Get a demo
Select the type of demo you'd like to experience with Indico Data.
Select the type of demo you'd like to experience with Indico Data.
Watch Christopher M. Wells, Ph. D., Indico VP of Research and Development, and Arthur Borden, Underwriting Process & Tech Expert in episode 9 of Unstructured Unlocked. Tune in to discover how underwriting leaders are solving their most complex unstructured data challenges.
Christopher Wells: Hello. Welcome to another episode of Unstructured Unlocked. Your host, Chris Wells, VP of R and D at Indico Data. My guest today is Art Borden. He’s a veteran with over a decade in the underwriting process and technology space. Art, welcome to the podcast.
Art Borden: Well, thank you, Chris, for bringing me in this morning. I’m looking forward to it.
CW: Of course. Yeah. This should be a good one. To kick it off, can you just tell us a little about who you are in your background?
AB: Sure, Chris. Well, thanks. I have many years in the industry. I lowly beginnings as a claim representative back at Allstate coming out of college and found insurance to be an interesting career path and have stuck with it since that time. So I have a full career in it. Most recently, I’ve been involved in a lot of the underwriting process and technology work in the industry. And so I’ve lived this space for quite some time and have, you know, have spent many, many hours as probably many of the folks who will end up seeing this have trying to figure out how to move the industry forward given some of the new technologies that are dropping into our environment.
CW: Yeah, it’s an exciting time for insurance. There’s sort of a tech renaissance going on. As an undergrad, I did not find the insurance space Exciting. Although I’m starting to change my tone on that. What was it that sucked you in?
AB: You know, what I’d say about the industry is that you can get in and do many different things. So I’ve been inside of the claimed space and then account management and, and direct customer-facing roles and, and then kind of moving into business architecture and technology and, and all of the development opportunities to try to bring technology to bear in the underwriting process. I’ve managed in my career to do a lot of different things, and I think that people have, who have been in the industry for a while, discovered that there are a lot of different opportunities within the industry. I can say. Honestly, I’ve never been bored in my career. I’ve always found something that excited me about the work that I’m doing, and it’s carried me through my career.
CW: That’s exciting to hear. Yeah. For those of you who are starting your career, you know, don’t take, you know, don’t sleep on insurance. A lot of interesting stuff going on.
AB: Believe me, it was the furthest thing from my mind when I got out of college. I’d never even considered it a possible career. But, you know, once you’re in, you sort of realize the breadth and depth of it and the passion that people bring to the industry. It’s, it’s a huge marketplace out there. It’s, it has huge opportunities for anyone coming into the industry. So my view is to give it a shot because there’s a lot of good stuff in the industry today.
CW: Yeah. Okay. On that note, I know that you’re at this point in your career, and you’re in another exciting transition time, but I wonder if you could sort of turn the clock back and tell us some of the highlights of your career so far. You know, places you’ve been, things you’ve done, people you’ve worked with.
AB: Yeah. And, working out in the field in front of the glass, you know, trying to process business, you get this opinion that you could go into a home office and change it all and fix it in a day. And what you realize is the complexity there is just a bit overwhelming. And, when I got in, you know, it was kind of the.com kind of space, and I was very interested in the technology piece, and that’s what drew me into some of the process and technology work I ended up getting into. And honestly, the things that I found attractive were business architecture and how you think about how people, processes, and technology fit together to make the process and the company successful. And, you know, if you don’t have all three legs of those stools working in harmony, you have a failure to launch many of these initiatives.
So I liked that space, and when I got into it, I did a lot of process redesign work, a lot of connection to emerging technologies and how those sort of maybe support each other and what is the organizational change management implication of all of that, because, you know, it can be seen as a stodgy industry. Still, you’re constantly, in a state of flux, in a state of change. How do you get folks to understand the new possibilities in terms of what the technology can bring to you? And, and, and so I sort of worked through that as I came into that side of the industry and did a lot of business process redesign work and, and work carefully with the field trying to understand their pain points. You know, kind of carry that works through my more recent cons career path, which has had to do with a lot with technology and transformation.
CW: Interesting. There’s a whole lot there that I want to poke at. Let’s start with what you mentioned going back to the.com era. Insurance products are complicated animals, as you mentioned, and there’s a whole lot that goes on behind the curtain to make sure those products get delivered well. How have you seen the role of technology in that stack change in the last couple of decades as both the industry has matured and you’ve advanced in your career?
AB: Yeah, so there’s been a huge desire to adopt more modern platforms for policy administration. So you see, entrants like Guidewire and Duck Creek have come into the marketplace and have driven a lot of product evolution, and product enablement work for most modern insurance companies is a good part of the marketplace, and they continue their work. You know, what you learn in that process, and something I had to learn as well, is you describe the complexity of the product and being able to enable that and how it’s so interconnected. The products are so interconnected within the insurance industry. Yeah. Right. So you have to be able to underwrite it, you have to be able to rate it, you have to be able to bill it. You have to be able to send downstream to the various states and bureaus all the information that they’re interested in keeping track of.
You have to be able to consume all of the data that comes from that actuarily and sort of revise your products and rates as you go forward. And you begin to realize the complexity that sits behind all of that and how hard it is to change while all this is happening. You know, you’ve got this evolution of technology from this very difficult-to-compose language of COBAL and such that, you know, really was the way that these systems were built to more sort of low-code, no-code type of approaches, and being able to have power users at, at a desk modify products and, and make those updates. And that in and of itself has driven organizational change management issues as you’ve looked to make those processes more efficient.
CW:
Yeah. So, that brings me to another item from your introduction, which is also a theme of this podcast. As I talked to, you know, folks from many different industries. How, how have you seen, you know, you talked about process redesign. How have you seen the relationship between the folks that are trying to automate the process and the folks that are trying to understand and maybe, you know, retool the process evolve in the last, I would say, let’s, let’s call it five years? How have those partnerships changed?
AB: Yeah, so you have in the, in the holistic delivery of, of these technology projects, you’ve seen the onset of agile style delivery and being able to have folks from the business side of the equation sit side by side in a very collaborative way with your IT partners Yeah. Has really been a, a huge shift. I mean, what you historically had was a more waterfall type of delivery where you’d have folks gathering a bunch of requirements going off and returning six months later with a product that didn’t look very much like what they had hoped for. And now, at least in, in the way that products are delivered now, you have this kind of collaborative type of spirit. I think that has continued to evolve within the industry. Now, some folks are early in their agile journey. Some have been doing this very formally for a number of years now.
I know for us we made that conversion, and it is a little difficult to, to get your mindset around set up properly around that. It changes how you gather requirements. It changes how you talk to one another, but once you get there, it really does make a huge difference in terms of delivering a quality product to, to the business. So, you know, those types of things have really changed how I would say technology and business have worked together. Again, it brings forth a change management issue, right? I keep coming back to that, but it really does force a change in terms of how folks think about building out some of the, the systems and capabilities that are needed today to be successful.
CW: Yeah. Yeah, I couldn’t agree more with that. It’s interesting you mentioned cobalt you know, on mainframes, and I think a lot of folks who you know, like myself, were outside of the insurance industry initially scoff at that. But the reason cobalt and mainframes has stuck around so long is because it works really well at scale now. It’s hard to change, but it works really well at scale. And I think thematically as we get started here, one of the things you need in, in any industry, but insurance may be more than anything, is stuff that works well at scale. And I think, again, outsider’s view, one of the things that, that the industry is starting to understand is that you have to have change management that works at scale. So I’m interested to hear more about that as we go.
AB: Yeah, those applications, those old applications are bulletproof. They really are. And they just keep, they just keep chugging along. And, but you’re right, when it comes to changing them and, and, and updating those products, just understanding, I mean, if you, if you’re inside of an industry that has a big legacy issue, it gets hard to even understand what the code is doing inside of the application. So you find yourself having to mine the old rules out of the old system just to even understand what, what you want that system to do your next system to do. And that’s really time consuming is difficult. And boy, I’ll tell you, try to get a new programmer that’s just come out of college who, who you then tell, Hey, for the next 10 years, I want you to really focus on coball. You’re, you know, it’s tough to recruit those folks too. You see many people coming out of, of retirement to do these types of projects because they’re the ones that learn that code and, you know, are still the ones that can help you.
CW: Yeah. And, you know, those new programmers coming out don’t have the context that, like, you know, the reason the languages you like, like Python, exist is that lots of people have figured out the tough stuff like cobalt and c c plus. So yeah, it’s a great perspective,
AB: But it’s interesting. Now, I’d say one more quick thing, the migration to the cloud and bringing this data out is sort of softening the blow a little bit in terms of the legacy because, quite honestly, if you can get the data over there, you can then lay in some of the newer technologies on top of the data sets that you’re Yeah. You’re building in the cloud and, and it does help you get around some of your legacy issues.
CW: Yeah, that’s a great point. The, the cloud really does help you decouple these things from one another almost by design. Okay. Let’s take a couple of steps back. Now that we’ve gone all the way to mainframes. Tell me, so a lot of our u a lot of our listeners are not in the insurance industry. They come more from the technology and automation and data side. So if you could take a, a little bit of a step back and just tell us what an underwriting organization looks like, you know, from the tech side, the management side, product side, just whatever would be useful in understanding where you’re coming from.
AB: Yeah, so I’ll, I’ll talk about my most recent experience. We had a centralized operations group that was responsible for receiving the submission data in the door. Maybe it came to the underwriter first, but it got flipped over to operations to get it registered in our systems. Okay.
CW: And if I could break in just when you say submission data, like what’s in that package?
