Watch Christopher M. Wells, Ph. D., Indico VP of Research and Development, and Michelle Gouveia, VP at Sandbox Insurtech Ventures, in episode 32 of Unstructured Unlocked with James Wright, Head of Technology at Beazley Digital.
Michelle Gouveia: Hey everybody. Welcome to a new episode of Unstructured Unlocked, Michelle Gouveia
Christopher Wells: Co-host Chris Wells,
MG: And we are joined today by James Wright, head of Technology, Beasley Digital. James, welcome to the podcast today.
James Wright: Thanks guys. Thanks for having me.
MG: Very excited to chat with you today. So James, I’ve known you for a while, but could you for the benefit of Chris in the audience, explain what the head of technology for Beasley Digital does?
JW: Sure. So hi everyone. James Wright. I’ve been at BZ about 19 years, long time, been in various roles and I’ve split some of that time across the US and the uk. So I led the North American US team and then really about three years ago came back to the uk. What I look after is all about small commercial insurance, predominantly specialty risks. So stuff like cyber, DNO management, liability. We’ve got some bizarre smaller lines, so pleasure craft in yacht business, but essentially everything in the small commercial space. That’s my role. I sit on the leadership team and we’re quite unique, I suppose in many ways digital was formed almost in the model of a smaller technology company. So we’re quite, the division’s got technology operations and underwriting in one team. So I sit on that leadership team that oversee those areas of our business.
CW: Sounds like you have your fingers in a lot of pies. What’s the day-to-Day like?
JW: Day-to-Day for us is a lot of my time is spent on trying to figure out how we digitize those products in those countries. So we operate in uk, France, Germany, Spain, and Canada, about 15 products and a lot of that time that I spend a lot of my time figuring out how do we automate the underwriting of those products. The second part of it is how do we connect those products to distribution? So how do we get our brokers to access those products in a very efficient and digital way? So yeah, most of my time is spent figuring out those problems.
MG: And James, the way you described the org structure and all the different roles that you have as part of the digital team, obviously working in specialty lines of insurance, when you talk about automating or digitizing the underwriting process, you’re looking to automate or digitize processes that have different inputs, different data requirements, different submission intake processes, probably coming from different distribution partners and things like that. Is part of the way that you’re all structured to be able to have all of those capabilities so that when there is a technological difference in where data’s coming from or an operational difference in how that data needs to be managed, it can get done faster. Just talk to me a little bit about how that org structure benefits the nature of such specialty lines of insurance.
JW: No, absolutely. So when we started this team, Beasley had digital trading across its product lines in different areas, different parts of the world. Some were working really well and some were not working so well. But what we realized was that from a broker’s perspective, that was really a bit schizophrenic. It’s a bit scatter gun. It wasn’t a consistent digital interface to our business. The second part of that conversation was when you go and visit a broker or talk to a customer, they don’t only want to talk about the coverage the product, but they want to talk about how can we access that and make it work efficiently. For us, brokers and carriers are obviously trying to be as efficient as possible and customers just want to get coverage quickly. So you need to bring to the table not just underwriting skills, but those operational technology skills at the same time to solve those problems. So that was the two kind of sides of the coin that we were trying to fix with the formation of the division.
The last part of that is owning the p and l. I think a lot of carriers have got a lot of technology initiatives going on across the teams and to really figure out are they returning value, are they really making money? Is quite a hard thing to do. So we’ve put ourselves on the line here a little bit. We’ve carved out the p and l, we’ve carved out the skills and it is very evident now to us where is digital trading working and where it isn’t and where it isn’t? What should we do to fix those problems? I think ultimately it’s well received by our brokers. When we show up, we’re showing up holistically. We’re like, here’s our products, how can we solve your problems? A big part of the strategy for us is being organized around our brokers. So meeting our brokers where they want to trade. Go ahead Chris.
CW: Yeah, I was just going to ask, it sounds like you must have not just a technology mandate, but a product mandate and I assume a product organization that falls under you. Is that right
JW: In terms of the mindset and the way we structure the division
CW: To
JW: Build what we’re building?
CW: Yeah, that’s interesting. You don’t need me to say this, but that sounds like the movement and it sounds like it’s working.
