Episode 8 - Metromile Enterprise's Amrish Singh on Reinventing Claims Processing
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 Amrish Singh, general manager, Metromile Enterprise.
Amrish is a lifelong technologist, and has dedicated his career to helping companies operate with greater efficiency, and provide a premium customer experience. I specifically selected Amrish for today's podcast because of his strategic knowledge and experience launching the Metromile Enterprise Group, and how they reinvented claims processing.
Hello, Amrish. Thank you for joining my podcast today.
Amrish Singh Hi Laura, thank you so much for having me. I'm excited to be here today.
Laura Drabik Tell our listeners a little bit about who you are, your role, and your responsibilities.
Amrish Singh Yeah. I'm the GM of Metromile Enterprise, which is Metromile's software business unit. Much of my work and my team's work is based around taking the digital experience technology and the predictive analytics models that Metromile has built and powers its own claims experience, and licensing that to other insurance carriers.
Laura Drabik I had the privilege of working with Metromile many years ago, as they were evaluating Guidewire in the early days. One of the quotes that really stuck with me from your management team was, "We don't think of ourselves as an insurance company," meaning you're focused on providing value add services. Could you elaborate on this statement, and the Metromile value proposition?
Amrish Singh Yeah. I think that statement is interesting. I would argue that we think of ourselves as a technology company that's focused on the idea that modern approaches in machine learning and digital customers experiences can make the entire insurance experience just fundamentally better than we've seen it in the past. We're really focused on bringing the concept of fairness into insurance by focusing on pricing by the mile, rather than other factors.
Laura Drabik Now, you provide multiple service options for intake, including smart phone, web, and a central number. Which option is the most utilized, and why?
Amrish Singh Our principle is that customers should have the freedom to access the services they want at their convenience, so the medium that they prefer. There isn't really any one option that is always winning out over others, but what we have seen more recently is there's an increasing adoption of our digital channels. We're seeing more and more people access those, specifically on the claim side, using web or mobile rather than calling the call center.
Laura Drabik Do you think that has anything to do with COVID-19, and accelerating a digital process? Or, the use of self-service?
Amrish Singh I think this was happening much before COVID-19, though obviously there's been an acceleration.
I think the reason for that is because consumers, in general, are looking to interface with all of the things around their world using a digital medium. So those expectations are also, now, entering into insurance. If you give an option to someone to wait five to 10 minutes listening to a music playing on a phone line, versus opening up their phone and accessing the same service over the phone, we're just seeing more and more people prefer the latter option. Obviously, with COVID we've seen that just get accelerated at a much faster rate.
Laura Drabik How do you simplify and expedite claims processing, while also providing consumers with choice of how they want to be served?
Amrish Singh So, I think it goes back to that idea of choice, I think that's the right word. The idea to access the services they want, and at their convenience. From our perspective, we've built on a technology called Streamline that essentially powers the digital experience layer at Metromile. The idea is really to module out the entire claims process, and back that up with predictive analytics.
We're able to use this technology to power our claims experience, essentially allowing our claimants and our customers to file claims, and get a really seamless, quick process without waiting a lot for us to process out the claim or for us to seek information from them, which typically slows down the claims process.
So it's really a combination of our digital tools, along with this technology Streamline, where we've built predictive analytics and decision making in the back end, that's helping us triage the claims faster and route them more effectively throughout the lifecycle of the claim.
Laura Drabik What types of claims and activities are best suited for Streamline?
Amrish Singh We're seeing an emergence, especially in the auto sector, of the idea of direct to cash non-repair. Which is, essentially, the claimant goes in, they start filing a claim, and they choose to get a cash payout rather than going through the repair process. That was our initial focus, which is offer a digital medium for them to file their claim online, through a digital FNOL, then use predictive analytics to predict the likelihood of this being a candidate for straight through processing. Then, ask the options from the claimants, and then process it out completely digitally. Essentially, they're on their phone, they're sharing their options, they're submitting photos using their phone, and eventually we're presenting estimates and issuing the payout in that very experience itself. That's where we initially started.
Now, we've expanded broader, into the entire claims business process. We can focus on really simple claims, which is this direct to cash non-repair. We can focus on a lot more complex claims. We can also focus on partially automating highly complex claims.
An emerging thing we're seeing is the idea that claimants want to be able to book their repair ship appointments online. Or, book their rental, or get payouts faster. And our technology, really, is able to now help them with that entire claims business process, all the way from intake, to straight through processing for low complexity claims, over to issuing payouts quickly and seamlessly.
Laura Drabik Before we continue, listeners, if you're enjoying this podcast be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. And, you can rate and review this show on Apple Podcast, 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 Amrish Singh, general manager at Metromile Enterprise.
Amrish, can you tell our listeners about how Replay provides adjusters with an integrated view of the loss?
Amrish Singh Yeah. I think it goes back to the idea of using data for decision making. Replay is the tool we've created to help adjusters recreate the facts of loss, using independently verifiable data. Telematics is a big part of how Metromile offers its insurance product. We're able to tap into the telematics signal that's coming from the car, and recreate accidents, understand the intensity of crashes, confirm the information that we've received in a claim. Often, this is very much to benefit our claimants, because we're able to get the information we need to process the claim faster and more efficiently.
But, this also really helps the claims adjusters, because at the end of the day, claims is a lot about getting access to information so that they understand how to effectively process the claim. Replay's essentially replaying the entire trip, and understanding when the accident occurred, how it occurred, really and helping us to figure out how do we get help back to t customer as easily and seamlessly as possible.
Laura Drabik I recently read a statistic that you'd increased your automation percentage from 4% to 40% of claims process. How were you able to do this?
Amrish Singh You know, the idea of touchless claims, or automation on the claims, is an emerging trend. The challenge that most insurance companies face, and that also Metromile faced, is how do you start. When you have to start with a gradual process, pick the low hanging fruits and automate those, and then slowly increase the addressable claims that can get automated.
So at Metromile, we did exactly that. We focused on the super simple claims. Get the claims in, triage the really simple claims into an automation business process, and model out that automation business process where our machine learning models are helping triage that claim effectively. We then applied the learnings we got from those really simple claims, and then we started expanding to more and more complex claims. That, for Metromile, has meant we've gone from four, to now 40% of the claims, being touched by automation, which means some aspects of those claims are either partially or fully automated away. Really, eventually, honestly, helping our claims handlers address the more complex claims more effectively, and be more available for our customers when they're seeking help.
Laura Drabik According to McKinsey, high performing organizations are three times more likely than others to say their data and analytics initiatives have contributed at least 20% to EBIT. What role does data and predictive analytics play in ensuring that Streamline maintains an accurate automated workflow?
Amrish Singh That's a great question. You know, honestly, when you think about the benefit that you're looking to provide to the customer, which is a great customer experience, in order to provide that benefit you need to be able to offer the right kinds of services, and the right kinds of help, at the right point in time. That, for us, happens through these machine learning models that are making these predictions as accurately as possible.
So data is core to that, we tap very much into our core claims system in order to get that data. We use, then, these machine learning models at different touchpoints within the claims business process, such as triaging a claim, which is using a model to figure out how to effectively assign the claim. Or, predicting the likelihood of fraud, based on the information that's coming in through the claim. We are also using ML models to effectively route the claim through the organization, based on the information we got initially when the claim came in, but also based on the information we're collecting over time.
And finally, I think what we've learned is, at the end of the day, we're providing these tools to the claims org, ensuring the claims handler is more equipped, really, in doing a better job and being more effective a processing the claim. And much of that happens with access to as much data as possible, which a digital medium offers, but also having these models making accurate predictions based on this data that we're getting from our systems.
Laura Drabik How do you learn fast? And, how do you transition data recommendations into actions?
Amrish Singh It's built very much into how you establish a data science program at your company. There was, initially, this idea that we'll apply this data science model, and we'll press play, and it'll do all of its magic on its own, without ever needing any more improvements.
What we actually learned is that these predictive analytics models need a lot of nurturing, and care, and improvement. So now, whenever we establish a data science program, we provide these models, but we also provide a decision support system, a tool, that enables the handlers to provide feedback to these data science models. The way to become successful with any data program is to have this back and forth feedback loop, between the actual experts at the company, which is typically the claim handlers and the claim adjusters who have decades of experience, and how effectively they're able to provide feedback back to these automated models. This feedback loop is what we've learned is the secret to success at any claims organization that's looking to establish a data science program.
Laura Drabik How do your data scientists and adjusters work together, and thus stay in alignment?
Amrish Singh You know, claims is a very context sensitive domain. The people that are in the business have learned, over decades really, of how to process claims effectively. They understand the nuances with claims that somebody with a data science background wouldn't actually know.
So our idea is to create these teams, where we pair a data scientist with a team of claim adjusters. Really, the data scientist is the one that's modeling out the model, but they're getting insight and feedback from the experts in the field, which are adjusters. The adjusters are the ones that decide which signals to access, they explain to the data scientists how decisions are made. The data scientists job is really to codify that approach into the data science model. The key to success, really, is how effectively these two organizations are working very closely and collaborating. And for us, it's this idea of creating this internal team that's a mix of both data scientist and claim handlers.
Laura Drabik Just a reminder, if you're enjoying this podcast be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts.
Amrish, how have you infused fraud detection into your digital claims processes?
Amrish Singh A very, very, very, very large majority of humans are typically honest and fair, so we want to make sure that the services we're providing to them start with an assumption that there is a mutual trust between both parties.
We do have technology in the back end that's helping us effectively detect the risk associated with a claim. So we specifically focus not on the idea of calling it fraudulent or not, we focus more on surfacing rich insights back to the real experts in the claims organization, which is the claims handlers, and letting them take these automated insights, and then actually make a determination if the claim coming in is fraud or not.
At the end of the day, you're looking to make the claims process incredibly seamless. The way to make it seamless is to ensure that the claims that come in are processed out as quickly as possible. We focus on models, also, in not only predicting of its fraudulent or not, we're more importantly predicting if these claims are not fraudulent, which is sort of a counterintuitive term. We're a lot more focused on ensuring, if this is a claim that comes in where there's really no risk associated with it, let's put it through an automated process such that the claimant gets access to services, and gets the help they need as quickly as possible.
So both of these things, which is the idea of identifying as much risk as possible, but also the idea of ensuring that this particular claim has no risk, is powered by this core data science model that we have. Which may be called a fraud detection, we call of it more of a risk assessment model, to identify the risk associated with a claim.
Laura Drabik How did your modern technology landscape position you to respond to COVID-19, and the shift in consumer behavior to digital self-service?
Amrish Singh This has always been our approach at Metromile, offer access to services at the customer's convenience, and this has just accelerated, with COVID, a large majority want access to services using a digital medium. People are feeling a lot more vulnerable now than ever before, and when someone files a claim they're already at an incredibly vulnerable point in time of their life. So the principle, really, here is give them whatever they want, at the convenience that they need it, and digital really helps us in enabling that customer experience.
Laura Drabik Amrish, thank you very much for your time today, and for your incredible insight into how you have reinvented claims processing. You've showed us it's not just about ideas, it's about making ideas happen.
Amrish Singh Thank you so much, and thank you for having me.
Episode 7 - Betterview's David Tobias on Geospatial Analytics in Insurance
Laura Drabik Welcome to InsurTalk. My name is Laura Drabik and I am the Chief Evangelist at Guidewire. In this episode, I have the privilege of interviewing David Tobias, Co-founder of Betterview. I have had the pleasure of working with David in our industry for several years now, and I specifically chose him for today's podcast on geospatial analytics because of his in-depth knowledge on today's subject matter. Full disclosure, Guidewire is an investor in Betterview. I also recently discovered that David is my neighbor, neither had any bearing on me selecting him and Betterview for today's show.
So with that, hello David. Thank you for joining my podcast.
David Tobias Yeah, thanks Laura. Thanks for having me, it's great to be here.
Laura Drabik Great. So geospatial data is data about objects, events or phenomena that have a location on the surface of the earth. Geospatial analytic providers like Betterview gather geospatial data and imageries from sources like satellite and manned aerial. You analyze, manipulate and display the data collected in a consumable fashion useful for insurers.
David, how did you come up with the idea to start a company that employed geospatial analytics for insurance companies?
David Tobias I grew up in the insurance inspection business of all things, and in that business, we saw insurance carriers coming to us for more and better roof data. So roof claims were starting to become a really big problem for them, hail claims, wind claims, accounting for about 40% of all property lost dollars paid out going to these roof-related claims.
So on the underwriting inspection side, they were asking us for more and better roof data, and Betterview originally started as a drone company to utilize drones to go capture this imagery and this data for the insurers. We pivoted into using manned aerial and satellite imagery to do that because it was more scalable. But that was the real initial impetus to starting Betterview, was really just solving this customer pain point around roof data.
Laura Drabik So, that makes a lot of sense. I also appreciate that you grew up in the insurance industry, your father founded the insurance inspection company almost 50 years ago, and then you took over the business, grew it and sold it. How has your insurance experience helped evolve Betterview's value proposition?
David Tobias It's interesting. I think coming from the industry and really starting with the problem first rather than the technology or the solution, has kind of guided us along the way. I think that's part of the reason we're still here, is starting from that side of it makes it a little bit different. I mean, I think in any company or startup, it's a bit difficult to create a solution, whether that be software or hardware or whatever it might be, if you haven't lived and breathed that problem yourself.
So, I was lucky enough to have that opportunity to see the problem from the inside and see how it was impacting insurers. I do think it's important, especially in this industry. We're in a regulated industry that has a lot of rules that need to be abided by and fundamental ways that things have been done for a long time. Not to say all those things are correct in the way that they happen today, but I think you really need to understand those if you want to present a solution that's scalable to the industry.
So I think it's helped us quite a bit. Obviously, I'm biased, because I do have that background, but I do see other insurtechs struggle at times. Even if their technology is great, if they don't really truly understand the problem, it becomes hard for them to sell it into an industry they don't really know. I think we've been really lucky from that standpoint, and I think it's helped us quite a bit along the way, creating a better product, but also being empathetic with our customers.
Laura Drabik Yeah, that makes a lot of sense. Especially when you talked about regulation, it is really important to understand that there are rules and legislation in place that insurtechs need to work with or around. So, the global revenue for geospatial analytics will reach $166 billion by the end of 2028. What are your thoughts on that number? Too aggressive or not aggressive enough?
David Tobias I mean, it's really been amazing to see, even in the short lifespan of Betterview in comparison, how much geospatial data is out there and increasing, and the rate at which it's increasing. So I think when we started, we had one manned aerial provider, as an example, now we have eight. That's just amazing to see how this is happening, right? There's more satellites going up, you can thanks SpaceX for some of this and others. There's more manned aerial happening, there's interesting things happening on the drone side to capture more imagery. There's companies flying high altitude balloons capturing imagery now.
For us, this is all great, right? The more imagery we can get, the more geospatial data we can get. It's not just imagery, it's property data, it's flood data, so on and so on. It helps us and in turn our customers with a better product, a better experience. But this geospatial data is touching every bit of our lives, even as consumers these days. Think about self-driving cars, delivery, urban planning, and then of course insurance, right? Those are just a few.
That 165 billion number, I think it's probably accurate or small. Obviously, we'll leave it up to the economists and the folks to figure that out. But I think you're just going to see more and more geospatial, and not only more of it, but more recency on it. I think that's a really important thing to remember. Just having an image for example, that's great, but having an image that was taken a month ago or a few days ago is even better, right? For all those things, not just insurance, but self-driving and delivery planning.
So I think you're going to see more geospatial data as a whole, and you're going to see more recency on it. I think those two things are really going to help the insurance industry.
Laura Drabik Before we continue, listeners if you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. Now, this is Laura Drabik and let's get back to our conversation. I'm talking with David Tobias, Co-founder of Betterview.
David, on the underwriting side of the business, a clear benefit of geospatial analytics is providing insurers with insight into their risks prior to issuing or renewing the business. Without sending someone onsite, the carrier can see whether the roof needs replacing or what the proximity of structures are to brush, et cetera. What other benefits does geospatial analytics provide the carrier?
David Tobias So we talk a lot about roof and you've heard me talk about it here, but there's so many other pieces to the data that can be unlocked. We use building permits for example, we have distance to brush exposure, whether there's a pool on the property or not. All these things that are relevant to the rating of the property, the pricing of it, but also the risk factors that are there. So it's really not just about the top down view, it's about everything that's on that property. Is there debris in the yard? Debris in the yard's a leading indicator for many carriers to potential loss in the future, to the maintenance of the property and also liability hazard.
I was talking earlier about the frequency of the imagery. In many parts of the United States, we have images getting captured three to four times a year, so we know that progressive timeline of a building. So, one example of where we have carriers using this today is looking for premium leakage. So there wasn't a pool and now there is a pool, and that wasn't maybe reported, maybe the insured forgot to report that to the carrier. Well, the risk factors have changed, the pricing has changed and now the insurer can have a productive conversation with the insured, or the property was 3,000 square feet, and now it's 5,000 square feet and the premium should be adjusted to support that.
So, I think there's just so many different elements around this data and it goes beyond roofs. It's the property as a whole, and we're seeing a lot of interesting ways that this data can be used for things that you might not think. Not just property, liability and work comp and other use cases as well. So really, really interesting when you think more holistically around geospatial in general.
Laura Drabik What about the benefits to the policyholder? I mean, one I can think about is that insurance companies could provide consumers with perhaps a list of repairs or replacement items required to help reduce the chance of loss. What are you hearing from carriers as what is the benefit for their policyholders?
David Tobias The industry as a whole has really lived in this world of repair and replace, right? We're here to pay claims when they happen, we're going to repair it, we're going to replace it, we're going to make you whole again. I think that we're seeing this trend in the industry of trying to get more from repair and replace to predict and prevent. I think that shift is really beneficial to the insureds, because we see a lot of these carriers today really trying to make the insured a partner in risk management, verse this reactive repair and replace format.
Customer expectations have changed, right? We're used to instant gratification. I can go on Amazon, I can get my groceries delivered in a few hours, right? Insurance is lacking there, and you see some of the newer insurtech insurers purchasing home maintenance companies to help support their insureds. So we think this trend of predict and prevent is going to continue and it's going to deepen the relationships with the insureds, which ultimately makes for a better customer experience.
I think that's really good for everybody, right? You don't want to deal with having your roof damaged and having water leakage and replacing your valuables. But something like your roof, in many cases, it's out of sight, out of mind. If the insurer can come in and say, "There's a problem with your property, we can help you fix it and show you at least what's wrong with it, and maybe here's some places you can go to fix it." I think that's going to be a value to the insureds, it's going to help differentiate the insurers that really are more proactive than reactive.
Laura Drabik So, geospatial analytics can help empower adjusters to desktop adjudicate catastrophe claims. As a former adjuster who worked catastrophe claims, why don't you tell us a little bit about how your technology supported the Kilauea volcano eruption and how you helped to improve the safety of the adjuster?
David Tobias Yeah, we see people using our geospatial data like this on two sides. One's in the CAT response side and another's individual claims. Individual property claim comes in, they'll use a platform like ours to look at the historical imagery on that property, see if there was prior damage, see what the permit history is before they even get boots on the ground, right?
Then on the CAT side, that volcano example's a great one, because this whole area was cordoned off, you couldn't get people in for a long time into these areas. So, we had an insurer call us and say, "Hey, do you have any imagery? We just want to be able to tell our insureds who own these homes, is it a total loss, partial loss, no loss, because they want to know. They had to leave in a heartbeat and they have no idea."
So we were able to use the satellite imagery, because the satellite is going to have more refresh. We can get satellite pretty much anywhere in the world, usually a few days old at the most. So, we were able to tell them which homes were gone, which ones were partial losses and which ones hadn't been touched yet. That was really valuable for them. It was valuable for their insureds as well, and it helped situate the CAT teams to where they needed to be once they did have access to go in. If it was a total loss, they didn't necessarily even have to send in an adjuster.
So, this kind of triage element can really help on the CAT side, and again, provide a better customer experience, because you can get that customer information sooner than they would have gotten it otherwise.
Laura Drabik So, David, can you tell us why you got out of the drone business?
David Tobias What it comes down to for us is we've always been really customer-centric. You heard me speaking about starting with the problem, not starting with the solution, and the drone was a solution. We really created a lot of value there, we had 30, 40, 50 customers on that platform. The thing that we kept hearing though is these drone reports gave us the level of detail that the carrier wanted, it solved their problem, but it was too expensive and too slow.
So, we really just listened to the customer and were able to say, "Okay, we know what the problem is, what are other ways we can solve this problem for our customers?" We said, can we do this with manned aerial and satellite? We're able to use a lot of the computer vision technology that we had created, but it was, at the time, a somewhat painful decision. We built a lot of technology around drones, we had solved the problem, but it wasn't scalable.
So for us, it wasn't that hard of a decision if you listen to the customer. We've seen a lot of technology startups go down with the ship because they were wedded to the solution. We're not married to the solution, we're married to the problem. I think that's an important distinction.
Laura Drabik That is an excellent example of starting with the business problem and constantly going back and strategically probing and making sure that you are resolving that business problem in a cost-efficient and accurate manner. Thank you for sharing. So, how has geospatial analytics helped carriers in the wake of COVID-19?
David Tobias What COVID did is it made it a necessity. Something that was maybe on a six-month or 12-month timeline to implement, now became a two week or three week timeline. I think we saw carriers move very, very quickly, because they had to. Boots on the ground inspections were postponed, canceled, totally stopped for most of these carriers across the United States.
The challenge that the carriers we work with face though is they still had policies coming up for renewal, they still had new business coming in that they needed to quote and bind, and they still needed data. So, their traditional workflow of maybe sending their own risk engineers out and things like that could no longer be done.
So we had carriers who came to us and said, "We need to put 300, 400 people in your platform and we need to do it in a week or two weeks." It became critical that we really helped people get information about property that they couldn't otherwise get. So, this type of technology has helped fill a gap in lieu of the traditional ways with COVID happening.
Laura Drabik Just a reminder, if you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. Let's get back to our conversation with David Tobias, Co-founder of Betterview.
Aerial and satellite imagery is predicted to improve dramatically in quality and quantity over the next few years. Taking that into account, tell us what underwriting and pricing will look like five years from now.
David Tobias As we think about this dramatic increase in the quality and quantity and the recency of this imagery, it's really allowing for this type of data to be moved sooner in the process. So, what I mean by that is something that maybe today or in the past was only happening on renewables because there was more time to deal with it, now can happen earlier in the process.
So, we were talking earlier about really being a partner in loss control and risk mitigation with the insured issuing recommendations, cut back the tree overhanging over your house, for example. If you can get access to that data earlier in the process, even during the quote, it can change pricing, but it also can change the experience with the insured, traditionally that kind of information or recommendations would only happen post-buy after an inspector had gone out to the field.
What this influx of imagery and recency of that imagery is allowing for you to do, and then of course the computer vision, the machine learning to turn that into something actionable, is to use that data much earlier in the process, right? Even all the way up to the quote. So imagine issuing a quote to somebody and saying, "Yeah, here's your price. But before we bind this policy completely, you have to cut the brush back from your house. You're in a high wildfire zone, we can see from this image that you've got dry brush surrounding your property within 10 feet. Cut that back and this'll be your price, and you have 30 days, 60 days to do that."
In doing that earlier, ultimately, I think, will lead to more accurate pricing and set the stage for bringing that true partner in risk control very, very early. So I think over the next few years, I see this type of information really being brought forward in the process and not being used as the secondary piece.
Laura Drabik I really like your two points. One about transparency into the process, it's usually a black box that consumers really don't understand, and then secondly, starting to establish that trusted partnership between the carrier and the consumer. What critical piece of advice would you share with carriers exploring the usage of geospatial analytics? What do they need to consider and prepare for before implementing?
David Tobias The carriers we've seen be most successful, they have many, many processes in place, right, that might be set in stone over the course of 50 years or longer. So, the people we've seen be most successful are the ones that take that process or those processes, and they take this geospatial and they insert it into the processes that already exist as a step one. Trying to do a wholesale shift of an entire process on day one is very difficult for any organization to do.
So, my recommendation and based on what we've seen work is to take these processes that are already in place, maybe start with your renewals as an example. Start in one spot, make it successful there, and then go through the rest of it. I think we see a lot of move towards automation and straight through processing, but we also see a lot of carriers using data like this to really make their humans superhuman.
At Betterview, we're trying to help people find needles in the haystack and we're trying to shrink the haystack, so that the stuff that the humans really need to get involved with and use their highly skilled brains for is a smaller stack, right? A smaller haystack. So my advice is to start with a couple of processes that exist today, insert this into those processes, make them much better before you move onto the next one and the next one and the next one. I think it shows incremental wins too, which is helpful for the organization. So, that's what we've seen work so far.
Laura Drabik Yeah, taking a focused approach, that makes a lot of sense. So David, thank you very much for your time today and for your incredible insight into geospatial analytics. You showed us it's not just about ideas, it's about making ideas happen.
David Tobias Thanks Laura, appreciate it. Thanks for having me.
Episode 6 - CAA Insurance’s Matthew Turack on Pay-as-You-Go Insurance
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 Matt Turack, president of CAA, Canadian Automobile Association Club Group. Matt is a senior executive leader with 19 years of insurance experience, and has led teams in underwriting, actuarial, claims, as well as sales and service. I had the privilege of meeting Matt during CAA's evaluation of Guidewire, and I specifically selected Matt for today's podcast because he drove Canada's first pay-as-you-drive insurance offering, CAA MyPace.
Hi Matt, thank you for joining my podcast today.
Matthew Turack Thank you very much for having me, Laura.
Laura Drabik CAA Insurance Company is part of the largest Canadian Automobile Association club, with over two million members. You already offer extensive services like roadside assistance, automotive services, as well as insurance coverage like home and auto insurance. Why offer as a pay-as-you-drive offering? What were your business drivers behind launching this line?
Matthew Turack Well, we went out to CAA members, and we really asked them what are they looking for in their insurance. And, what are the things they want to see their insurance product, auto or home, look like in the future? This was back in 2018.
What we heard from customers, and we heard from members and insurers, is that they want flexibility. They want some choice, they want control over how their insurance operates. We used that feedback, that insight, to help use create CAA MyPace, where we could provide consumers with that choice and that control over how their auto insurance works.
We knew that low mile drivers, or drivers who have multiple cars, or driving less want to be able to pay less for insurance, and want to see the discount that comes along with the less exposure that they have. We know that there was a shifting lifestyle focus, where people's lifestyle looked and felt different from each other, so we wanted to create a product that really spoke to those life stages. And really helped put the consumer first, allowed the consumer to have that ability to flex their cost, allow the consumer to control what, ultimately, their insurance, how it behaves and what it does cost them in the long run. And, allow them to make some decisions along the way.
So we used all that information to really create, as you said the first, and still only, pay-as-you-drive auto insurance program in Canada.
Laura Drabik That's a really powerful message, create a product for all life stages. What differentiates CAA MyPace from other telematics offerings in Canada?
Matthew Turack Well, the other telematics offerings in Canada, whether it's with CAA or other companies, really looked at a discount based on driving behaviors. So how you speed, what time of the day you're driving, acceleration or hard braking. CAA MyPace, while it uses the same technology, we don't use any of those variables in our rating. CAA MyPace is simple, it's really just about how much you drive, the number of kilometers you drive.
So it allowed us to create a program that lets people manage their mileage, and therefore save money when they're driving less.
Laura Drabik So, I like that, the beauty and the simplicity. It's about simply tracking the number of kilometers.
Matt, when I think of the typical pay-as-you-drive consumer, I think immediately of a Millennial. But, that's not the case with CAA MyPace. Who ended up being your main consumer, and why?
Matthew Turack You know, when we designed the program, one of the first thoughts that came into our mind was the Millennial generation, who is using multiple modes of transportation, potentially living in downtown cores and cities that they can walk, they can cycle, they can use public transit. And, aren't driving or aren't picking up on owning a vehicle in the same way as people did in my generation or in other generations. We thought of that same cohort of customers.
While there are Millennials that are buying into MyPace and still want to own a car but drive it on the weekend, or use it along with other modes of transportation, what we find more and more is that we're getting the generation that is more established in their ways of driving. That they have two or three cars, that they use one car for more of the transportation, and the second car that may be weekend drives. Some of the older population, where driving is not as frequent. Maybe they drive to the grocery store or to see families, but are more at home. And really, any motorist who drives under 9000 kilometers will save money on this program, so it speaks to various different life stages and lifestyles, and various different age groups.
I would add to that, that with the pandemic going on, it's added a whole new dynamic. People working at home are not driving their cars as much as they used to, and this program allows people to save while they're working at home, and while we're in the current environment that we've been faced with.
Laura Drabik So, no one could have predicted such a massive shift in commuting patterns a few months ago, but so many people are working from home. It's reasonable to expect that pay-as-you-go models would become more common post pandemic. What are your thoughts, Matt? And, any other observations you can share with us on trends?
Matthew Turack Absolutely. CAA MyPace continues to gain tremendous attention, and we are seeing growth year over year of about 250%. Which tells me that consumers find the CAA MyPace program very applicable to their life stage and lifestyle now. Over the last few months, we've seen driving down 50%, and they still expect kilometers to be less even with some of the stages of reopening going on. So pay-as-you-go as an auto insurance innovation is very relevant for a lot of consumers, and a lot of what our future holds for us. It allows that flexibility, it creates the choice, and allows insurers and consumers to really be in charge of how much their insurance is going to cost.
Laura Drabik Before we continue, listeners, if you're enjoying this podcast be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. And, you can rate and review this show on Apple Podcast, it helps others learn about and discover the show. This is Laura Drabik, and let's get back to our conversation. I'm talking with Matt Turack, president of CAA Insurance.
Matt, you support multiple distribution approaches with MyPace. The CAA agent, direct to consumer, and broker advisory service. Is it the exact same product and rates offered through all distribution channels?
Matthew Turack It is, and we're extremely proud to be in multiple channels, and have the broker channel and the direct agent channel. It is the exact same product, the exact same price. We believe that consumers should have a choice in the channel in which they wish to shop.
Laura Drabik When you have to pay commissions on the human based channels, how does this affect your profitability for this line?
Matthew Turack The cost base between the two channels for us is the same. We don't see a difference in terms of what channel costing more than the other channel. Our direct channel, we have in-store agents and we have call center agents. So we have a blend of a call center and a captive agent in retail location model, so our cost base to run that channel versus the cost base of paying commissions works out to be the same. Therefore, it doesn't change our overall profitability or the ability to rate the program appropriately.
Laura Drabik The launching a new line like CAA MyPace must be a complex undertaking, involving actuarial modeling, regulatory collaboration, and marketing efforts. Give us a glimpse behind the curtain, and fill us in on the effort involved in creating and launching this new line in Canada.
Matthew Turack Yeah, it started with us really working on what consumers want, and what would fit within the consumers need from an auto insurance product. Then, we started working with our regulator to say, "How can we take this idea, and this innovation, and make it work with the regulations that we have in Ontario, or any of the provinces that we underwrite in?" We worked very closely with the regulator to vet out the ideas, vet out the concepts, to work out how can this program work and ensuring that it accomplishes the consumers choice, the flexibility, and provides the same rating program regardless of which channel, or whether it's a standard policy versus a MyPace policy.
We then started looking at okay, how do we explain this program to consumer? And how do we get this out in simple and plain language? So that consumers will understand enough about the program so they will call and talk to a licensed advisor, or a broker, about the details of the program. Without trying to market “insurancese,” the insurance language that nobody understands. We really tried to keep it simple. And make it so that consumers can understand that you buy in 1,000-kilometer increments. We know exactly how much those 1000 kilometers are going to cost you. And that you're never going to pay more than a standard policy, you will absolutely know that there's no downside to this. There's only upside which, if you drive less than 9000 kilometers, you will save.
Laura Drabik There is no downside if you're never going to pay more than a standard policy.
Matthew Turack The only time in which a customer wouldn't see the MyPace as advantageous for them is if they are driving a lot. You're going to pay in 1000-kilometer increments, so if you drive 2,000 kilometers in one month you're going to pay twice for that 1000 kilometers, you're going to buy 2,000 kilometers.
So if you do drive a lot, this program may not be the right one for you because you're going to be billed more frequently. That would be the only time in which I would say to a customer that a standard policy may be the right one for you. But, in any occasion, you will not pay more than a standard policy.
Laura Drabik Yeah, that's a really great way to sum it up. Heavy drivers, this might not be a great fit for them.
So French and English are the official languages of Canada. Are you supporting both? And what is the added complexity of creating a new line for multiple languages?
Matthew Turack Yeah, we do support both, French and English. You know, the complexity of that is, really, you have to make sure your app if you have it, or any webpages you have, provide both documentation in French and English. Any forms that you have are both in French and English. And, that you can provide customer support, whether that's through a broker or through an agent. CAA is very proud to provide both French and English support on all of our programs.
Laura Drabik We need to take another break. Just a reminder, if you're enjoying this podcast be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. Now, let's get back to our conversation with Matt Turack, president of CAA.
Matt, how do you involve your brokers in coming up a new product ideas? Can you tell us about where you get your new ideas for products?
Matthew Turack Yeah. Our product ideas come from brokers, agents, internal staff, underwriter, and a lot of research that we do in the marketplace, and globally. But, we do meet a lot with our distribution people, our brokers and our agents, and talk about what they're hearing from customers. What are they seeing the need is? Where do they think insurance products need to evolve? What are the things that have worked really well within the insurance products, that they're seeing customers appreciate and they like? And, what are those service moments? What are the things that are making a company or an interaction with a customer difference? That they know that they want to keep repeating, because that's what builds loyalty, and that's what builds response, and that's what, ultimately, they sell and they win on.
So we take all of that feedback and those ideas, and we try and look at, well what can we do from a product suite perspective, and what fits what CAA does? We don't want to put products out there that we can't back with the customer service, and risk, and pricing prowess that we have. We want to make sure it does fit what we specialize, and what we do really, really well. Ultimately, we need to make sure we can provide that amazing claim moment. That moment of truth where, when customers need us, we know we're there, and we're there with white gloves on. And that we live up to what CAA's purpose is, which is really being obsessed with both amazing product and amazing claim service.
Laura Drabik Now, you won a Guidewire Innovation Award for your MyPace initiative. I'm a judge, I voted for you. What I appreciated most about your submission were the metrics. You had an increase in 30% in auto policy sales, an increase in cross sell opportunities, and increased diversification of your customer base, with 70% of your MyPace sales being new customers to CAA. That screams success to me. How do you measure and define success?
Matthew Turack We designed the program by asking the question what do consumers like, and what would they benefit from, and we created a product that revolves around it. To me, the success is in delivering a product that answered those needs of the customer, that delivered on what the customers were telling us. That we could center on what is customer focused, and delivering a product and a program that meets different lifestyles. That, to me, is what has made CAA MyPace successful. And we've been able to create a product, and innovate a product, that isn't usually innovated in Canada, or in auto programs, to really spark the start of what you can do with auto insurance.
Laura Drabik Yeah, I really like what you just said. Ultimately, an innovation initiative is a business opportunity, and it comes back to ensuring that you're meeting the needs of your stakeholder, which is your customer.
In a recent survey of large companies by Harvard, they found that politics and turf wars were the biggest obstacles to change. What was your biggest obstacle to change, and how did you neutralize it?
Matthew Turack We don't really get into the turf wars at CAA. The biggest challenges we see for companies is looking for ways to pursue a pay-as-you-go model are the lack of technology. The lack of an appetite for risk, or the fear of cannibalization. We see those aspects, technology, cannibalization, and risk appetite, as really the main challenges for companies launching into pay-as-you-go.
It really draws on the core value of an organization to keeping moving forward, to keep looking at what customers needs are, and to really take the leadership and deliver something that really answers that question for consumers. And doesn't focus on the fear behind cannibalization, or the risk appetite. Yes, I may reduce my premium by doing it, but I'm going to end up retaining more customers, selling more policies, and delivering better value.
So, we've been advanced at CAA from a technology perspective, we put Guidewire in as our policy management system, in 2012, on a full suite basis. That has enabled us to do amazing things, like innovating our product, and continuing to evolve and create things like CAA MyPace. If we didn't have Guidewire, we definitely would not have been able to build MyPace, it really is part of the foundation that lets us innovate. I thank you guys and the company, for everything.
Laura Drabik Well first of all, thank you very much for the call out, we do certainly appreciate your business. And, I have one final question for you. What is the one critical piece of advice you would share with carriers planning to launch a new line?
Matthew Turack I think companies traditionally create product and figure out the math and profit margin, and the rating structure, and then try and figure out how to sell that to consumers. And, convince them that it's the right product for them.
The advice I would give to the industry, to companies, is that we designed CAA MyPace in the opposite way. We looked at what do consumers want, and we looked at studies and focus groups, customers told us the way they want auto insurance to work. And then, we challenged ourselves to say, "Does the data support that? Do we have data to say that driving less equals the probability of less losses?" Or, drives lower frequency. And then, once we were able to come up with all of those answers, then we went out and created the new product line.
I would encourage companies and insurance professionals to not get stuck down on regulatory meetings, to not get stuck down on creating a profit margin and the math behind an idea first. But, to really be innovative, create suggestions, create products, talk about it with your regulatory, and really drive that innovation because it creates customer value. Customer value is what we all need to think about, continuously.
Laura Drabik Matt, thank you very much for your time today, and for your incredible insight into launching the first pay-as-you-drive line in Canada. You've showed us it's not just about ideas, it's about making ideas happen.
Matthew Turack This was great, I thank you.
Episode 5 - Shift Technology's Jeremy Jawish on AI's Impact on Claims
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 Jeremy Jawish, Chief Executive Officer at Shift Technologies. Jeremy is not only the CEO of Shift, he's also the co-founder of the company, which specializes in AI-powered fraud detection. I had the privilege of meeting Jeremy as part of our Guidewire partner evaluation process. I was impressed with his knowledge of the industry, and also with the Shift fraud detection solution, Force. I specifically selected Jeremy for today's podcast because of his knowledge of artificial intelligence, and its potential to simplify claims processing. Hello, Jeremy. Bonjour. Thank you for joining my podcast today.
Jeremy Jawish Hi Laura. Nice to be here with you, and thanks for having me.
Laura Drabik Artificial intelligence is the development of computer systems able to perform tasks that normally require human intelligence. A couple of everyday use cases I can think of include speech recognition, digital assistance, self parking and driving cars. Jeremy, you started your career at a leading multinational insurance carrier doing mathematical modeling for fraud detection. What was the impetus for creating a solution, and company that leveraged AI, and how do you employ AI in your solution?
Jeremy Jawish The first job I got with my co-founder was, like you were saying, working in insurance company. We got lucky that we ended up very quickly in a special investigation unit. There, working with experts, we realized that, in the industry there are amazing experts that been detecting fraud for several years, knows really all the business knowledge. We realized that, with all what is existing in terms of artificial intelligence out there, we could really help them scaling their job. We created our company on the idea that we wanted to create a software service solution where we can help them by taking all of their knowledge, their expertise, and have a machine that would learn on that. Then, give them outputs of algorithms that takes their knowledge, and apply it on a huge number of data.
Jeremy Jawish You would see it as, you put all the data you can get from carriers data, external data what's online, and then you would take their knowledge and get outputs that are coming from, if this person would have been able to look at all the data at once, at every single data that is connected to this claim, or to the network that has claim has in it. What would be the outcome, what we would have received.
The way we employ AI in our solution is, we really take a machine that learns on the behavior on expert, and then we look at simple data, but also complex data like invoices, pictures, documents, external data, online data, weather data, maps data, and really behave the same way a human would have behaved looking at these data and say, "Oh, this looks suspicious," or, " When I connect this invoice, and this circumstances, and this description, something is not adding up," or, "When I look at 20 claims that are connected because of phone numbers, or because of address, or because of connectivity between parties, we should look at it. This is the reason why."
The hard part about applying AI to insurance is that, if you take a generic AI and you don't give it the knowledge and expertise of insurance, you won't have the right outcomes. Being able to detect it and say, "You should look at it because of that," this is the really difficult part, and why we need AI to be efficient on this front.
Laura Drabik I love your statement where you said, "We got lucky, and we landed in the SIU department." I'm sure SIU units around the world are applauding you for that. I also like how you describe your solution as taking the SIU units, the humans', tribal knowledge, and then gelling that into a framework, or into those algorithms. I can see a number of benefits, including improved scale, as well as accuracy of detection of fraud. What other benefits are there for leveraging AI in fraud detection?
Jeremy Jawish What we've seen with several years working with carriers is that, what's important is being able to detect what's suspicious, and being able to give an outcome to handler saying, "You should look at it because of that." The more we were training our algorithm, the more the handlers, and adjuster trusted our algorithm. The more, on simple claims, that you could really accelerate, for instance, on simple travel claims, simple physical damage claims in auto, or simple property claims. In order to accelerate it, it would look, did the machine said it was suspicious, or did the machine seen anything that is not normal? If it's green, they just fast track it.
We've seen that, in really small, simple claims, our customers are really using more and more our platform in order not just to detect fraud, but really select the ones that will accelerate and not spend a lot of time on it. Also, not ask too much question to the consumer. We really accelerated the time between first notice of loss and settlement, thanks to the fact that these claims are not suspicious, or we know that they won't be severity, or we know that there's nothing to question.
Laura Drabik I really like your angle there about improving the customer experience by reducing the number of questions that you ask them. AI has evolved to include machine learning, and deep learning. Can you explain to our audience the difference between the two?
Jeremy Jawish Yeah, I think the easiest way to explain the difference would be to see how we apply it, and how we use it in our platform. Machine learning algorithms, we mostly use them to learn on looking at how handlers behave after looking at these alerts. Should we put more of these alerts into the pipe, or should we put less, or when this type of information is there, is it more suspicious or less suspicious? It's really about learning on previous behaviors of users in front of alerts. Also, is it more or less suspicious, or is it easier to prove is it fraud or not?
Jeremy Jawish Deep learning algorithm, we really use it in understanding really complex data. Like understanding really long claims description, understanding documents, looking at pictures, and does the picture match the circumstances? Does the picture has damage that is matching what the consumer's describing? At the end, you need both to have the best quality of detection in terms of types of fraud you detect, and accuracy, but combining them with learning on previous, and understanding very complex data with deep learning, gives us the best outcome.
Laura Drabik AI can streamline different phases of the claim life cycle, as you know, by automating those manual redundant activities. For example, intake; vehicle sensors can trigger first notice of loss, and from there, AI can automate claim triaging, activity routing, and scheduling of repair services. Automation improves cycle time service, and claim cost. What other areas of the claim life cycle are ripe for disruption with AI?
Jeremy Jawish It's pretty interesting how insurance is being more and more targeted by very talented AI teams. It's one of these verticals where a lot of scientists realize that there's so much data, and there's so many thing we can figure out that a lot of them have invested in. We've seen a lot of teams now, really spending time on understanding more automatically parts of the claim. Like having an accurate estimate for some claims, having a better on patient preventions, for instance.
Jeremy Jawish A lot of carriers we work with are spending a lot of time on, can we predict that a claim is going to happen, and can we call our customers and tell them, "Before the claim happens, you should check this. You should be sure that you've done a diagnosis on your water pipes. Your electricity's to be checked." They're really trying to be ahead of the claim, because at the end, if you're able to have an expert saying, "If I was you, I would check something in your house, or check something in your car, or do some better maintenance," everybody wins. The consumer would win because a claim is not the best thing that could happen to them, and carriers win because they have less claims to pay.
Laura Drabik Before we continue, listeners, if you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. 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 Shift Technologies, and we are talking about the impact of AI on claims processing.
Jeremy, biases find their way into AI because of bad data, which contains implicit racial, gender, or ideological biases use to train the AI. Bias in AI can lead to poor decisions. It can erode trust between humans and machines that learn. How do we eliminate AI bias?
Jeremy Jawish This is something that is very important in the AI community, and in insurance fraud detection it's even more important. We're very transparent at Shift. We give real answers, not politically correct answers, so it's a real, real problem. Especially because we think it's important to have non bias in AI, and there is bias in fraud detection AI, because some teams would automatically think that, oh, because a person is from this nationality, there's more likely this person would fraud. Or, because this gender with this type of car would fraud more than this gender with the same type of car. This is how some people think, and this can introduce a bias in your algorithm. Your algorithm will learn how human behave in front of data.
It's a big, important topic. When we receive data from carriers, we remove nationalities, we remove location of birth, we remove gender. However, we took the liberty to do some tests in geographies where it's allowed to do it; in some Asian geographies. The good news is that ethnicity, gender, age is not a relevant feature in your algorithms. However, in some geographies, in some teams, they introduce this bias because they think it's a relevant feature. We are doing a lot of control and checks to be sure that our algorithm don't get biased like that.
Laura Drabik It is estimated there will be 30 billion IoT devices by the end of 2020. The global market value of that is projected to reach 7.1 trillion dollars by the end of this year. A common use case for sensors that our carriers employ is the home, to detect water leaks. Combined with AI, alerts can proactively be sent to the homeowner in order to mitigate loss. What are some of the other impactful areas that IoT, combined with AI, could positively impact our industry?
Jeremy Jawish Other than the usual connected cars that everybody think about, we, on top of that, have a lot of expectation from smart alarms, for instance. Any surveillance tools in homes for preventative of theft, for fires, we think that there is a big opportunity for insurers to do, like we said, prevention, or to react in real time, or to also detect fraud while using these surveillance alarms, smart alarms, smart connectors in homes.
Laura Drabik What critical advice would you share with carriers exploring the usage of AI in their claims department? What do they need to consider, and prepare for before implementing an AI strategy?
Jeremy Jawish What mistakes we're seeing in the insurance market is, when carriers say," I want to do AI because it's cool, because it opens a lot of doors," and try to find problems to apply AI. Usually, you ended up by trying to find the simplest way to apply things that are cool, and looks nice, but in reality doesn't produce a lot of results, so it's more fake problems. Ask yourself the right questions, and formulate the best way to problem. Then, see if this problem can be answered by AI. It maybe sounds like a small trick, or something not very important, but I can tell you to change the way you apply AI on some problems.
The hardest part we've seen is understanding, and realize that taking generic, global algorithm that can apply to anything, and just applying it to a huge set of data without thinking about what makes sense in terms of insurance, is something that usually fails. It's very important to realize that, whatever problem you take in insurance, you ask yourself how to tackle this problem. Then, is AI a solution?` If it's AI solution, how can I make it more insurance-specific? This is the part that is not easy. Every time we went up with other companies saying, Oh, we want to apply AI, we do on e-commerce for detection to insurance fraud detection," and they realized that there's a huge gap, and a lot of effort and energy to be sure that they managed to match AI with insurance expertise.
Laura Drabik Start with the business problem, and then work backwards. In fact, that's great advice for carriers to consider when they're looking at any Insurtech solution.
Just a reminder, if you're enjoying this podcast, be sure to subscribe to InsurTalk on Apple Podcast, Stitcher, or wherever you get your podcasts. Let's get back to our conversation with the CEO of Shift Technologies, Jeremy Jawish.
Jeremy, what will claims processing look like five years from now, when AI becomes more ubiquitous?
Jeremy Jawish It's something I spend my weekend thinking about, because I'm passionate about it. I really like comparison. An example I can give you is, looking at claims that would look alike in other industries. Lately, I test something, which is, I took a ride share before the crisis, and I realized that the driver took the wrong path and, by the way, changed the location, and changed the course, and ended up charging me more.
I reported a claim in the app, and it's amazing because, automatically, just on the bottom it said, "Claim: Driver changed the course." It automatically looked at what was the initial location. It automatically computed the difference of the route between the first address and second, and how much longer was the second was the difference of payments, and automatically refunded me with the difference. It was just a click. It did the whole screening. It looked at a lot of external data. It looked at the history of my rides, and automatically paid.
I think insurance would be the same, even without asking it. Sensors, and IoT will detect it, automatically send a signal to the insurance, will automatically take care of it, and even send someone to repair it in real time. Why not think also about having everything taken care of, even if you're not here? You come back home, and everything is repaired before even know it got broken.
Laura Drabik I really love that idea. If I were on vacation and my house got broken into, how wonderful would that be to come back to a home and not even notice that there was a loss? That's really insightful. How has AI health maintain, or improve claims handling in the wake of COVID-19?
Jeremy Jawish We were with carriers on the frontline COVID-19, and we really have seen, other the fact that, of course, claims dropped significantly within a big way in geography that got impacted by COVID. The behavior we've seen is that, because of COVID, a big crisis hit financially these geographies, and we've seen a very quick new trend of fraud that are because of COVID in geographies that are getting out of COVID.
We're seen a big, big increase in numbers of fires, for instance. From small businesses, we've seen a lot of claims that are death claims from stocks, from any assets they have. Also have seen in auto body shops, and workshop, really being very aggressive on fake claims, fake invoices. We really needed to react very fast, because it's things that are being done across the board, being very creative, being very adaptive to the fact that the crisis hit very hard, all the businesses. We couldn't react as fast as we're doing right now if we didn't have this AI platform.
Laura Drabik Jeremy, could you share one of those interesting fraud use cases with us?
Jeremy Jawish When you look at different geographies we're in, and when you look at the fact that the lockdown, and the crisis, they're all getting out of it, and not at the same cadence, not the same time, but at least we're seeing the same type of exit. You can see that a few weeks after everyone is getting out of the lockdown, small businesses go back to operating. The typical example is where a small business goes back, like a restaurant, and the restaurant realize that, even getting out of the lockdown, customers are hesitant going to restaurants the first weeks. These restaurants realize that it's going to be a very tough year.
One of the things that they would see for the ones that are really aggressive on fraud would be to set fire to the kitchen. We've seen a big increase in numbers where they would set the fire to the kitchen, because there was a lot of food stocks that they would not be able to sell. They realize that customers are going back very hesitantly to restaurants, and the easiest way, and the best way to get benefit of this crisis, and the fact that they have a big stock that could not be sold, is setting this on fire. The time is quite tricky, because you get a lot of fires. Also, some of them are not fraud. A lot of them are some people going back to the restaurant, and so that accidentally [inaudible 00:17:11] a fire, because nobody was there for more than a month or two.
Being able to separate the real ones, then the suspicious ones is something that you really need accurate algorithm that can look at a lot of data in real time in order to give answers. There's a big peak in claims going out of lock downs, and you need to really, in real time, be able to very quickly assess which one, where you want to dig in, and the ones that these guys had real claims and they need a real assistance, and help.
Laura Drabik Jeremy, thank you very much for your time today, and for your incredible insight into the positive impact AI will have on claims processing. You showed us, it's not just about ideas, it's about making ideas happen.
Jeremy Jawish Thanks a lot, Laura. I really enjoyed this podcast.