AB: Yeah, so if you think about underwriting as a lot like a mortgage application, right? So you have a lot of information that describes the company that you are looking to underwrite, and oftentimes that’ll be a description of their operation. It might be a list of all the locations that they have, the number of employees that they, they employ, the types of coverages, and things that the company is looking to purchase from you. Maybe a list of automobiles that they have. And I’m thinking about this more from a commercial market, a little bigger sized risk, but that’s kind of a profile, right? There’s a lot of information to know, just like the bank needs to know a lot about you, they need to know your assets, your, your wealth, a description of you, description of the property that you’re looking to underwrite.
All of those things are kind of true in a similar kind of view of what insurance is about. And that information comes in and it’s basically a request from a producer, a maybe what might be called a broker, somebody who’s trying to place the business with you and help that customer, okay. Work through all of this. So they assemble all of this information, and they sort of send it into the insurance company asking for a quotation on that particular risk. And once that hits our doors as an insurer, we have to do a few things. We have some legal obligations to make sure that that company that’s coming in the door is legit not foreign own, you know, all of these types of things. Those are rules that we have to follow.
CW: So know your know your customer types
AB: That’s it. And then we have to get it into our systems because the, the goal is to start to move this through a process that, that sits within our organization that might start out by getting it through the operational piece first. There’s some data gathering and analysis. The underwriter has to do preliminarily to say, Hey, this one kind of a quick thumbs up, thumbs down to see if it’s gonna move forward or not. And then, you know, if you decide that you’re gonna go after it, and again, on a commercial space, you would then start to throw a lot of resources at it. You’d send it to someone to actually enter all of the information about that risk into a system. That system can calculate rates. You will do some analysis about the nature of risk. What types of exposures do they have?
Are they driving cars around the city or is it out in the country? Are they doing hazardous type of work or are they office workers who are, who are less exposed? So you start to, to get at what is the nature of the risk that you’re ensuring, and those carry different rates, different different considerations that you, the underwriter needs to go through. Now, if you think about every type of industry in the country, there’s a huge diversity of that type of question set that you’re applying to Yeah. Each, every risk. And you really, you know, that’s where the rub is for the underwriter. They’re having to know kind of looking at that risk, all of the possible things that can go wrong, right? Yeah. and, and that’s a difficult thing, right? There’s a lot of information they have to keep in their head, and that’s the role of technology in this, right?
How do you use, use the technology to present to the underwriter a set of information that allows ’em to accelerate what I call decision velocity. You want them to be able to make good decisions fast about what they’re seeing. Now, for smaller risks, you do that through straight, what we would call straight through automation, right? So you build up enough rules in a system, or you use enough models to simulate what that might look like, and you push that through without a human touching it. As the risk grow, the ability to build those models to to confidence starts to strain, right? You can’t necessarily count on underwriting Ford Motor Company through a, a model so easily. Okay.
CW: Right. So let, go ahead, let me, let me hit the pause button there. What sort of, what fraction would you say are, you know, go through straight through based on rules or, or whatever else?
AB: Yeah, I think everybody every company has a little different cut at that. Yeah. you know, we tried to run 75 to 80% of small business and maybe 10% of middle market business. Okay. So there’s a, a big difference, right? If I’ve got a a storefront shop that’s doing luggage repair, let’s call it, and I’ve got this small, small shop, I could pretty effectively use straight through processing to figure out what is the likelihood of loss there and come up with a price that’s competitive. You know, there just isn’t, there aren’t that many exposures there, you know, but when, when I’m trying to underwrite a, a light manufacturer who’s building parts that might go into an automobile, for instance, the potential complexity there grows kind of geometrically. Okay? And so you have to be a little bit more careful about pushing that straight through, and maybe the underwriter needs to touch a few parts of that that decision stream. Now, you can get a lot of it to go straight through, but it starts to get more complicated.
CW: Okay. And so that, that’s a great point. I was gonna ask you like, which axis, you know, you decide to automate or not where you make that decision. So complexity is a big piece of it, but of course, you know, models and technology can sort of, you know, keep pace with complexity. So what is it about, what is it about this complexity that makes it hard to automate? Is it, you know, is it something regulatory? Is it business need to have a human eye on it because of the relationship with the customer? Or is it just we haven’t seen enough big deals like this, they’re rare enough that we, we don’t really have a good sense of what the rules might be?
AB: Yeah. So I’m gonna say yes,
CW: Yeah.
AB: Fair. Because I think it’s multivariate, right? You have a lot of different inputs to that equation, right? You might have you might have severity issues. The potential for a large loss could drive you to have to touch that risk, right? So, ah I’ve got some exposure that that could be significant, and it could car charge, it could cause a large loss, but maybe the maybe the, the automobile fleet that they drive is not risky at all, right? That they’re just, they’re just gonna smoke, you know, kind of tooling around, and you sort of know what they’re gonna get into. And, and so you can underwrite that risk pretty effectively. So there is a lot of complexity by product. There’s a lot of complexity simply by the nature of the risks that you face.
CW: Interesting. Would you say that there are certain products that are, let’s, let’s, let me ask this. Two ways historically have been ripe for sort of automation, and then, you know, what do you think is the next, the next one that’s gonna be, you know, taken off the wall?
AB: Yeah, certainly, certainly that’s true. There are certain products that’ll, that are a little easier to underwrite. You know, auto tends to be that way, though. There’s a complexity there that, that is pretty astounding at times. Property can be very straightforward, it can be very complex if you think about some of the global warming impacts to, to the, the world, right? And so flooding risk, and fire risk and such is really kind of morphing a lot. So it, you know, products can go in and out of favor. Cyber liability is one for the longest time who really worried so much about cyber liability. But now every company is, you know, think about Amazon getting shut down because of oh yeah. Of getting hacked. It would be a huge, huge issue. So, and
CW: The, yeah, even last pass just got hacked, right? Yeah. Crazy.
AB: I know. You just, you would not have expected that. So, so here you go. So yeah, there’s a lot of variability and, and why things are more risky than others. And honestly it’s not static. It can change. You can suddenly have a new emerging issue. I will, I will hazard that batteries in in electric cars may end up being something that people start to be concerned about because containing fires and things like that. And, and honestly, that’s never been a concern because we just haven’t had it before. You know, that type of thing/
CW: Yeah, yeah.
AB: Interesting. So but, but what, what is the industry trying to do? Well, yeah. You know, that idea around, you know, can we get all of this information gathered efficiently so we don’t have to have a lot of folks typing it into systems? So can I, can I machine read my submission? Yes. And if I can re machine read my submission, then could I go out to third party data sources who know a lot about this particular risk or this particular industry, and gather information electronically that keeps the underwriter from having to go research it themselves? Yeah. And can I put that into a model and can I output something that I can feel comfortable with? Those are the types of things that the industry is trying to get to, to save the human human-centric sort of, you know, wasted energy that people go through today to round out their understanding of a particular risk.
And, you know, it’s a, it’s a very difficult to find a good underwriter these days. The, the economy as such, you know, when you’ve got an unemployment rate ticking along at three, I think in the insurance industry, it’s less than two. Wow. Some people would argue that it’s zero or negative. Yeah. that it’s very, very difficult to get a good underwriter. And oh, by the way, teach ’em everything about your archaic legacy systems that they need to know, which is a very difficult thing. So getting, getting to a point where you can leverage the technology is really a big deal.
CW: Yeah. That’s interesting. You you mentioned a couple things that I want to tie together there. We at Indico, we talk a lot about automation and, you know, processing, you know, by bots or AI or whatever it is, not as being something that displaces the human from the process, but puts them at the center with a bionic arm so they can move more faster. And it sounds like the insurance industry is thinking about these processes the same way. In fact, it sounds like you need more people you just need them with better tools at the same time.
AB: Yep. Yeah, exactly. And I’ll go back to the phrase decision velocity. You really are trying to enable an underwriter to make their decisions faster with more confidence and less touch back to the insured or the insured’s broker. Because if you’ve got two equal companies that are pestering your producer for information that one can gather electronically and never has to ask the question, and the other ones on the phone to them every single day are, are firing off emails, trying to clarify de details. The broker, the producer, is gonna be more interested in working with the efficient company. And it’s faster, it’s less hassle for them. They have to have fewer people on staff as well to support their business. And, and it has a, a cascading effect. So it’s not only making the d our underwriter is more efficient, but it’s also putting a face out to the marketplace that says, Hey, we’re easy to deal with. We understand you, we go get information we don’t have to ask you about, and we’re able to make decisions very quickly and get you an answer. Because getting an answer quickly for them is a really important part of their jobs.
CW: Yeah, yeah. That, that totally makes sense. And I, I want to swing the spotlight back to you a little bit. And I have a feeling that that concept of decision velocity is a key to answering this question, but how was success measured in your role?
AB: Yeah, yeah. So for us you know, I had a, most recently had a couple of areas of accountability. One was product management. So all of that product complexity that you talked about in terms of that is my success was measured by making the systems flexible, a as flexible as possible when I updated product and brought it into to the system. And so I had accountability for implementing those products. And, and that for me was a big part of it. So what does that all entail? Yeah. Well, it entails figuring out how to best define those products, how to work best with my IT partners to build out the systems that enable those products. All of that is part of the role of someone in, in my role. And then secondly, the, the other piece was how was I measured for success with respect to my peers is another okay aspect of this.
So my peers would be the product owners for those particular products. So, ah so I have to help them build out updates to their product suite and give them new technology to speed up the underwriting process. So if I was able to implement a process or a procedure or a technology that helped speed up the process for the underwriters, that was a win. And they, they would count on me to do that, right? To actually implement systems that would make the underwriting process more efficient for them. So that was bringing in new technologies, it was thinking through their process, taking out waste and, and such in the process and, and building an additional automation to support them.
CW: Okay. Interesting. And that, that sort of guides the conversation to the next phase, I think, which is we’ve talked, you, you mentioned submissions and intake. And so I wanna, I wanna, there’s a bunch that I wanna get at there. The podcast is nominally about unstructured data, and I have learned recently that you know, insurance submissions are full of it. So talk to me a little bit more. You, you, you started talking about it a little bit earlier. I want to, I wanna drill in now. What is in a submission and where does it come in the door? Is it a, you know, is it in the mail? Is it a fax, is it email? What’s, what’s in detail? What’s in that bundle of stuff that comes in?
AB: Okay. I’ll say yes, all those be part of that. There’s, there’s various sizes of risk, but I’ll, I’ll deal with a little bit more complexity here to try to some sense to it. So if, if I’m dealing with a mid-size manufacturer, and that submission comes in, I started to, to sort of tick it off before, but there’s a financial document that explains their financial health, because we need to underwrite companies that are solid financially. If they’re not solid financially, they could go bankrupt. That’s a problem for us because the the care and custody of their employees could suffer as they’re going through bankruptcy. And we wanna be careful about that. But then we would have a list of the, the exposures that they’re bringing to us. So if they have an auto fleet, they want us to ensure there would be a list of all the vehicles, including the VINs, the type of vehicle, the value of the, of the vehicle.
All of those would come to us if it’s a, if it’s a property exposure, it’s gonna be a list of properties that they want us to ensure what goes on in those properties. Square footage type of construction of that property. Is it a frame structure? Is it a concrete structure? How close is it? What’s up to us to, to figure this out a bit? But how close is it to floodplains and such? Yes. So the, you we were talking about what is the information that’s coming in on a submission? So I would say to summarize, there’s e for every exposure type of coverage that you’re looking for the insurance company to underwrite for you, you have to describe what the nature of of the exposures that you’re asking us to cover. So the list of vehicles, the, the location of the buildings, the size of the buildings, the nature of the, the workers that are working in your, in your plants, et cetera. All of that comes in on an application. Financial information comes in your loss history, how many losses have you had? History. That’s another important thing because if you’ve had lots of losses, that’s of interest to us, if you
CW: Yep. You’re more risk,
AB: Yeah. Perfectly clean. You know, that’s important to know as well. It affects your rates. All of these types of documents and maybe some description of your operations, some information about your leadership team, all of that could be as part of a package. There could be pictures of your locations, there could be financial documents, there could be spreadsheets, there could be structured application forms. All of that could make up a submission that we receive. So from your perspective, like how do I electronically read that, right? Yeah, it gets to be a pretty complicated piece, right? So there are what we would call structured documents that you know, the industry has agreed upon application formats that you can use called Accord and Accord tends to structure what information is requested of the insured and, and, you know, on that accord form, what you can expect to see.
And then I would say there’s a lot of unstructured data, which people send spreadsheets and Word documents and so on. And you really want to be able to read those documents electronically and sort of know what you’re seeing there. And you know, I see submissions that are a hundred pages long. It’s not unusual. Wow. It could be a, a mishmash of all types of information that’s come in there, some of it conflicting some of it not accurate. So you really do have a, a quite a challenge when that thing hits your inbox. They tend to come in by email. Interesting. And so you’re looking at this a hundred page documents, and you’re having to figure out, well, is it a b ABC company, ABC co. Is it, you know, and, and you know, you get all of these kind of wacky sort of questions.
Yeah. We have addresses that don’t match. We have a lot of inconsistent information you have to reconcile. And that’s something that traditionally a human being has done that. Yeah. But you’re looking to use technology to do more of that. And the technology has come along in such a way that, you know, maybe, maybe it can read through the document and flash up on a screen for a human, here’s your two choices. Which one do you think it is? Here’s the confidence interval around that, and you should choose A, not B, that type of thing. And so those are the types of rule sets that you begin to build into this submission intake process to try to understand that submission.
CW: Okay. So let me, let me, I’m drinking from the fire hose here. So let, let me try to put some detail in here. So say the whole process, I don’t know, does it take for a complex one? Is it an hour? Is it six hours?
AB: Yeah, it again, you, so I’ll talk about process. So what is it that you wanna recognize? First do you just wanna get it in the door? That could take you just a couple of minutes to get it preliminarily in the system. Okay. Now I, is that enough to send it over to an underwriter to look at without actually extracting all of the information? Maybe. And then that might take five minutes, send it to the underwriter, then take a quick scan through it. Do they know this risk? It’s very possible. They do. And they already know the company, and they’ll say, yes, I wanna work on this. Then it starts to, to to be more complicated. Right? Because you are now looking at real rates, real analysis. Yeah. You gotta be very precise. You wanna make sure that you don’t make a mistake and misclassify this particular risk into such a way that could cost you later.
CW: So, and maybe the answer is a lot of different places here, but where in that sort of chain of custody do you first get a sense of, this submission has everything I need in it, or I’m gonna have to kick it back to get more information?
AB: Yeah. Again, what we’ve been trying to do is push that decision point to the very start. Yeah. Yeah. That’s what we would like to do. We’d like to know right out of the gate if, if it’s complete, if there’s key missing information, is it something we can go retrieve, or is it something that we have to go back to the insured on? And that you know, that’s been a bit of the challenge over the years, is to move that decision point all the way to the front so that when the system looks at that submission on the, on the intake, it knows exactly what it is, can strip off all the relevant information, the systems will need to rate it to price it, et cetera. That’s the goal. And that’s where the complexity is as well.
CW: Interesting. So what do you think is the, I’m sort of jumping ahead in the script here, but what do you, what do you think is the, the, the big obstacle there to pushing that all the way to the front? Is it a, is it a tech obstacle? Is it a people obstacle? Maybe both. Probably both. What is it?
AB: So I think the been challenged by what they, what it has to do, right? Okay. So it has, it has to look at that hundred pages and, and really decide what’s the, what’s the accurate information that needs, it needs to move forward. Now there’s a lot of potential in third party data sources out there where you could go somewhere and, and call a service that tells you everything I ever needed to know about ABC company.
CW: Okay. Yeah.
AB: Now, the question is, is the accuracy of that data. So you, you have to sort of shift your focus to say, I’m no longer maybe be hindered by the technology. I now am hindered by a process to monitor data quality, which is a little different. Yeah. Cut. So if that, if that data provider is 70% accurate or 80% accurate, which 20% is wrong, right? Yeah. And, and, and, and those 20% questions can really get you in trouble if you underwrite the wrong risk based on the wrong data. So, you know, that, that is a big concern of the insurance industry is, and there’s a lot of suspicion about, well, we can’t trust all of this data out there just yet. So there’s an evolution that’s going on in the industry to tighten that down and to build better models that sort of negate some of that interesting. And that’s a challenge for the industry.
CW: Yeah, yeah, yeah, of course. Bad data and bad decisions out, I guess. Yes. so that raises the specter of the question how accurate are the fully human processes and is that being monitored as well? Like, you know, is 80% quality data good comparatively, or is it sort of par? Right? I don’t, I don’t know enough to, to answer that question.
AB: No, no, that’s you’re, you’re right at the heart of it. So for instance you know, we talked a little bit about brokers and producers being sort of in the middle of the process. Well, yeah. They don’t necessarily have perfect information either. They pick up, okay, they call their insured and their insured they may be, may have not updated that particular bit of information or data in their systems in some time. So you start out with sort of imperfect there at the source, at the customer level. And then the folks that are interpreting that from the producer level they may very well not fully understand what they’ve just heard, and then they put it on an application and send it to us. Okay? So are we sit, are we sitting at a hundred percent perfection at that point?
Certainly not now is something that we looked up in the marketplace through third party data sources better than what we’re getting in on our applications. Again, there’s a whole methodology that you have to build to test those and retest those things because yeah, the data quality from the marketplace is actually getting better every year. So I may say, well, they’re 80% this year, but next year they’re gonna be 82. The year after that they’re 85. You know, so that’s, it’s not a one and done kind of yeah. Process. So you, when I talk about change management, do you have a department or a group that’s responsible for monitoring data quality? Hmm. Well, maybe you didn’t historically, because you never really called third party data to do this before. Yeah. But in the future, don’t you need it? Yeah, you probably do. You need to make sure that that stuff is well understood because you’re using, using it to underwrite your business.
CW: Interesting. I, over the years, I’ve taken lots of ML and automation projects to production and yep. I always get the question, how accurate is it gonna be upfront? And I <laugh> I always respond, well, how accurate does it need to be and how accurate, you know, the implication being, how accurate are the humans doing it today? It’s always Yes. You know, shrug, right? I dunno.
AB: Yep. Yeah. Right. That’s a big challenge.
CW: And, and you’re, you’re pointing out one of the reasons that this occurs is because you’ve never had to think of it. You know, it’s always just, well, we do the process and of course, we’re making money doing the process, so it must be, you know, a hundred percent accurate. But
AB:
You know,
CW: What, if it’s not, if you don’t check,
AB: What if your competitor builds a better process for validating that data and you are competing against them and you’re writing risks that you probably shouldn’t write because their data data process is better than yours. I mean, this is the hidden competitive pressure that the industry feels, right? It isn’t just that our price is better, it’s what goes into that decision, you know, are, are you a better analysis company because of the infrastructure that you built than your competition now that you may not win because of that in the first year, but over the course of time, you’re gonna win if you’ve got a better mile strap that way.
CW: Yeah. Yeah. Interesting. Okay. Let’s keep, let’s keep pushing in there. I, I interrupted you at some point you were talking about that handoff from the data intake to the underwriter. What happens from there?
AB: Yeah. So again, it, it really is how much technology to do you provide the underwriter in their analysis? So you know, we talked earlier about the diversity of companies that an underwriter might be asked to underwrite and all the, and having good knowledge of the exposures. Well, there are guidelines that most companies have about what’s a good risk and what’s a bad risk. Yeah. But there are so many of them. How does the underwriter keep those fresh in their mind at any point in time? Right? Right. So, can I provide technology that picks out the right rules that push red flags in front of the underwriter for exceptions that are inside of that submission that they’ve received that say, Hey, this particular location is in an area that you might wanna give them closer attention. Right? And, and look at that and ask more questions of, of the company about have you accounted for all these risks?
And those are the types of things that make for an effective underwriter, right? And, and so the questions for the underwriters start to be data driven, right? So they’re, you know, that an exposure exists within the submission, but now you’re trying to push out in front of them some guidance as to, Hey, this is why this, this is important because they could face this type of loss, and you should ask about this. Now if they miss that and they write it and the loss happens and not much you could do about it, right? It’s, it’s, it’s water under the bridge, so to speak. So those are the types of things that the, that start to happen after it’s gotten through its initial pa paces where you say it’s a financially sound company, it’s a type of market or a type of risk we wanna write. Now I have to really dig into it, and I can look at the losses, their prior losses, and the, that’s the more actuarial driven type of approach. But you also want to get at some of the more some of the softer analysis around the, the actual exposures that you’re being presented with.
CW: Interesting. So it, it sounds like it’s a, it’s a bit of art and a bit of science. Is that, is that what I’m hearing? Yeah,
AB: And that’s, that’s, it’s often said that way,
CW: And how, I don’t, you may not be able to answer this question, but how does a good underwriter become better at the art side of it? Like I can, I can imagine how the science gets better, better data, better tools, but how do you get better at the art side of it? What makes that easier?
AB: Yeah. Well, I mean, you know, tends to be, that’s where you’re looking for more experienced underwriters when you’re out there, someone that’s maybe seen some of this type of risk before you do it by segmentation to a certain extent. So say I, I want to write a lot of construction risks. Well, I build up a you know, a group of underwriters who have done a lot of this type of work, and they know the questions to ask. So you have specialization, but the problem is, is you can’t specialize in everything. So what tends to be the case is you’ll specialize in, in, where you have enough volume to specialize, but it’s very difficult to, to maintain that. So what’s the solution? The solution is sort of smart knowledge management that can actually push out in front of the underwriters questions and topics for considerations. Ok. That’s a really important part of it.
CW: Interesting. All right. Let’s swing back to a few things that you mentioned earlier. One of them you named, dropped a couple of technologies like Guidewire and Duck Creek, names that I, you know, people are in the industry have heard you know, in your career, which, as you’re thinking about augmenting and automating some parts of this process, what, what are the tools and solutions that you’ve had success with?
AB: Yeah, so you know, process flow is a big part of our world. And so, you know, you’ll hear about Pega System systems and Appian and Uncork. Enco is really hot right now because of the no code type of approach. So setting up the right processes to move by exception work to the right person to make those decisions is, is, is a big body of work inside insurance companies. Like, I’ve got this exposure. I might want to have this person take a look at this, and so I wanna move that particular body of work over to that person. And getting that business process management correct, is a big part of the insurance industry, right? Is making sure that that’s done correctly. Now, you know, the Duck Creek and, and Guidewire though, they, they would claim that they’re more than this, but they’ve really focused largely on policy administration.
So they, okay, they look at, you know, all of that complexity on product that you described earlier. Building those products, maintaining ’em, and building systems that make it easy to rate, rate policies and, and issue policies or things that Guidewire and Duck Creek have historically worked on. What they’re trying to do. Now, Majesco’s another one, and they’re trying to spend a lot more time on the underwriting risk assessment side of the house. They’re trying to move into that area in more of the underwriting decision support, less of the actual sort of mechanics of generating rate and, and forms. They’re trying to do more of the, well, is this a good risk or a bad risk type of, of assessment? But, but I’d say they’re earlier in that journey than than others. There are a number of companies out there now folks like send, which is a new entrant to the marketplace, or Fed Auto folks that are focused on the decision support things that we talked about, okay.
Pushing alerts to the underwriters and helping them manage their desktop. You know, we’re fresh from a a project of talking to a lot of underwriters out in the field. And yeah, you know you know, they, they talk about the scavenger hunts. The industry is life with, you know, trying to find the right data at the right time is really a real challenge. They could have six or eight windows open on their desktops at once looking for information. They could have a, a rating spreadsheet open. They could have a, a website open. I mean, it, it’s not, it, it is not surprising to see that activity go down for an underwriter in a lot of companies. And that’s that’s a real challenge for them. So there are folks that are trying to build more rational integrations that would sort of present this information underwriters a lot more effectively.
CW: Interesting. All right. So let me, let me try to put that all together. So in terms of like a high level architecture, and correct me, or I’m wrong here, you sort of have, you have like a workflow engine, right? Yep. and at the beginning of that is, you know, some sort of way of getting the data in as a, as a raw package. And then that I, I assume, you know, there are various sub-components of that workflow, then the policy administration system has to integrate with that workflow, I, I assume, at various points along the way.
AB: Yes, that’s correct.
CW: And then all of that has to route to the underwriting systems, whatever they may be, and there has to be back and forth with that workflow as well. That’s great. And I’m guessing there’s enough variability between companies for all of these components that in the near term, there’s not gonna be sort of a single system that rules them all. You’re gonna need the flexibility in each component such that sort of wiring these things together is, you know, is what’s viable rather than here’s we, we installed X and x just as all of it. Is that a, is that a fair read?
AB: You’re ready for the industry. But that’s appreciate <laugh>. That is, that’s, that’s a really good quick summary of, of the complexity that’s facing the industry is to try to make those various components talk to one another, and each, each of those components tries to step out of their, their little area and take on chunks of the other systems, right? And okay. Yeah. And so you, you have to
CW: Border border conflicts.
AB: Yes. Yes. Fit for purpose becomes a really important you know, kind of consideration as you’re choosing to purchase these systems and to implement them, you know, which are your systems of record. It’s a huge, huge topic in, in our world in terms of figuring out which system is really gonna govern the interaction with the underwriter, and that that is a real source of confusion and debate as you’re building these systems.
CW: Interesting. All right. We’re coming up on our time together. I feel like we could spend the rest of the day talking about this, and maybe, maybe I’ll, maybe I’ll ask you about that later on. But bringing it back to unstructured data you talked about integration of these various components, sort of hub and spoke being really, really important. Where does the ability to automatically, you know, ocr, r ai, extraction of data, automatic classification of documents is ready to go or not, all of that sort of package of things that you could do with a computer vision solution or an ML solution in terms of what’s hard and what’s valuable for this industry, where does that type of capability fall, assuming it exists, and assuming it works well?
AB: Yeah, so we talked a lot about the intake process. I think in the long run, that is the place, right? So I’m gonna take that submission. I’m gonna strip, I’m gonna read it electronically, and I’m going to figure out which systems of the three that you sort of described. Is it a workflow issue? Is it a policy, administration issue, is an underwriting issue? It will look at those and say, these, these, these are the, the data points for each of those activities, right? So I’m gonna need to rate this. So my list of data needed to rate ends up needing to go into my policy administration system so I can rate it. If I need to learn more about the nature of the operations, maybe that goes into a third party data call. And that ends up in the underwriting system so that when the underwriter looks at it, they have a good understanding of the nature of that particular risk and can move forward effectively.
And your business rules logic can kick in behind the scenes to sort of push alerts to the underwriter. Mm-Hmm. <affirmative>, you know, the workflow component is like, is this, is this an exception? Does this need early attention? Do I need to go go, well, that might need to go to a different person, and where is the data missing and how does that feed a BPM tool that might send it over to another party who would round out the data and then move it back into the process? So those, those pieces. So I say all of that is really targeted, really at the upfront when I get that submission in the door.
CW: Interesting. And, and would you say, again, outside of the sort of integration being the, the, the tough part today, would you say that that, you know, sort of speeding up, perfecting, tweaking that intake part, is that a higher priority than say, data integrity from third party systems or better underwriting, you know, rate engines?
AB: Yeah, and it sort of depends on the pain points that you have within the industry. Like, if you ha within your company, if you have a lot of operational challenges, then that becomes your driver, right? I can’t get things rated and issued very effectively. Well then maybe I need to spend a lot more time on that. And, and, and using these tools to actually accelerate that, to get things rated faster might be something that you really focus on. If I’m more, if I’ve sort of got that one solved, and really what I’m trying to do is sort of accelerate the decision velocity, then maybe what I’m doing is I’m taking that information in and I’m feeding it into a module that will go call data sources and bring that information back for the underwriter. And maybe that’s where I’m gonna spend all my time in this process because my operations are pretty tight.
You know, you, you sort of have to look at what is the company perspective on this, because I think there are a number of angles on that technology. But yeah, getting a fast answer back to your agent, and you have to turn this back around to look outside in and say, well, what is, what is your customer’s? Yeah. Biggest concern, right? It’s taking forever to get a quote or the accuracy of the quote isn’t good, or the pricing isn’t matching, or the quality isn’t there in terms of what policy documents they receive. I mean, there’s a lot of, there’s a lot of cuts on this.
CW: Fascinating. The insurance is more interesting every time I look at it. <Laugh>. Well at this point we should probably wrap it up. My guest today has been Art Borden, who is a vet in the underwriting process and technology space art, thanks for you know, just really laying out your expertise for us today. It was a great conversation.
AB: Great. Thanks so much, Chris. Thanks for your time.
CW: Yeah, absolutely. Take care.
Check out the full Unstructured Unlocked podcast on your favorite platform, including:
