Episode 9-11 Transcripts
Laura Drabik Welcome to InsurTalk. My name is Laura Drabik and I'm the Chief Evangelist at Guidewire. In this episode, I have the privilege of interviewing Rao Tadepalli CIO at Siebels and insurtech mentor. Rao has been a senior leader in insurance technology for over 25 years. Kick-starting his career at great American Insurance. Rao has been a mentor at Plug and Play a Guidewire insurtech incubator for a number of years. And I specifically selected him for today's podcast because of his knowledge of insurtech and how to leverage it to make innovation ideas happen. Hello Rao, thank you for joining my podcast today.
Rao Tadepalli Thank you, Laura. I am really honored to be on the podcast.
Laura Drabik You have been an advisor and a mentor at Plug and Play since 2017. Why did you start mentoring and what is involved in mentoring insurtechs?
Rao Tadepalli Well, Laura, in fact, very interestingly Guidewire was instrumental in getting me started on the insurtech mentorship. I believe in March 2017, I attended the CA was Symposium posted by Guidewire at Palm Beach, Florida sessions included speeches by CEO of Slice Labs and Plug and Play insurtech incubator. I was already thinking about the digital transformation and they insuretech evolution. Point of view, insure tech savvy, the insurance, our technology as their background. Not necessarily both with my experience in both insurance and technology, I decided what better way to influence the insurtechs, and make sure they are working on solutions that matter to the insurance. Then joining them and educating them on these insurtechs specific challenges. I talked to Plug and Play and became a mentor, and I worked with multiple insurtechs and provide guidance and recommend areas to improve in their pitch deck, to investors and customers and help them validate their value proposition.
Laura Drabik I'm glad you enjoyed my panel in Palm beach. I've continued having insurtech panels as well as interviewing several insurtechs on my podcast. And when I asked them what is critical for success, they all highlight starting with the business problem. What else is critical for an insurtech value proposition to succeed?
Rao Tadepalli First and foremost, clear message of the value proposition is extremely critical. Most of the time the carriers are looking at three things, either increase the premium or decrease the losses or decrease the expenses. Of course of late in addition to those three, there's a fourth element, which is really delightful customer experiences. Obviously this will enable increasing the renewal retentions. So making sure there is a clear value proposition and at least the value proposition should focus on one of these four things. I just mentioned, if you can do all four, fantastic, but at least one of them should be there. And then assembling a diversified leadership team is extremely critical. That should be multi-generational and the team should have challengers and enablers with networking connections to the industries. And of course, sales and marketing.
These are extremely critical and also don't underestimate the capital needed. Lot of insurtechs underestimate the length of the sales cycle and not allocating enough money for sales and marketing and the social media promotion. The other thing is one needs to integrate with the core systems and build accelerators to core systems with vendors like Guidewire. And also one has to understand emotion and not always the reason that drives most decision-making. So try to use, I call it two EC's, easy stands for Economic Connection and Emotional Connection. These are extremely important. If you don't have a story, someone else will make up a story and that could be your competitors.
Laura Drabik Really exceptional advice in particular, I appreciate the building and integration to a core system like Guidewire because insurtechs aren't systems of record. We are. So for insurtechs that aren't partnering with insurers, they have the added challenge of understanding our industry. And as you know, the insurance life cycle, as well as our regulations can be a little tricky, can stand alone, insurtechs thrive without insurance partners. And if they can how?
Rao Tadepalli Absolutely Laura, they can try. They got a few things. First of all, I recommend they go to bring in an advisor to the leadership team or the board with extensive insurance expertise and industry connections. In addition, they can work with incubators like Plug and Play or other innovation centers like Global Insurance Accelerator, and also partner with the insurance software vendors like Guidewire and develop accelerators to facilitate the speed of implementation to their solutions. And by integrating to the core in short Excel, to realize that they cannot implement these as a standalone solutions, they have to be part of the ecosystem that they ensure us how. So APIs are critical to the ecosystem. And one has to kind of make sure the insuretech solutions are using open architectures to facilitate its ability to integrate. And honestly, I believe that one of the success factors for Guidewire itself is its open architecture and be able to easily integrate with partners. So I highly recommend the insurtechs make sure the solutions are open and be able to integrate with other partners solutions.
Laura Drabik Before we continue listeners, if you're enjoying this podcast, be sure to subscribe to insured, talk on Apple podcasts, Stitcher, or wherever you get your podcasts. Now let's get back Rao where we're talking about innovating with insurtech. So according to McKinsey, due to the pandemic, we have vaulted five years forward in digital adoption. One of the digital insurtech value props that has dramatically increased in adoption with our customers has been imaged based remote inspections to support both the underwriting and the claims process. What other value propositions are you seeing increase in adoption due to the pandemic?
Rao Tadepalli Well, Laura, definitely the pandemic changed the whole landscape and accelerated so much of the digital transformation efforts in the industry, digital payment solutions for both premium and also claims payments have been very popular and they have seen an increase in adoption due to the pandemic and companies like a Warni and InsurePay have been really busy in trying to implement some of these digital payment solutions for various carriers. I also see increased adoption of big data usage to collect underwriting information as opposed to agents going to prospect to ensure to get answers to the underwriting related questions. Wearables and I will be devices are also popular to track the moment and identify the location. Overall, anything that facilitates a remote operation have been in demand since the pandemic.
Laura Drabik That makes sense. So my favorite insurtech value proposition is conversational AI. It allows carriers to deliver personalized inspection intake and policy servicing. So what is your favorite insurtech value prop and how does it better serve the consumer?
Rao Tadepalli Well, Laura, I have quite a few favorite one. It's kind of hard to select some of them. However, I must say the smart homes that prevent the property damage using the insurtech technologies is one of my favorites and also the apps that collect miles driven and the driving behavior. So the customers can pay only for the actual usage that us paying for an estimated usage. I work in devices and sensors bring measurement of usage and transforming the entire industry from a subscription-based business model to a consumption based business model. Also any insurtechs that enhance senior living as there is a great need with so many baby boomers retiring. I believe we have a social responsibility.
Laura Drabik I couldn't agree more, especially considering that in the next nine years, every baby boomer will be over the age of 65. So thanks for drawing attention to that. Based on your experience, what is the most effective way for insurers to work with insurtechs?
Rao Tadepalli Well, insurers should engage in dialogue with insurtechs and partner to do pilots to understand the fit for the insurance business problems and evaluate the value proposition. One should ask questions like, is it scalable? Is the timing right? And also if we move forward, will we be good at it? Do we have the expertise in the appropriate technologies? Can we explore the opportunity for the long term or really be a short term advantage? Finally, digital transformation is not about building technology. It's about transforming the business. So these are some things that need to be kept in mind for both insurtechs and the insurer.
Laura Drabik We need to take another break, just a reminder. If you're enjoying this podcast, be sure to subscribe to insured, talk on Apple podcasts, Stitcher, or wherever you get your podcasts. Rao, what is critical for success when implementing insurtech technology within an insurance?
Rao Tadepalli Just like any systems implementation first and foremost is set the expectations upfront. Also create the character environment and promote a culture that fosters, faster decision making and also be flexible. Do not enforce lot of rigid policies and procedures. And see if some of the bureaucratic hurdles can be reborn and track the progress at various milestones, be compatible with the gray because not everything is black and white. Sometimes things may be gray. Document the lessons learned and do not be afraid to pivot after each milestone perform a critical reflection. What and what areas to improve starting an innovation group within the company. Remember that the irony may be that the structure policies and procedures that are put in place and help the company prosper and succeed may be coming in the way of insurtechs adoption and innovation.
Laura Drabik You mentioned something really interesting and I think really important, which is you should start an innovation group within the company. Rao do you need external people? Or can you work with internal people in the innovation group? Or is it a mixture of both?
Rao Tadepalli I would say it should be a mixture of both. The reason for that is you got to have people within the organization who are familiar with not only the business problems, but also you have to be aware of what assets are magician has, whether it's technology assets or whatever it is. So when you bring other people into the innovation lab, post-fall, you need to make sure that the team comprises challengers and enablers, and most of the carriers will have executors inside. So you might want to bring them, but then you might need to go outside for road challengers and people who can come up with new ideas and be able to question the status quo.
And also with all the new technologies that the innovation might need, whether it's so blockchain, IOT or any of these AI and ML, if you don't have the expertise, then you might want to bring. But also you need to remove some of the obstacles that could come in the way, sometimes bureaucratic things that are put in place to protect the organization, but you need to protect now what I call newborns, just to make sure that they flourish and become very successful solutions.
Laura Drabik From an insurer's perspective. What do you appreciate about working with insurtechs?
Rao Tadepalli The positive energy is contagious, and the creativity and the passion. They have a willingness to learn and adopt, and most importantly, speed of delivery. A lot of times within the large organizations, the speed is not as fast as we all would allow to have. Insurtechs I'm always impressed with the speed of delivery. And also... If you want to go to the moon, you can't fight the gravity. You got to learn to mitigate the gravity. What I learned is quite a few of the insurtechs look from outside, looking in and anticipate the customer needs not just to respond to the customer needs. And those are the things I really loved working with insurtechs.
Laura Drabik The flip side. What are some of the challenges of working with insurtechs that carriers need to be prepared for?
Rao Tadepalli Well, some of the challenges are limited documentation. A lot of carriers are used to having really nice documentation when the vendors bring some products to them, but that's not always the case with the insurtechs and also lack of limited formal structure among the teams in development versus maintenance and QA and things like that because quite a few insurtech team members wear multiple hats and the governments and security sometimes are concerned with some insurtechs. So those are some kinds of challenges insurers face when they try to bring some insurtechs into the organization. But as long as both teams are flexible, I think there's a great opportunity to collaborate and partner. And I think in my opinion, it's a win-win situation.
Laura Drabik Thank you very much for your time today and for your incredible insight into innovating with insurtech, you showed us, it's not just about ideas. It's about making ideas happen.
Rao Tadepalli Thank you, Laura.
Laura Drabik Welcome to InsurTalk. My name is Laura Drabik and I'm the chief evangelist at Guidewire. In this episode I have the privilege of interviewing Jeroen Morrenhof CEO and co-founder of FRISS. A company which specializes in fraud detection software for underwriting and claims. I specifically selected Jeroen for today's podcast because of his knowledge of fraud and how early detection and prevention can transform insurers into fraud-fighting superheroes. Hello Jeroen. Thank you for joining my podcast today.
Jeroen Morrenhof Hello, Laura. And I'm very much looking forward to it.
Laura Drabik $40 billion per year, that's what insurance fraud costs our industry and these costs are passed on to the policy holder, in the form of increased premiums resulting in the average family like mine, paying anywhere from 400 to $700 per year to cover fraud. That's what gets me fired up about insurance fraud, Jeroen, what gets you fired up about fraud and why did you start a company focused on battling it?
Jeroen Morrenhof Well, Laura, the same thing gets me fired up. We're in a situation where 90 plus percent of the people are completely honest, but those who aren't drive up the cost as you mentioned. And, I don't think it's fair that carriers have to consider this as a cost of doing business and I especially don't think it's fair that honest people like you and me have to pay extra for it, and that is exactly why I founded the company now almost 15 years ago. Not only because I was upset, but also because I know there was something we could do about it. And, only can they catch the bad guys right away, but it can also provide such a better experience for the fast majority of their customers who are relying on the coverage in time of needs. Last year across our customers we have been able to save more than $1 billion for the industry.
Laura Drabik So, how do you leverage AI to detect fraud, how does your solution supplement and support the adjuster?
Jeroen Morrenhof Well, a remark you made in one of our conversations stuck with me. You said FRISS helps to make the adjuster brilliant in the moment. And, I think using our solution is like having all of your most experienced experts, looking at every single claim in a consistent fair and unbiased way and give all their insights to you in that split second. And, that's also why we opted for a white box or explainable AI framework, allowing the user to trust the AI and give them all the information to make the smartest decision on the spot. And, we do this by running structured and unstructured data from our customers and combining that with all types of external data sources through our self-learning AI fraud models. And, these models compare the claim information to known and evolving fraud patterns and anything is flagged immediately and is available for the claim adjuster.
The other major thing we're doing is building complex networks. So, we can see any suspicious connections between people, assets, medical providers, third parties. We Are the only vendor in this space that's able to do this in real time. So, you can catch every suspicious element right away. And, finally our clients give us feedback and that feedback is used to immediately retrain these models. So, the AI models are literally getting smarter every single day. And, the beauty for the adjuster is that they can do all that without leaving the environment they know for example Guidewire. Looking at the FRISS score, determining the next best action, taking that action and giving feedback and better yet most of this can be done automatically.
Laura Drabik So, FRISS is an international company. How do views about insurance fraud differ across countries?
Jeroen Morrenhof Fraud is accepted more in some countries I noticed. Sometimes we have only 5% fraud and sometimes even up to 20% of claims of fraudulent. And, I think to commit fraud you have to be able to rationalize doing so. And, in some countries it's still seen as acceptable in a society to commit fraud, but there are also countries or States with very specific fraud schemes. These are often unwillingly introduced based on legislation like the no fault PIP States that we have here in the US or the ABI claims in Canada or schemes like ghost broking that are only relevant in a few countries or States.
And, last week I spoke with a customer who have actual proven fraud cases due to threats against their company or employees. And, it's really a moments like that that reminds me that this is not a victimless crime. And finally, another difference that we see across countries is the willingness to share data. Obviously we're all concerned about data privacy and I think that's really a good thing. However, it's very easy still to many countries for someone to commit fraud against one insurance carrier and then moved to the next and do it all over again.
Laura Drabik In McKinsey’s insurance fraud report. They know that fraud is not the focus of top management. They purport that many insurers view fraud management more as a specialist than a top management topic. This leads to higher tolerance in the event of a claim. What is your point of view here and how does your solution affect change in the perception of fraud?
Jeroen Morrenhof Well, this definitely is a concern that managers of SIU departments have been struggling with for years. I do believe that's slowly changing and when I look at reports from for instance, I learnt about topics on the agenda today, these are digitalization, automation and of course AI. While SIU departments care deeply about the detecting and preventing fraud. My advice to the managers of these departments is not to only focus on finding the bad guys, but to position their efforts towards improving the customer experience for the good customers and supporting those digitalization and automation efforts.
More and more insurers understand that it cannot automate their processes, without knowing that this claim or application is actually legit. You need to have some form of verification to ensure someone is not lying to you through the internet and hiding behind their screen. One of our joint customers EMC is a great example of a carrier who's extremely innovative to the point that it's a company value and that extends to fighting fraud, to keep their customers happy and treat them fairly. And, this is what I would like to see more of.
Laura Drabik Before we continue listeners. If you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple podcasts, Stitcher, or wherever you get your podcasts. And, you can rate and review this show on Apple podcasts. It helps others learn about and discover the show. Now this is Laura Drabik and let's get back to our conversation. I'm talking with the chief executive officer at FRISS and we're talking about fraud. Jeroen, what is an insurance company's biggest obstacle to detecting fraud and how do you help them resolve this issue?
Jeroen Morrenhof We recently finished surveying the industry with our 2020 fraud survey and the top obstacle carrier set they were facing was limited IT resources. And, another main thing we always hear is about data availability and the quality of the data. However, we've found that companies are underestimating what they have and are not using it to its full potential. We're also sometimes see that a carrier has developed an AI model in the lab setting and that this model shows a fantastic area under the curve, but it never gets used into production, or it gets used in a batch process and it doesn't get looked at until after the claim has been settled.
Integrating the fraud score and the core processes at the right time and in real time is key. And, you asked how we help resolve these issues, and that's why exactly why we love working with Guidewire. We are able to take hidden or silo data and connect it together. We are able to incorporate models of any carrier's data science team, and we are able to do all of that instantly. And, as soon as the information comes into policy or claim center, we are able to provide these insights right back. The implementation time and the use of IT resources are greatly reduced and the benefits are coming in much quicker.
Laura Drabik The global insurance fraud detection market size is expected to grow at a compound annual growth rate of almost 26% until 2024. What does this growth in CAGR tell us about the growth rate of insurance fraud?
Jeroen Morrenhof Fraud is never going to go away, but the technology we have to detect it and better yet to prevent it is well-proven now. While nevertheless a CAGR of 26% is steep and it means that the investment in fighting fraud is growing. And, I'm very proud of the industry for stepping up and making strides.
Laura Drabik How has COVID-19 affected insurance fraud and can you share with our listeners an example of a new or unique example of fraud spurred by the pandemic?
Jeroen Morrenhof Well, COVID-19 is one of the worst economic times we've lived through Laura. And, to be honest I don't think we've seen the worst of it yet and specifically when it comes to fraud, the effects will last a long time. This is going to vary quite a bit in different States and different parts of the world. It will vary based on how different governments respond with both restrictions as well as financial health. But think about someone who just lost a job and they may be lucky and have some savings or they may have unemployment benefits, but eventually the bills come due, the rent or mortgage is due, the car payment is due and now you may be paying a ton of extra money to keep your health insurance.
And then, in the back of people's mind they think "I've been paying my premiums for this insurance policy for years and what has it gotten me, maybe it never paid a claim." And, they start to become easy to justify ways to recoup that money. They've seen other do it why shouldn't they? So, the mounting financial pressure rationalization that is okay, are two of the free elements needed to commit fraud, which third being opportunity. And, we have seen more fraud the last month for the most part we've seen the same schemes but just more of them. People who no longer can afford their car and maybe don't even need it anymore will stage an accident or theft, or you can actually even hire someone for a few hundred bucks to make your car magically disappear no questions asked, or another typical fraud scheme is the business use of personal Lance vehicles.
In the past we saw this with for instance Uber drivers, but now we see this with kids that want to help out their parents’ restaurant business by delivering food with the family car. Presenting a totally different than actually a really uninsured risk. We are also seeing a huge influx in workers complaints for instance, extended injuries. And, it's easier for people working from home to submit these false workers complaints. We are seeing an increasing amount of fraud and we strongly believe this is just the beginning of it.
Laura Drabik Jeroen thank you for humanizing the rationale or how fraudsters actually perceive insurance fraud. I appreciate that. You advertise 1000 plus insurance years onboard it for us. How important is it to have insurance experience versus deep AI experience, how do you leverage both so you get the best of each into your solution?
Jeroen Morrenhof Well, leveraging both is key. I believe AI is very powerful, but just to shout those two letters and expect it to solve every problem in the industry would be shortsighted. AI is very effective when it comes to known fraud cases and for several unsupervised use cases as well. But when there are for example, aren't enough proven frauds for a certain scheme, we use AI to automatically retrain on daily feedback that is coming from other techniques for instance business rules, third-party data network analytics to alert these cases and start getting more data in. So, these models will start to pick up on these patterns and false claims.
Laura Drabik We need to take another break. Just a reminder. If you're enjoying this podcast, be sure to subscribe to ensure top on Apple podcasts, Stitcher, or wherever you get your podcasts. Let's get back to our conversation with the chief executive officer at... Jeroen, what is the most effective phase in the insurance life cycle to target for fraud detection?
Jeroen Morrenhof The best answer I can give you is that you need to detect fraud at every stage. Over 60% of our customers use FRISS not only at claims, but also at underwriting. And, 95% of our customers use FRISS at claims, but also in SIU department at the same time. They all use this analogy. Let's say your sink is clocked up, but your faucet is still running. Eventually it overflows and you grab mop or some towels and start cleaning up. And then, you look up and a sink is still running, what do you do next? Do you grab more towels and keep moping or do you reach out and turn off the sink? Well, it sounds silly, but many carriers are still letting the sink run while they keep mopping up the mess. And, this is to say that they're still letting fraudsters into the book of business, even though there's a way to stop it.
So, the next question is where to start, and that differs per country. And, it depends on the privacy regulations and it also differs per carrier, their size, how fast they're growing through what channels, et cetera. So, if you have for example, a digital label focused 100% on doing business over the internet with little active policies yet, but spending a lot of money on acquiring customers, it's clear that they will start at the underwriting first. On the other hand we'll see a large mutual carrier with a loyal customer base a little to no checks in place. They will more likely up to start at claims.
And, when we're detecting fraud and claims it should not only be at the and ,[inaudible 00:11:51] stage. Every time new information is added or changed, you have to do an auto check. We see examples of all the time where legitimate claims comes in, or we want to pay this as quickly as possible. But then more information comes in, perhaps a passenger magically appears with some injuries. We need to check all of that to make sure we aren't pay on a front exposure of that same claim. And, that's the beauty of modern screening solution. You can do it quick and you can repeat it over and over again to cover all of your bases.
Laura Drabik You started for us in 2006, how has fraud evolved over the last 14 years?
Jeroen Morrenhof Well, previously I made that distinction between the more optimistic and organized fraud. And, in terms of evolution you would say the optimistic fraud hasn't changed that much really. It is typically also a not planned fraud scheme and done by taking advantage of the moment. The organized fraud on the other hand remained to do their... What we call market research. And, as the industry is getting better to detect these known fraud schemes, fraudsters are moving to either carriers that don't have those controls in place or lines of business where it's less policed, for example travel insurance, pet, insurance, and commercial arts.
Laura Drabik What critical piece of advice would you share with insurers looking to implement an AI fraud solution?
Jeroen Morrenhof Well, first of all, make sure you tie it to your strategic initiatives. There is no digitalization, no automation or improve customer experience without the ability to detect fraud and make the distinction between the dishonest and the honest. The majority of your customers that deserve to reap the benefits from all these strategic initiatives are depending on it. And remember, this is not only a one size fits all solution. It's not simply throwing AI at something you need the experience you can't build it all yourself. By the time you did, it would already be outdated. And, this isn't the best place to start a do it yourself project. We've heard from many carriers regret the time and resources they lost trying to do it themselves. Lean on a partner who will give you the honest answers, to good advice and be there throughout the entire process to make sure you're getting what you deserve out of such a solution.
Laura Drabik I really liked your comment about tying it to a carrier strategic initiatives, because that increases the probability of adoption and also being able to follow it through and measure it. Thank you for sharing that. Hey Jeroen, what's your experience been like working with Guidewire's a solution partner, do you think insurers appreciate our plug and play integration between the solutions?
Jeroen Morrenhof Well much like Guidewire at FRISS we embrace the ecosystem. And, this is also why we are making extreme efforts to be an open API based platform. And, besides integrating with Guidewire, we also integrate with a lot of data vendors, investigation platforms already in intro tech companies. And, the insurance ecosystem can be very complex and working with Guidewire really simplifies that. Our solutions connect across the entire enterprise and making these connections as easy as possible for the carrier is a lifesaver. It drastically reduces the implementation times and costs and that's walking through as one.
We want our solution to make your workflows easier and more efficient. And so, we're very proud to, to being a longtime Guidewire partner, working hand-in-hand with you to deliver the very first fraud analytics integration for Claim Center and carriers absolutely appreciate how much easier that makes it to turn on a solution like FRISS and for them to take full advantage of it every day.
Laura Drabik Jeroen, thank you very much for your time today and for your incredible insight into how insurers can become fraud fighting superheroes. You've showed us it's not just about ideas. It's about making ideas happen.
Laura Drabik Welcome to InsurTalk. My name is Laura Drabik and I'm the chief evangelist at Guidewire. In this episode, I have the privilege of interviewing Gary Hagmueller, president and CEO of CLARA analytics. Gary has been a leader in the technology industry for over 25 years, with a deep focus on building AI and machine learning applications for the enterprise market. Full disclosure: Guidewire has invested in CLARA analytics, but I specifically selected Gary for today's podcast because of his knowledge of AI, and CLARA's impressive results in delivering artificial brilliance for commercial claims. Hello, Gary. Thank you for joining my podcast today.
Gary Hagmueller Well, thank you so much for having me, Laura. Really looking forward to the conversation.
Laura Drabik Would you please fill our listeners in on CLARA analytics' value proposition?
Gary Hagmueller CLARA analytics is focused directly on claims processes. We help carrier, self-insured enterprises, and even third-party administrators to figure out things that are going on within claims that can help them reduce loss costs, can help them improve outcomes for claimants if it's a workers' comp claim, or otherwise figure out how to find the best services, or best attorneys, or best actions to take on a particular claim, and using that to reduce the cost and improve outcomes.
Laura Drabik So why did you focus on commercial claims?
Gary Hagmueller So in the early days of the company, what became apparent was that a lot of the earlier customers that we had interacted with really had simplistic models that they were building to understand severity, and understand who the right providers and other things like that were for their commercial claims. So what we realized as a company was that we could go and draw on a large pool of data. We generally sit on a very large pool of closed claims, and we could basically understand what was going on within those claims, both the phenomenon that would lead to higher or lower severity, and the people who would be able to service those claims most effectively, and really give our customers back something that was a game changer for them. Since that time, we've really spent a lot more time and effort really refining it, and so today what we've seen is, we're able to reduce leakage. We're able to do a lot of the sorts of things that are really out of reach for a lot of customers and to boot, it was a market that nobody was really addressing.
Laura Drabik Early intervention triggered through the detection of claim issues can dramatically reduce the open claim time and better control claim costs. How does AI leverage in the intake process, proactively identify, and resolve claim issues?
Gary Hagmueller So let's take some workers' compensation examples. For each of our customers, we ask for at least five years, and typically 10 years, worth of closed claim, and anything that's open. We put it into a collected data lake. So what winds up happening is we're able to look at millions upon millions of closed claims. And what we do is we cohort together claim types, so think knee injuries for men that maybe have diabetes. The machine helps us figure out what the right cohort is. We can then begin to establish what we call the optimal path, so the series of things that happen in a particular claim cohort, and then we can assess, "Okay, well, what does good look like?" And good can be lower cost, or they get back to work faster. And then we can also see, by the way, what's not so good, so the things that can go wrong in all of that. Once you have an establishment of what the optimal path is, you can then begin to instrument the different vectors that a claim can take at any point in the process, right?
And the earlier you can understand the claims that are likely to have problematic outcomes, or high severity, or may involve an attorney or other things like that, the more attention you can give it, and the better the outcomes that you can have, because you can get ahead of the things that could wind up going wrong. What AI allows you to do is, basically, have the machine go through an enormous amount of different systems and different records and really help you crystallize, "Okay, these are the actions you need to take. These are the things that need to get done, and here's the people that you need to call in order to get that done." And it really allows you to get ahead of all of the things that might otherwise slip through the cracks, because you just didn't know to look there.
Laura Drabik How do you measure the accuracy of the AI and its recommendations?
Gary Hagmueller The way to assess accuracy on these sorts of things is really to go off and do a couple things. First off, we can go back and look at benchmarks for what existed before implementation and what existed after. And the good news is that, because in many cases, what we're doing is we're dealing with the time that a claimant is off of work or the cost that's incurred, we can very clearly go back and say, "Okay, the score that we've given, let's say, this particular provider usually winds up very tightly correlated to exactly how much that cost is going to be." So that's one way that we can very easily understand what the accuracy is.
The second piece of it is that over time, as you begin to add more data to the data sets that power our models, what you're looking for is to make sure that you have stability, so that if you don't have stability, that's an indicator that there are still phenomenon that you don't really understand. But if you add more data and it brings out or accentuates what you already have, then you know that you're accurate. And then the final thing is we have a pretty large team of data scientists whose day in and day out job is to go through and figure out new and more interesting ways to make the models evermore accurate.
Laura Drabik Before we continue, listeners, if you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcasts, Stitcher, or wherever you get your podcasts.
Now, this is Laura Drabik, and let's get back to our conversation. I'm talking with the CEO of CLARA analytics, and we're talking about the use of AI in commercial claims. Gary, how does AI streamline workers' compensation or injury claims in general?
Gary Hagmueller So first off, we integrate into a large number of different systems that our customers run. What we're listening for is things that the adjuster needs to know in order to understand what's happening in the claim and what actions to take. Let's say, if you had a knee surgery, and you went to a doctor the day after the knee surgery and asked for a pain med. That's not an unusual sort of thing, but let's say a month goes by and you go to the doctor and you ask for some strong pain medications, that probably is an unusual thing. And you want to know that, as an adjuster, so that you can take some action, because if you don't take action, maybe the next week the claimant can go back to the doctor and say, "It didn't work. I'd like some Oxy," and now suddenly you have an opioid-attached claim, which is going to cost at least $10,000 more to get rid of the opioid addiction and has a much higher likelihood of leading to an attorney or other things like that.
So the simple knowledge of knowing something is happening out of sequence, and that you should take action. Maybe you should call the claimant and find out what's going on so that we can get ahead of it. Maybe recommend some pain mitigation strategies or other things like that, and the aim there is to have a better outcome. And it's in doing that sort of stuff that the AI can really help the adjuster understand more of what's going on than they might normally have in the normal course of operations.
Laura Drabik So how does the AI ensure, then, that the right medical provider, nurse case manager, or attorney are assigned to the right claims?
Gary Hagmueller Remember how I described earlier that concept of an optimal path? The optimal path is really the most powerful thing that we have in our arsenal, and keep in mind the optimal path is established over thousands, if not tens of thousands, of different claims in a particular cohort. So you really have a sense for a particular type of surgery, or a particular type of injury. "Here's what is supposed to happen." And then you can then match up the performance of particular doctors, particular attorneys, or whoever's going to wind up touching the claim. You can wind up matching up how they've performed against that optimal path. What it allows us to do is to give them all scores. So obviously what you want is you want as many high-scoring providers or attorneys, or whatever, interacting on a particular claim as possible. And you'll obviously want to avoid those that are going to be lower-scoring.
Now, keep in mind, they may be lower-scoring for a particular type of claim. They may not necessarily be lower-scoring for every single sort of claim, but certainly for the type of thing that you're working on, you want to make sure that you're selecting the right doctor, and the scoring is what allows you to do that. And again, it's against that optimal path.
One other thing that I would draw out, though, is that in the case of an attorney, it's not enough to just know how a defense attorney is going to perform for a particular claim type. Keep in mind, the AI does do that, right? It does look at, "Okay, let's look at all the different characteristics of the claim and figure out which attorney is likely to be most successful with it." But we can then also look at the plaintiff attorneys and figure out, "Okay, well, which ones are really specialized for a particular claim type?" Because keep in mind the applicant attorneys are really good at selecting cases that they know they're going to win with. And you want to know what it is about those claims that they think is going to lead to a positive outcome, and then you also want to be able to make sure that you are as fairly represented as possible by somebody who is going to be equally as capable of defending against that.
Laura Drabik So what is explainable AI, and how does it improve commercial claims processing for, again, claims like workers' compensation?
Gary Hagmueller What explainable AI is all about, is basically making sure that people aren't just told that, "Okay, this doctor is better than that doctor," but really giving you a sense for, "All right. Well, why is this doctor better than that doctor?" When we provide a score, we also have a lot of component scores. So things that talk about, "Well, how long is the claimant going to be off of work with this doctor versus other doctors?" So you can really begin to break apart, "Well, what is it about this particular provider that has led to this score?" Over time when people begin to understand, "Okay, I see how the score evolved, and I see why this was determined to be a better provider or lawyer, or whatever," they build trust in the system. And when you build trust in the system, the adoption goes up.
The other thing that trust in the system also gets is speed. Over time, you don't wind up questioning as much anymore, "Well, how did it get to that score?" And you basically say, "Okay, well, I trust this machine. It has given me the right outcome umpteen number of times before. And so therefore I'm just going to go ahead and execute on it." And usually when you act with speed, you generally wind up getting better results, as long as you're well-informed as to what you're doing.
Laura Drabik If you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcasts, Stitcher, or wherever you get your podcasts.
By the year 2025, millennials will account for 75% of the global labor force. I have to meet with carriers who want to understand how technology can help them attract this talent pool. How does the use of sexy technology like yours, and AI in general, attract millennials to claims positions?
Gary Hagmueller Here's my sense of what AI really does. It frees the adjuster from the stuff they don't generally like to do, and it allows them to then concentrate on things that probably got them into this role to begin with. What the AI can really do is make their jobs much more interesting, much more meaningful and much more effective. It's an opportunity to basically create much more fulfilling jobs that are going to make them much happier in terms of what they're doing. It's also something that... People like to be on the cutting edge, right? They like to tell their friends, "Hey, I'm using this cool technology. I'm using AI." There is that cachet that if you implement a solution such as this, you're able to attract people who are technology-forward, and who want to be at a company that wants to do things on an innovative basis.
Laura Drabik How is CLARA analytics adjusting to the changes in business behavior caused by COVID-19?
Gary Hagmueller Given the fact that our focus has been on workers' compensation, there was definitely a period of time where people were trying to figure out, "Well, what does workers' compensation even mean anymore, in a setting where a lot of people are sitting around on Zoom calls or not working, or other things like that?" So I'd say that was the early phase of what will end up happening. But the interesting thing is, what's happened since then, is that it has spawned a lot of our customers to think hard about expediting the adoption of this system. As I go and I talk to prospects and existing customers, there seems to be a sense that there's going to be a whole bunch of things that are going to come back into the workers' compensation system very quickly. There's a few plays that'll happen to that.
Number one, there's going to be new types of things that are going to show up, and nobody has any great experience in terms of how it winds up working today. And that could stem from remote working, or some of these other things that people are doing. Number two, I think there is some worry or concern that there will be claims that will open up that really are maybe filed for economic reasons, versus somebody who wouldn't file that sort of claim in a normal condition, and so there's a variety of things that are showing up that our customers are concerned about that.
But the overarching thing that it's done for us, frankly, is really allowed us to highlight the fact that the sort of solution we have, which brings in data from over 20 carriers at this point, and the number grows every couple of months here, it's best illustrated in how the COVID response has rippled through. Any individual adjuster in any carrier may have one or two cases of COVID show up at any given time. Let's say they're working 100 claims, or 200 claims or whatever winds up being. Even if you go across the entire adjuster base of the carrier, maybe they're seeing a couple of hundred COVID claims show up at any given time. Well, if you have a couple of hundred claims, that's probably not enough data to really help you do any of the predictive sort of things that we talked about earlier. It certainly isn't going to tell you which doctor's going to wind up being better at it. It's certainly not going to tell you how the lawyers are going to respond. And severity and other claim indicators, you're just not going to get a handle on.
But if you aggregate that across 20 carriers, now you're in the thousands, if not tens of thousands, of claims. And if you're at tens of thousands of claims, you have a pretty good sense for the sort of phenomenon you're going to see. You're going to have a better sense for treatment patterns that you need to undergo and all that sort of stuff. So what COVID has really done for us is really pointed out the need for this sort of industry-wide type of solution, and the value of speed to understanding complex phenomenon and how they interact with the claims process and how you can get it closed. So that's probably the best description I can give you for what COVID has meant for us.
Laura Drabik Gary, thank you very much for your time today and for your incredible insight into the positive impact that AI will have on commercial claims processing. You've showed us it's not just about ideas. It's about making ideas happen.
Gary Hagmueller Cool. Thank you so much, Laura, for hosting me today. Really appreciate it.