think that's so beautiful. Now, on the other side, how do you avoid getting into analysis paralysis where you just have too many options? So, you've identified a problem. You know, let's say your problem is claims and forest fires. There are probably many solution providers out there. How do you actually investigate once you know what that core problem is, the smallest chunk, how do you avoid analysis paralysis?

Nick: Yeah. Well, are you asking like in terms of evaluating the tech, or how do you get through that decision?

Ron: You know, if you have wisdom on both or either, you take it where you wanna go.

Nick: Okay, well, I'll start with decision making. I'll give you a specific example. So, our specialty at reThought Insurance is flood, right? So, there are a handful of key data points, that's probably going to be sufficient to analyze upwards of, getting up to like 95% accuracy when it comes to flood. But one of the key ones will be for a particular property, how high is that property lifted off the ground? We call it the first-floor elevation. Are you basically walking up steps to get into the property? Or are you walking in, for example, like a McDonald's, you swing the door open and you just step in? And there's no first-floor elevation or there are stairs, for instance? And that is an extremely important parameter in making decisions, right, for a flood, for us to analyze the exposure. Now, that is a data point that is, like I said, difficult to get. And from the third-party perspective, we have yet to find any third-party that is aggressively collecting that information. So, it's up to us to try to retrieve it. And we have different sorts of technologies that we can use to estimate what that 1st floor elevation is.

But let's assume we run into a situation where we can't collect it. There are trees blocking the view or whatever. There's some impediment that's preventing us from collecting that information. Does that mean we stop? In my opinion, that does not mean we stop because I can always, in my mind, say, well, "What's the worst-case scenario?" The worst-case scenario is that the first-floor elevation is zero, the building is actually on the ground. That allows me to continue underwriting because I know with any additional information, the risk is actually going to go down and not up. So, I can actually compute a premium from that assuming a worst-case scenario. And so that is one aspect of analysis paralysis is you're collecting this data and you're using these models or whatever. And what if they aren't accurate? What if you get paralyzed in your decision making because you don't have enough information? Well, the appropriate approach in those cases is at those specific decision nodes where you need a particular piece of information and not having it is preventing you from moving to the next step, you can always assume the worst scenario and keep the process moving.

And so that is something that we do on our side is we can get all the way from A to Z by identifying those specific decision nodes in our framework and figuring out is there enough information out there? Is there providers of that information? And if they can't provide all of it, what then? If we're gambling on this stuff, how do we gamble this, so we're not harmed? We assume a worst-case scenario and we keep moving forward. The second thing that we were discussing is in evaluating insurance tech's or evaluating different tech solutions, and how do you not get paralyzed with all of the different solutions? And I think it comes down to, again, to the framework that exists. What are you actually trying to solve? And are you willing to roll up your sleeves and actually examine the technology that's there?

So, I'm thinking of potentially different options here. But let's say, for example, a building footprint for instance. I am now inundated with companies that can use artificial intelligence that can scan down imagery and can identify a building for me and actually give me the building footprint. I could get paralyzed in the different options that are available. But my job is to kick the tires with all of them and actually figure out which of these companies can actually provide this at scale, which ones can actually fit into my underwriting process?

So, I need to actually roll up my sleeves and figure out will this seamlessly fit in? Where are there gaps? Where does it go wrong? And does that actually move the needle? Does that actually affect my underwriting process? So, it's not just even a question of the technology, but can that solution provider, can they provide that data, provide those models, provide that technology, in a way that fits in with the workflow that we're designing? And so I don't know if like it ultimately, if you I guess, to bust out of the analysis paralysis, you have to, I think filter down your potential solutions and really dig in deep to all of them and find out what the best overall solution is.

Ron: You said a couple interesting things that I think I hope people were noting down. One thought that came into my mind is a negative slope into somebody's house, like especially if you live in a basement. So, I was wondering, you said sort of if the worst-case scenario is zero, but have you ever come across like a false positive or a case where making the assumption of zero was actually not even severe enough? Like they should have been more severe. It's a negative slope.

Nick: Well, I mean, that's actually an interesting use case. So, what does zero first floor elevation mean? For us, it is we can identify the ground grade. So, we can identify the elevation of the grounds around a particular property. And so even with a negative slope, we know that the topography, whatever solution provider we're using, they can recognize that there is a negative slope. And what we're saying is that the first floor of the property is zero compared to the ground that's immediately present around it.

Ron: Gotcha. So, very important that you have trans... not transparency, but that people especially we have the same problem have different definitions sometimes for the same word. And so, I completely understand with your definition, like that makes sense. Now, would you consider this an example of frictionless insurance?

Nick: That's a good question. I'm not sure. I don't know if I have an answer to that.

Ron: Yeah, I mean, for those who don't know, when we talk about frictionless insurance in the spirit of defining what we're talking about, straight-through processing is another word that I often hear. And I don't know if...yeah so that one clicked for you.

Nick: Yes. Okay. Yes, it did. Yes. I was trying to think of the context of where that would fit. And so, I think this nicely segues from the framework that I was describing to your mission. And I think ultimately, you have buyers of insurance. They do not need to see any of that sausage-making that's going on in the background. So, all of that decision making, right, our job is to make it as easy as possible for them to buy insurance, right? So, it's our job just to determine does that buyer fit into the type of ideal client, ideal customer, ideal risk, that we're looking for? And it's also our job to figure out what set of questions do I need to ask them and how much other data do I need to retrieve in order to make all of those decisions? Anything that's outside of just the base level questions that I asked them has to be done by us. And so, the whole frictionless element of the workflow, the underwriting process, is absolutely ideal. And it must exist for the insurance buyer simply because they have lots of options. They have lots of options in order to go and buy their insurance. And it is I think pressing upon us in the industry to make sure that we make that as frictionless as possible. We ask them as few questions as possible.

And frankly, I think in many regards the advantage or it's advantageous for the insurer to ask as few questions as possible, because we have so many other data sources in order to capture that. And so, for me asking a property owner if they have a negative slope, asking the property owner what their first floor, elevation is, those are technical questions they may not even know the answer to. And so, delaying or adding friction to the buying process, it's just an impediment for them to buy the coverage, one. I think secondly, you may not get the right answer anyways. So, you may get something that feeds into your model or your framework, that's kind of either red flags it or kicks it out. And I think that's ultimately bad for insurance. In my area in natural catastrophes, ultimately, I will know a heck of a lot more about the property and the risks associated with natural catastrophes than the property owner will.

And so to make this as frictionless as possible, it's my job to figure out what the important aspects are, go fetch those, go fetch that information, analyze it in such a way, and make sure my property owner, my customer, my brokers never see the sausage that's being made.

That's my job is to keep it clean, but it's also my job to let them know, here are the key pieces of information that I use in order to generate a premium and figure out how much coverage and how much a limit or what the deductible is. That way they have a fighting chance of understanding, the broker has a fighting chance of understanding and knowing what I'm looking for, what's ideal, and where they can continuously send me the type of business that I feel like I can be extremely competitive with and not something that I'm not gonna be competitive with, where it's just a waste of everybody's time. So, it was a long circuitous route to describe frictionless in my mind. But to me, that's sort of how I see it.

Ron: And it's so true because as policyholders, as people on the receiving side of the policy, oftentimes those technical questions are very complex. And so, if you can take that away, that helps. And on the flip side, we do a lot of work with brokerages and one of the things that they would love more insight on is what is your appetite? Sometimes they actually don't know. So, I think the way that you do business with them is actually advantageous.

So, we’re going to take a 20-second break to tell you where you can find more information and insights about insurance innovation. We’ll be right back.

[If you liked this episode of AI Wisdom, subscribe to our blog, Writing the Future: AI in Commercial Insurance at for feature articles, interviews, opinions, and more.]

Ron: We're back with our featured guest, Nick Lamparelli. Let's jump right into it, Nick. So we've been talking a lot about data and keeping data sort of as much as possible away from the consumer so as not to overburden them because we have more data than they recognize and more data than they would be able to provide themselves. Would you actually say that we have too much data? And is there such a thing?

Nick: Probably yes. I mean, there's so much data out there, and it's growing at an exponential rate, most of it is probably noise, right. But it's hard to know ahead of time what is the noise and what is the signal? What part of the data is actually valuable? And what is the type of information that you can or the type of data that you can ignore? So, it's the trying to balance off what is data and what is information? Or trying to take the data and convert it into information. So, the answer is yes and no, honestly. Like, there's so much data and most of it is probably not useful. But it's our job to figure out what is the useful part of it? How can we convert that data into useful information that can get someone coverage? Or if you think of all the risks that exist, all of the exposures that exist in our universe, the vast majority of them are not covered by insurance, right? It's the overwhelming amount of risk is not covered by insurance. And so given all of the data that's out there, how much of it can be converted to some useful information that could take an exposure and allow us to create an insurance policy, allow risk transfer to occur within the insurance industry to then cover it?

So, there's these massive business opportunities because most risk is not insured. There are these massive business opportunities that exist if we can just take all of this data and figure out ways to manipulate it, to analyze it, to model it may be so that we can generate the correct information which can lead to insight which would then allow us to do to create insurance products out of it. So, it's both. There's probably too much data but it's our job to kind of sift through it and figure out what, within this mountain of data, what is useful.

Ron: That's right. That's so true. And it’s part of our focus at Chisel is 80% of data, based on most estimates, is trapped away in documents. And there probably is too much data out there and the hard part is getting to it. And once you actually get to it, filtering out the noise and having the right data there. So, I love where this is going. So, let's take a slightly different approach because we've talked a lot about underwriting. And I'm curious to hear your thoughts on whether and how AI has an impact on underwriting. And maybe it's part of the core technologies you were talking about that you have available to you that people don't even know. But if you can go into a little bit more detail, what are your thoughts on AI in general? And how do you see it being deployed today?

Nick: Well, there's a mountain of data that a human being could never's impossible for them to interact with. There's just so much data. And so, AI, I think, is scaring a lot of people because they think of robots and losing their job. And I guess I think of it completely differently, in that I need help analyzing this data, and I'm looking for a competitive edge. And if someone can design a software package that's going to be somewhat intelligent and speed my ability to sift through that data and glean insight from it, I'm all for that. Because there's just too much information.

From an underwriting standpoint, AI hasn't even scratched the surface yet. We're just starting to figure out and try to use it in ways that just can allow us to breathe, allow us to get our heads above water when it comes to just this gigantic volume and a tidal wave of data that's being thrown at us.

So, we're just scratching the surface. But with that said, I think we're many, many, years and probably decades away before it's as ubiquitous as other technologies have been. So, I think it's just starting to kind of enter its way in, but I think we can circle this back all the way to the beginning of your first question. If you don't have a framework, if you don't know why you're in business, then AI is not going to help you. You know, the AI is not this magical thing that's gonna come in and just say, "Oh, you should be writing in this business. You should only write these risks." If you can't feed the appropriate questions and be able to explain what your mission is and why you exist and what you're trying to accomplish in your business, AI is not gonna be helpful by itself. It's not that magical bullet. But if you have it, if you know where you're going and you know what path you wanna go down, AI can really do a lot of the heavy lifting when it comes to getting you to that next step or helping you traverse the chasm that you see in your particular business.

If you know where you're going and you know what path you wanna go down, AI can really do a lot of the heavy lifting when it comes to getting you to that next step or helping you traverse the chasm that you see in your particular business.

So, I may have a different view on AI and and what it can do for business, but I think that's only because I think we have direct. We know what our mission is, we know what it is we're trying to accomplish. So, we can immediately identify the mountain of data that we have to sift through that we think is going to help us cross that chasm.

Ron: And are you just saying that or are you referencing the famous book that every businessperson has probably read and has on their bookshelf somewhere, Crossing the Chasm?

Nick: Probably subconsciously, I didn't even realize I was doing that. So yeah.

Ron: So, I'm curious, because obviously, you mentioned you have a little bit of a background on the real estate side. I'm curious if you've seen may be outside of the world of insurance and real estate, if you have seen technologies being used in the real estate industry that other industries could learn from? Namely insurance, but even more broadly than that, is there something that you're seeing that they're doing that isn't really happening elsewhere?

Nick: No, I, AI has been used, you know, we have AI in our pockets, right? So, we pull out our smartphone and whether it's Siri or one of these other tools, we're already using it. And we're already like growing accustomed to it. And so I think that the biggest problem that we're running into is that the technologists that are building AI for, you know, whatever industry they're in, but let's take, I think, let's just take tech, like search AI that we have on our smartphones, for instance, what I see happening is they're trying to take that and transport it, globalize it to other industries. And that's not going very well, because these other industries just have different pain points. The way that data flows and the workflow that occurs there is just vastly different than what happens in technology. In the technology biz, it was custom made to do what it's doing. Its custom made for AI, everything's digital, or it can convert everything to digital. And other industries can't do that. It's just not that simple.

You reference how roughly 80% to 90% of information with an insurance company needs to be converted into some form before it becomes useful. So that's the audio, the video, the paperwork, it's just sitting there, and it's basically wasted data. It's sitting there.

So, whether it's insurance or whether it's in other industries, I think, they'll be...the ones that can convert to digital, convert their data to a digital format, will most benefit. So, where I think AI could be most valuable is in step one, phase one. And I think this could be any industry, but I'll speak specifically for insurance is helping to convert the analog data to digital. So, then we can then go to step two. That I still think is gonna be like a decade away from really being ubiquitous. There's a lot to sift through. And companies really have to figure out, again, going back to that framework, they have to figure out what is it that we're trying to do with this? If all we're trying to do is just convert all of our paper to digital, okay, that's step one. But if they don't have a long-term plan, then all they did was convert a mountain of analog data to a mountain of digital data. They still have a lot of work to do.

The AI promise, I think, it's so vivid, right, in technology, like we see it and we're using it now on a day-to-day basis. We're now seeing in cars and stuff like that. But in some of these other industries, real estate, insurance, other areas, it's gonna be a much longer curve before we're there. But I think the companies that are willing to be early adopters will be the ones that have a mission. Again, getting back to that framework. They know why they exist in business. They know what they're trying to do. They know what they're trying to accomplish. It's not just about making money. And I think that maybe that's like, ultimately the biggest sin is that where you're not in business primarily to make money you have to solve the problem first, before you can make that money. And if you can't identify that, why do you exist, then there's no amount of AI that's going to help you.

Ron: That's right. And it comes full circle. You have to know your problems; you have to be able to identify them. And it actually just brought something up that when I meant to say when we first talked about this, which is sometimes what people don't realize is that they're looking for multiple solutions. Because companies often like to leverage their buying power and the fact that they can get economies of scale, which is always good, and you should always try for. But at the same time, you also have to recognize that sometimes there isn't a one size fits all solution. And I'm thinking of a couple projects that I've heard about where companies have gone out looking for somebody that can do, just arbitrarily as an example, claims intake, and let's say submission triage, and policy checking, and underwrite properties, and do analysis of angles. And there is no solution out there that can solve all of that. And that's where breaking it down into what are the problems you're trying to solve is really, really important. And so, I think this has come full circle, so let's wrap up on this. If you could give one piece of wisdom, and now it doesn't have to be related to insurance, it can be anything, any piece of wisdom that you'd like to share with our listeners, please feel free to.

Nick: Well, I think this ties into business and not business, which is in I think the vast majority of decision-making cases, there is no right and wrong. It's trade-offs. It's what do you want? And what's the optimal way for you to get it? So that means making decisions where you're going to have to sacrifice something, right? And in order to get something else. So how do you, you know, those trade-offs or how do I manage my risk, my time, my resources, my opportunity costs, what is the priority that I'm trying to accomplish here? You can't have it all. That's basically what it boils down to. And when making decisions, you have to sacrifice something. Because you're not gonna be able to get it all. You're not gonna be able to get all of the information that you need. And the information you might want or might need might be too expensive. So how do you balance all of those different trade offs in order for you to really focus and deliver on your highest priority?

And so that's probably, I don't know if that's wisdom but that's basically what I've learned. Being in business and living on this earth is that, for example, your priorities might not be mine. Mine might not be yours. Therefore, we can't make the same set of yes or no decisions on everything. And so, you have to filter those things down and balance your trade-offs.

Ron: I think that message was spot on. So, Nick, if people wanna find out more, where can they find you?

Nick: They can find me; the best place is on LinkedIn. Just type in Lamparelli. Hopefully, I'm number one on the list. I am active on Twitter and my Twitter tag is @Nick_Lamparelli. Those are the two best places to reach out to me.

Ron: Awesome. Well, Nick, thank you so much for taking the time. And as always, you guys can find us on Twitter, and you can find us on LinkedIn, and of course our website. Nick, thank you so much.

Nick: I appreciate it. Thank you.

Ron:  That’s a wrap for this episode of “AI Wisdom” hosted by Chisel AI and me, Ron Glozman. Thanks for listening.

If you like our podcast and want to hear more, check us out at or tune in and subscribe wherever you get your podcasts: SoundCloud, Spotify, iTunes or Stitcher

Join us next time for more expert insights and straight talk on how AI and insurtech innovations are transforming the insurance value chain. See you on the next episode!



Browse different topics

Recent Posts