JW: I think obviously in an insurance company you’ve got the product, which is the insurance product, but what we have done is we’ve organized the team by the delivery channels and in that we’ve got product owners and product managers and they think about what they’re building as a product. So an example of that would be our API team. I think in most carriers, an API team has probably seen as a technical back office group of people that integrate stuff. Our API team talk to brokers and they think about the APIs as a product, and that’s definitely something we’re still working on, Chris, but that mindset’s really important.
CW: Yeah, I love it. Product’s hard. I have great respect for product people. I could not do the job not well.
JW: It’s really hard because we have to find people that are curious enough to learn both sides of the coin. They need to understand the insurance market, the insurance products, and in equal proportion they need to understand what’s feasible with technology, how to solve problems, and how to organize that in a way that creates value. That’s a hard job,
CW: And users don’t always know what they actually want. They will just lie to you unintentionally or intentionally. It happens.
JW: So we’re getting better at doing things like MVPs value testing. We’re doing some work at the moment on an ESG service offering, and we didn’t really have any confidence of what we should build, what customers want. So we put together a piece of work that’s been very iterative. We’ve had people out there talking to brokers and clients, testing painted cardboard thickness screens, iterating that into an MVP and just next month actually we’ll be launching something for real life usage. But that’s not because we’ve got me saying this is a great idea. It’s us saying there’s a market need, we don’t how to solve it and let’s test and iterate some concepts.
MG: So it’s a little bit like your organization James is kind of like an innovation hub where it’s like you’re testing and learning, testing and learning, then seeing if it’s something that you move into a larger operation. I’ll say in terms rolling something out, what are the gates or how do you measure success? How do you determine what is working, what’s not? And then is that something that gets to the wayside? Do you just keep pivoting until you find something that there’s a market there? How does that work?
JW: Yeah, it’s a great question, right? I think there’s plenty of innovation teams that might create great things but that no one uses, which is a big problem. So we use the KR framework quite heavily. So the way that we’ve done that is that our leadership team, which I’m on, we’ve set five strategic objectives, the team, everyone on the team ops, underwriting technology, understand what they are. They’re then tasked with coming up with their key results that sit under those objectives. Now the objectives that we’ve set are relatively obvious, and for example, minimum touch. That’s one automate things. The other one is organized around our customer. So don’t just build it because you think it’s a great idea, talk to the brokers and figure out how they want to access us. The other one’s access the specialists. So obviously we’re not just a tech company or like a tech MGA, we’ve got some legacy underwriting knowledge and brokers want to access that. So how do we push that to the front of the proposition? So they’re the top level objectives. The cross-functional teams come up with their krs and that really keeps us all aligned. I naively thought this model, by the way, would be quite easy to implement it. It probably took us about 12 months to really embed the OKR model into the division.
Now we’re kind of in year two of operating it. We’re in a relatively good flow, but it is still hard to always think about outcomes over things. Everyone jumps to, I want thing, I want this thing, I’ve got this solution.
CW: Give me an app,
JW: Just give me an app. So we often say, well, what outcome are we trying to achieve and how are we going to measure what the outcome is? So it might be, for example, we want to increase the quotes, find ratio. Now there might be a thing to get there to do that, but let’s talk more about the measure and then we’ll figure out with a really good cross-marketing team how to make that happen.
CW: At what point in this whole process of getting these projects up and running, do you know what the target is for the OKR or is it always more vague? It should go up?
JW: We use baseline metrics, so we typically know where we are. The KRS are typically stretched, so I’m not saying at the end of every quarter we’re high fiving each other. There’s been somewhere we’ve made no progress, but we’ve learned tons. But I think the difference is that in the team that we’ve built, when something like that doesn’t work, it’s not like technology’s failed. It’s like, okay, how do we pivot? What happened? What have we learned from this rather than a fake project plan? Everything’s green until the very last day and it all goes red. We’re kind of not in that world.
MG: I love that you said that you hit my biggest pet peeve from my carrier days of just because you’re moving dates, I’m probably going to get a lot of crap for this. You’re moving dates, things look right. That doesn’t mean that things are going right.
CW: No, but then it’s agile.
MG: Yeah, that’s a whole other tangent episode. I think maybe that’s the B roll we need.
CW: Oh yeah.
JW: I think we’re all, obviously there’s a lot of technologies listening to this and I think we have guilty of this often are quite optimistic about what technology can achieve and sometimes you can go in a little bit hot with what the outcomes be. I think we have to remain optimistic and positive otherwise doing our jobs really hard. But you do need to be realistic about what the outcomes are going to be and how they need to be iterated. We’ve recently taken on some work around submission ingestion,
So dealing with unstructured emails from brokers and trying to extract the data such that we can optimize the underwriting process. Now that’s a completely different problem to building a policy admin system. Building a policy admin system is hard. It is complex, but you know can do it. So we can be kind of confident on that one. But on the first one I mentioned, it is a relatively new and developing area. There’s lots of stuff out of your control, and I think I was probably guilty in this scenario of being the overly optimistic technologist. And this year by looking at the key results and how they’re iterating, we’re starting to become a bit more realistic about the outcomes within technology failed. We’re just saying, look, we’re just not, it’s iterative. We will get there.
MG: And James, given that you team and your roles, both your current and previous ones at Beasley have crossed international borders. So what does that look like in the US versus the countries you’re in Europe in terms of what the submission process is, but then even the data that you’re capturing. So I imagine that you may have to create, you want the same outcome, but you may need to create different rules within the different markets you’re in just based on what’s available to you. Is that fair?
JW: Yeah, Michelle, you’ve hit the nail on my head. I can probably break it into four regions. So the us, the uk, Europe and Canada. I’ll give you some examples. So in the US market we predominantly deal with wholesale brokers. It’s a heavily email driven market. What we’ve seen in the last three years, let’s say it’s been brokers, they’ve got their digital trading strategies and they’re equally pivoting as well. So we’re starting to see the emergence of APIs become a predominant new way of receiving submission data. And that’s continuing. In the UK it is mainly portal driven business. So brokers are quite comfortable with portals in the UK and that seems to be persisting with the emergence of APIs from brokers. Europe is a melding part of different stuff.
There’s some well established broker hubs and broker technologies. There’s a fair demand in Germany, for example, for embedded insurance with banks. That seems to be an emerging trend there and there’s still a lot of email business Canada we’re still figuring out. So we have got a full portfolio of digitized products, but we do know that there’s portal fatigue in Canada and therefore we’re trying to be a bit different and figure out our go-to market strategy in Canada. So watch this space on our Canadian approach. But yeah, different markets, different levels of maturity, but the common theme is probably APIs as carriers and brokers start to get more proficient in operating and those ones, which makes sense, right? We’re just exchanging information with each other. So I’m hopeful that that will become a predominant way of trading in the next few years.
CW: Interesting. Portal fatigue is a new one for me. I hadn’t heard that before. I like that. So are you finding specifically with underwriting and claims, do you find that the challenges in that space are geographically sort of organized or are there common challenges across all geographies?
JW: There’s definitely common challenges. Let’s maybe give you some of the underwriting challenges, just more commercial specifically. We are looking to straight through process as much as we can to deliver that exceptional service. The challenges there are in data gaps and quality. So an example would be we sell a lot of tech e and o insurance.
CW: What’s EO insurance?
JW: Errors and omissions. So it’s where a technology company is their product or their services have caused some harm to their client and cause some degree of loss. To ensure that risk, you need to know what the technology company does. Right now there’s quite a big difference between someone that fixes laptops to someone that’s mining crypto for example. Now that could be hard to figure out. It’s not an obvious thing necessarily. So when that submission comes in, there’s always a question mark over the business activity. So we’re having to think of ways to ingest what the broker’s telling us, but also think about using third party data to confidence score that such that we can progress with confidence today. Even in our team, a lot of that’s done by underwriters. They Google the company, they look at what they’re doing, they’re looking for business activities that might be prohibited and then underwriting it accordingly.
There’s other considerations outside of tech EO, even in standard lines, you may have a restaurant, but it may in some way have for example on the side of it, a dispensary for marijuana. And it may be that we’ve decided that marijuana is not an appetite for us and that would then prohibit us from writing that risk. Now that’s not a problem for us, but it’s a simple example. You can imagine it’s industry class, it’s restaurant, but it does have something on the side that’s prohibited. So that’s why you need to get into a bit of the detail on the underwriting side and they’re the data gap that make automating this challenging.
CW: Interesting. Yeah, we’ve heard this several times over the course of the last few months that data gaps are a big deal and Michelle’s talked about some of the InsureTechs out there that are trying to bridge those gaps. So that’s glad you brought that one up. Would you call that the biggest issue or is that the one that came to mind first?
JW: I think that’s one of the data ones. The second one I mentioned already, which is for us the digital trading strategies of our brokers are at different levels of maturity. They’re pivoting, they’re unpredictable, so that’s super hard. We spend a lot more time now trying to keep in lockstep with our broken partners so we know where they are. That would be the second I guess. And I think the third one is as we get more connected and we’re seeing this cyber’s, definitely the canary in the cage in the US on this one, the cyber markets become digital very quickly. So it’s almost been the thing that’s uberized the market to some extent. It’s forcing us to be more connected. It’s forcing us to do things like vulnerability scanning and bringing that data into the underwriting process in a digital way, and that’s making the product move quicker. It’s making the rates move quicker and that’s been a big disruptor for us. That’s definitely something that’s difficult and hard right now for us.
MG: And James, when you are talking about underwriting talent in these specialty lines, and we’re talking about automating a process obviously, is there still how much, even if it is automated, how much underwriter touch on the backend for verification or validation do you anticipate there will be even as that process matures? Just given the complexity of the specialty lines?
JW: I think it’ll vary by product. Something like cyber in the sub hundred million dollars revenue range I think will be quite low touch. And so just so you know, we use kind of litmus testers if we think we can get to a referral rate of around 20% referred to underwriter, that’s really a digital product. So that’s 80% being straight through processed. And the reason I think cyber can be done is that we can learn quite a lot about the client by using third party data. So cyber vulnerability scans and other metrics that we can collect. Other products, it may be more challenging because the data’s just not, especially for small companies where they’re not public, it can be quite challenging to get the information you need to underwrite it. And I think in those scenarios you’ll see a referral rate probably a straight through processing rate, more around 40 to 50%. What we do is we look at our referral rates by product and we deploy resources to optimize them where we can. So we look at the question sets, we look at using third party data, we look at our own underwriting methods and see if we can optimize that process.
CW: Interesting. And are you building tools additionally for the underwriters themselves to help them bridge some of these gaps that exist or is it sort of at that point a handoff and they have their own toolkit?
JW: So we have a product engine and I would say probably the biggest thing we do to assist the underwriter, it’s not providing with a flashy workbench, it’s actually in trying to help them codify their knowledge so they’re not at the keyboard 24 7 processing risk manually. So we spend a lot of time sitting with the underwriters trying to understand what are the rules that are going through your brain when you’re selecting and pricing that risk. And then we try and put those rules into the product engine so we can increase the amount of risk that we can automate and there’s always going to be edge cases and rather than building for the edge case, we just say that’s always going to be a referral. It’s just too difficult, too complex, always refer. We look for those kind of sweet spots, we look for those industry classes and risks where we go. We could definitely automate that one and these things change as our claims experience develops, we can choose to flip it, run the other way, we can change something to referral and that’s just good underwriting, right? We’re obviously trying to protect that loss ratio as well.
MG: James, given the answer that you just gave, which totally agree with where do you see AI solutions being the most impactful in that process?
JW: It’s a good question. I think we’re all still figuring that out. If I see another demo, someone doing some text extraction on a large language model, I’m literally going to shoot my knees off. It is clearly going to be very impactful. I think it would be naive to say it won’t. The challenge we’ve got is how do we embed that type of technology into some processes that are not yet fully digitized? And I think most companies are still in the midst of their digital transformations. So it’s trying to figure out how you embed that level of intelligence into a workflow. One example that we are actually doing and not just doing city demos on easy in the submission world, the text extraction world. We’ve been since March this year using one method of extracting data from submissions, which has been quite heavy on the learning, quite heavy on the config. And about a month ago we started testing a large language model to help extract the information with greater confidence and that’s providing what we think is going to be about 20% uplift in over and above an older model. So that’s a great example of us starting to put that process into production. We’re not in production with that yet, but we will be by the end of the year.
CW: It’s so exciting and refreshing to talk to someone who’s thinking reasonably about large language models and insurance over the course of the past this year really talked to so many people in technology organizations, in insurance companies and the use cases they’re talking about the hype that they’ve bought into. It’s like to your point, you haven’t finished your digital transformation with technologies that already exist and are well understood, like focus, backup.
JW: I mean it’s clearly going to be really impactful. I think there’s space in the market for maybe some specialized LLMs in insurance. There’s lots of language, there’s lots of terminology that’s nuanced and eg a limit in insurance means something to a limit in a casino. That’s a simple example, but I can see there may be space for a specialized open source language model for insurance potentially at some point. But yeah, tons of opportunity in that space. We’re just at the start of that journey. I think
CW: If I could figure out a way to get insurance companies to pool their data so that I could fine tune such a thing, I would be very excited. But I’m skeptical about that.
JW: A lot of us have got lots of days challenges in the middles back office.
CW: Yeah, fair enough.
JW: It’s a consequence of legacy and it is also a consequence of the cost of data quality. When you think about how much information there is in a submission or an application form, historically the cost of getting that structured just doesn’t economically make sense. So we haven’t got it. The economics are changing now. AI is going to help us probably reduce the cost of getting that data by a lot. So again, that’s an example of I’m hopeful, but we’re still not there.
MG: Actually, Dave, that brings up an interest. Well the question for me that, because I find what you just said really interesting is in standing up or being responsible for digital first products and distribution, but knowing that the broader organization is still tied to very legacy systems and data structure, how do you reconcile those two? Are the processes that you’re building, the automation you’re trying to implement, are you building things on top of those systems that from a go forward basis now you have a better infrastructure or are you still having to do all of this with the constraints of legacy debt underpinning what you’re building?
JW: Yeah, so B is going through a digital transformation at group level and then we’ve got our division as well. We work in partnership to a large extent what we’ve done, we’ve carved out enough autonomy in our division to control the bits of the process we need to control.
So that’s predominantly the product engine, the product rules, wordings rating, the distribution piece, how we connect to our brokers and the processes for the underwriters, the group components we back into via API. So we built that kind of layer. So finance claims, reassurance, exposure management, data warehousing, they’re all equally obesity going through their own journeys, but we can abstract ourselves away from that to a large extent in those data contracts. So I suppose the organizational model, we’ve carve that just enough to be as agile as we can yet still support the group transformation. That was a hard journey to go through from an organizational perspective. So I’m not downplaying the effort, but anybody listening, I highly recommend if you’re trying to do something and move quickly, convincing your organization to carve out not just technology and ops but all the people you need to deliver an outcome. But just enough. We don’t want to duplicate those other bits as well. It doesn’t make any sense for us to duplicate those big back office team. They’re very expensive and complex. So that’s kind of how,
CW: Yeah, it’s the old saying, right? You ship your org chart and it sounds like you’ve got the org chart, right? So you can ship the software that mirrors it. I think that’s great.
JW: Yep.
CW: Coming back to the AI question, obviously changing, adding this type of technology, which is not perfectly predictable the way ordinary Java code or something would be, it produces some challenges, probably some risks. So what have you seen at Beasley and what do you think the market’s seeing
JW: In terms of the risk side of it?
CW: Yeah,
JW: So I guess I’ve got two answers to that question. There’s the internal answer, which is we have a formal AI steering group now that’s specifically looking at the internal risks of this to ensure that our staff are not sending all the data to open AI and things like that. So we’ve got that control. And then on the product side, we have a really good product innovation team that sits in our corporate development function and they’re starting to think through what are our client’s risks with this and how could we potentially help them transfer some of that risk in an insurance vehicle? To us, that’s still early days, but we’ve done this before. I think examples are probably on the crypto world. We’ve got a couple of innovative products for crypto. It took us a while to figure out what we can and can’t s show in that space, but we’ve got a couple of new products in crypto product innovation team worked on a really interesting fertility insurance product which ensured outcomes of anyone going through IVF because it’s a very expensive process and again, that’s something that we’ve developed an insurance product for. So I think we will come up with some insurance products for the AI world, but we’re all still learning, right?
CW: Yeah. I got to step away so I can go trademark prompt insurance. I’ll be back.
JW: Yeah, prompt insurance, prompt engineer insurance.
CW: Yeah, exactly.
JW: I mean that will be covered under a sort of techie and I policy, but clearly to your point Chris way that we think about technology and o is that there are people coding every line of code and that’s clearly not the case anymore. Most tech companies are already using chat GT to create SQL scripts and that code and create Python code and that’s happening today.
CW: It used to be that every time I was trying to code something slightly novel, straight to Stack overflow and nope, I know. So when you think about implementing sort of an AI powered technology for your brokers as a part of a product, what do you see? Are there different challenges with that type of product in adoption or use or is it really just the typical, it’s a new tool and I have to get used to it kind of thing?
JW: No, I think there are challenges. So the first one is regulatory.
CW: Okay.
JW: We have to be able to evidence why a system has taken certain or actions and especially as you guys know in the US admitted market, that’s particularly true. So you’ve really got to think through your regulation. So we don’t use AI in any of our underwriting at the moment for that reason. I think the other hurdle is taking the underwriting teams on that journey of trust. The people change side of this is quite huge. Everyone knows in tech getting adoption is hard, but in this one you’ve got to take your underwriting team or your claims team on that journey with you. So we’re starting to do a little bit of that. We’ve learned that this year actually, that’s something we’ve got to do more of.
CW: I think most of the people listening to this would love to know what some of the stages along that journey are because everyone has to find them and they’re hard to uncover.
JW: So I think we made a mistake this year in that with regards to automatically trying to get to auto quoting for SME business, we kind of felt like that’s a predominantly an operations benefit and if we can get the data confidence high enough, we can get to an auto quote outcome. I think what we learned was that we needed to put a human in the loop in that process, but for that human to be an underwriter, because we need their subject matter expertise, we need their experience knowledge in looking at what’s coming through and providing the correction on it. The second part is if it’s ever for the, we want those underwriters to clearly become more portfolio based. They need to trust what’s coming into their portfolios. So we pivoted, we’re engaging them more in that end-to-end process. There is a cost to doing that because they’re the team that are producing business and we’re getting them more involved in training some machines, but it’s not just training the machine, it’s really them training the machine and them gaining trust in that new way of working. So that’s been our learning this year really.
CW: It stands out to me that you didn’t mention anything about them being afraid for their employment. No,
JW: I don’t think that’s the case. I think most people at the moment are, look, our team working so hard and there’s so much opportunity to grow. We’re not going to go out and hire lots of new underwriters can’t, they don’t exist. They’re not in the market, they’re working somewhere else. But the only way that we can grab portfolio is through increasing automation. So I think our underwriting team are pretty excited about getting the technology mature enough to take on more of the simple risks. We’re not there yet. We’re still in that process, but they’re excited about that.
CW: Go ahead Michelle.
MG: No, I was going to point out that I think it’s not funny haha, but just funny, interesting that you say the simpler risks coming through, but when I think about specialty lines, I understand some of the underwriting may be simpler, but I just always think it’s complex. I know that it’s not your standard list of questions for submission and I always imagine that there’s a lot of back and forth going on. So what I think is probably to your point James, is that the Androids are excited to get away from some of the very manual not value add part of it so that they can get to the interesting part of really dissecting what is this risk, how does it hit up against our rules? And just like what’s unique about it.
JW: No, you’re exactly right. Without giving away the secret source within specialty, it is all complex, but there are some patterns of risk that we perceive as being lower risk and therefore requiring less confidence in the data. Does that make sense?
MG: Yeah, that’s a good distinction. Yeah.
JW: I’ll give you an example. If you’re putting, let’s say a hairdresser into a certain hazard class, it might be that if it’s in has class two or three, it makes almost no difference to the premium. And let’s not get too obsessed about that data being of a super high quality. There’s other risks for other classes where it makes a material difference. So your data confidence needs to be a hundred percent and that makes it harder. So it’s just taking that, there’s a slightly different lens on how you get to automation.
MG: I don’t think we’ve hit on it yet in this conversation, James, but how does your team specifically vet third party support for that? Whether it’s incumbent large data solutions, whether it’s insurtech solutions, be it data technology plays, however, and how do you think about incorporating that into your process? Are you interested in POCs and pilots? Is it not a strategy you guys are looking to execute on? Just curious how that whole ecosystem plays a part in what you all are building.
JW: Yeah, we’ve vet them based on how good their PowerPoint skills are. Just joking. It’s not a recommendation. This is going to be a really boring answer, so I’m really sorry. But literally just using sample data sets, an would be NS code mappings. We’ve got loads of data internally because we’ve manually underwritten these classes and then we’ve got a load of new submissions, which we won’t have completed the NS codes on, and you can just do some simple sampling and testing and that’s how we typically get to understanding how good the data is. We typically see hit rates like data quality from third party providers, like most of them are in the mid to high eighties when you finish the analysis and then it’s just a debate, is that good enough or not, right? Is that 14%? That’s not accurate, a major issue.
MG: So you all are not, this is probably just a very broad general statement, but it sounds like you’re not looking for anyone to help you build the pipes to get things in the workflow done. You’re looking more for areas of impact within your existing workflow where third parties can be more value add than what you’ve got. Is that fair statement to make?
JW: Yes. Yeah. And we’re all about using partners. I’m definitely a buy versus build guy on this topic. There’s lots of people in the world with investment that are trying to solve difficult problems and there’s so many that are so amazingly bright and brilliant, so why not partner with them, bring them into the ecosystem, partner with us in that way.
CW: Amazing. We’re coming up on about 10 minutes left, so I want to give you a chance to tie a bow on all of the wisdom that you’ve shared with us so far. And then I want to ask where you see automation and underwriting and claims going in the next three to five years. So to tie a bow on this for your sort of counterparts at other organizations who are listening to this, what’s your advice for those who are struggling to automate these processes?
JW: Think about outcomes over things will be point number one, drive that conversation pretty hard. What is the outcome you’re trying to get to here? Because there are a million shiny things out there right now that you can levitate to. We spoke about one of them, right? Large language models, don’t be a magpie looking for shiny things. Go for the outcome conversation. And I think the second point would be build that cross-functional team. This isn’t just a technology problem. You need a committed stakeholder who’s an expert, right? You need some partners, build that team, get the outcome, and then empower that team to deliver some clever ways to solve the problem. That’ll probably be my recommendation to any practitioner.
CW: The title of this episode is going to be Don’t Be a Magpie, by the way that Awesome. Alright. So where are we headed in the next three to five years, do you think?
JW: I can talk mainly about small commercial markets. I think what I’m seeing in all those regions, right? us, Canada, France, Germany, Spain, the uk, things are becoming increasingly connected between the carrier, the broker, and the client. That’s happening at different rates in different countries, and I think that’s going to lead to faster product change, whether that be pricing, appetite or even the services that we add to the insurance products. That’s all going to accelerate as these markets become more connected. Whether that’s going to happen in the next three years or the next 10 years. I’m not going to try and I’m definitely not a crystal ball expert on this one, but that connectivity leading to faster cycles, faster innovation is definitely happening in all regions.
CW: Amazing. Good. Any other words of wisdom before I wrap us up?
JW: Wear green sweaters. Yeah,
CW: Green sweater today. Green sweater. Day on. Unstructured. Unlocked. Yeah. Everyone just listening on audio is like, what is James talking about?
JW: I didn’t get the memo for a green sweater, so I feel like an outsider. No, I’m joking.
CW: You look dapper in your black polo
MG: Yeah, you’ve got better weather going on. We are obviously already dressed for the cold that’s hitting the northeast.
JW: It’s true,
CW: It’s gloomy. The
JW: Only other word of wisdom to close on is also whatever role you’ve got in the organization, get as close to the cash register as you can really understand the customer. I’ve really encouraged my team to do that. And I think if you’re looking for that outcome, if you’re looking for really understand the problem, it’s not always easy. But definitely ask those questions, be curious and try and understand the client as much as you can.
CW: And with that, this has been another episode of Unstructured Unlocked. We’ve been talking to James Wright, head of technology at Beasley Digital, and he’s been sharing a lot of really good insights. I’ve been your co-host, Chris Wells.
MG: I’m co-host Michelle.
CW: Thanks James. This has been great.
JW: Thanks Chris. Thanks, Michelle. Pleasure.
CW: Absolutely.
Check out the full Unstructured Unlocked podcast on your favorite platform, including: