DayCast Transcript Episode 1: Alan Demers Interview on the use of AI in Insurance
Updated: May 24, 2022
In this episode we discuss how Artificial Intelligence is being leveraged in claims and underwriting for insurers today. We dive into the advantages of using AI, the issues in deployment of AI technology and the people impact from using AI.
Announcer: Welcome to DayCast where we discuss how AI is really used in business. No hype discussions with real AI users. Here is your host Corey Dayhuff:
Corey: Welcome to DayCast AI, where we explore how AI is really being used in business. I’m Corey Dayhuff president of the Dayhuff group. I’d like to welcome to our very first podcast Alan Demers president and founder of InsureTech. Alan is an advisor for several insurance technology startups and was recently on the claims advisory board for Lexis Nexis. Alan Has had a 30 plus year career in insurance with 24 years at nationwide insurance where he was VP of P&C claims, for personal, commercial agribusiness and specialty lines and most recently VP of claims innovation & technology. Welcome Alan.
Alan: Hey thanks Corey, I’m glad to be here with you today.
Corey: Absolutely, on a rainy Columbus morning so maybe we just start with how you got involved with AI to begin with.
Alan: Yes, so my former role was probably the first time I was involved with learning about AI among other things. So as VP for claims innovation my job was to figure out how we could create an innovation culture. We were also focused on what we called claims of the future. That work was really focused on trying to find more efficient ways to operate. We really focused on how to improve the customer journey and make efficiency for the company. And so, a couple themes of self service and automation emerged. Certainly, looking at AI and anything that would facilitate that would be a good example. For self-services we looked at omni channel approach because we know our customer would still call, some would go online, some would text, some may want to chat. So those were areas we were looking at in addition to several others.
Corey: So, what did you learn from that AI using chatbot technology experience?
Alan: The area that focused on the bot technology was our marketing group, they were looking at ways that they could help improve the claim contact center. They recognized that people were calling in and identifying that they had questions and looking for more information. I jumped in to learn what could we do to help bring that further along. The bot technology was quite good with the idea that people would call in or text for info. They styled it around an FAQ model, for example if someone wanted to know about their deductible, the answer that the bot provided was a definition of deductible. I looked at that and said that’s probably not what customers want. It may be more like ‘how much is it,’ ‘why are you charging me a deductible.’ It’s usually more involved than the definition. That was the downside of how it was applied, the upside was that we learned a lot from that experience because I asked that we keep the bot turned on so we can learn what customers are chatting about. It was amazing. People were asking all kinds of questions, ‘what is the status of my claim,’ etc. It gave us a lot of knowledge that people were comfortable using chat technology, but the bot wasn’t capable of giving the answers they were looking for.
Corey: It wasn’t necessarily that the bot didn’t work, it was the practical application and the functionality wasn’t broad enough to satisfy what the customers were coming to it for.
Alan: Exactly, it was the way it was implemented or the lack of creativity, or truly understanding what the customers want and need. If using a bot where you’re expecting consumer adoption you need to understand what they want and deliver, vs. throw the technology out there and assume the best. You learn from your failures and you learn from your successes. So, we took the best of both from this one and learned quite a bit.
Corey: Interesting. Were you able to quantify some of the savings you received from the service desk?
Alan: We didn’t show actual savings, but we did quantify what the impact could be. That’s one of the elements of any technology is can you create a good business case. In this instance were able to identify over 2 million calls were coming into the call center annually. About half of which were just asking information, from all sources, policy holders, claimants, medical providers, sometimes looking for something simple as a claim number. We could certainly quantify the reduction of those calls for a live service representative would do, so the business case was there.
Corey: One of the promises of assistants is by taking away some of those shorter questions you allow for more time to tackle harder questions that come in, give a better experience to customers that may have a bigger problem, than what’s the status of my claim. Were you able to see the conversations elongate on more problematic calls?
Alan: Yes, I think that’s the upside and the promise that the technology can give you. Instead of spending a lot of time and giving a high cost and high touch experience, especially when the consumer on the other end just needs information. I don’t want to listen to an IVR and go through a menu. Could I use the bot to give me that vs. having to talk with someone, all nice things but certainly not needed in that example. And I would say CSRs able to deal with more complicated things. It certainly is the case where the CSR could spend the time where the human parts are needed, like if someone did just have an accident or a loss and they need that empathy. That’s much different than someone calling in needing simple information quick.
Corey: Sure, they are not under the pressure to answer stuff quite as rapidly because they don’t have the volume of calls. There are so many AI technologies out there, available was there criteria you used to pick which ones to evaluate?
Alan: First of all, I say AI in the insurance claims space is still a fairly new revolution. Some that were more practical would be like image recognition. The good news about that both large estimating software companies and vendors have invested a lot of time and effort to use image analytics and machine learning to create automated estimates so the idea is to capture photo in this case of vehicle damage and run that through big data to identify what the extent of damage could be in that scenario. That’s come a long way. That’s actually being utilized by most insurance companies today. You see companies advertising that you can snap a photo and get a quick estimate. I think the great part of this story is You also have consumer adoption. It’s easy for them to do. They see the benefit of getting a quick estimate. On the insurance side, it’s much more efficient. You don’t have to send an adjuster out to look, you capture the information early on and even though it has its short comings using imagery early on in the process makes things much more efficient. Takes out some of the steps
Corey: Sure, how that works in the back end is let’s say you’re driving a Toyota you get into a car accident, front end passenger side. It searches through its criteria to identify other photos of Toyotas that have had damage on the front passenger side, it knows how much you paid on that claim, compares it to that photo and now gives an estimate.
Alan: That’s exactly right, that’s how it works. That image lines up with others. And at first you look at it and say How could that possibly happen? You have cars that are designed and manufactured in a certain way, points of impact tend to be common and then you talk about high volumes. Using your example, say it’s a 2010 Toyota Camry, it’s been hit in the right front. Chances are there is a whole bunch of data on 2010 Toyota Camrys that have been hit in the right front that have similar damage. The machine learning part will identify does that bumper need to be replaced? does the bumper cover have to come off? and the other good news is the estimating process is well structured, part pricing, procedures on the amount of time required, that’s very structured. This makes a really good use case to produce that estimate much more efficiently and quicker. And even if it isn’t a complete estimate, it promises that various parts of the estimates will be pre-populated, so that an adjuster has to just eyeball by exception or complete the estimate, save time for both ends.
Corey: So that’s kind of embedded in the entire claims process. The users probably have no idea AI is being used on the back end.
Alan: That’s a great point. I think both consumers really don’t consider it AI, and I think if you ask most people in claims today they wouldn’t call it AI, even though there is image recognition and machine learning that’s being powered, it’s been commercialized to the point where it has taken out some of that mystery, which when we talk about artificial intelligence as a topic I think people struggle with it because what does that really mean, there is a fear of will machine replace people, business leaders like the upside of that, which is I can do more with less, it’s more efficient for my company but I also have the human side which is how do I replace people and take jobs away and then when you think about the context of handling claims another dilemma is how do I give that empathy and that human touch that I’m looking for if I’m using artificial intelligence. I think this gives us a good story that If you find a practical way to use it, it’s good for the customer, good for the insurance company it demystifies the whole discussion about AI and becomes about photo estimating.
Corey: What you think that means for what a claims department looks like, what the impact on the claims department is?
Alan: Before COVID there was already a theme toward doing things remotely and doing them more efficiently. Since COVID, it has really accelerated the ideas of virtual, touchless. Those are big themes right now in the insurance claims space. And for good reason because being able to have a virtual inspection is not only efficient, and it Could be a better customer experience but it’s also a cleaner safer way to do it, where you’re more contact free. The other theme which is touchless takes on a lot of different definitions talks about how you have consumer adoption. Maybe it has turned into a bit of a self-service model. So, anything that’s going to empower customers be able to do more, have more access and information and be part of the equation for their claim, those are really hot themes right now. And so, a lot it’s digital, although I wouldn’t discount what’s happening with voice because that’s right on the heels of everything that’s being digitized right now.
Corey: Do you think consumers are adopting that kind of technology more rapidly than they would have in the past?
Alan: Definitely. The stage has been set. We’re all in our homes now. Most people have adopted Zoom, and didn’t know it existed until a few months ago. I think right now you have consumers that are willing to do more, and kind of expect that. They’re having these fluid experiences from other favorite, online and digital and mobile experiences that they only expect in the claim. Now everything has it’s point of diminishing returns, or limitations because the claim process can be complicated. People for instance when they have a claim, they still want to call in. While insurance companies worked really hard to create digital loss intake and have launched their apps and online presence, well over 90% of customers still phone call their claim in. There is probably a lot of good reasons for that. I think the insurance companies condition them that way, they haven’t done enough to show them there are other channels. People feel most comfortable by doing that. I think there is some room to go in the adoption. Maybe the pursuit here is not digital, it’s more voice driven, and it can still be automated.
Corey: Sure, and will probably only accelerate. Who would have thought 10 years ago you would even use a chat bot to try to get a status on a claim. You had to talk to somebody, you had to talk to the adjuster. The acceleration and how we use that technology seems to only be gaining momentum.
Alan: It always comes down to how good it is and how well executed and how much thought has gone into the customer and what they really wat. Those are additional pursuits. If they’re not done well, there is nothing more frustrating than that. But if they’re done well the adoption rates can be terrific
Corey: What do you think is inhibiting insurance companies from adopting AI more quickly?
Alan: Part of it is mysterious, what is AI? Again, taking the photo estimating story aside, I still think there are a lot of questions, what does AI mean, how do I deploy it, what use cases make the most sense, do I have to change my whole organization and workflow and structure to make it work and get the most from it? One of the examples we worked on is RPA, robotic process automation, for an area where medical bills which were repetitive in nature and it seemed like a really good place to do this, after a few months of trial and error we found the processes were very different from region to region. We had unclear processes and workflows that had adapted over the years and been modified. So, unless you’re willing to spend time to clean up those processes, it’s hard to automate. Those are just some of the barriers that get in the way, but the good news is every year that goes by there is more awareness, there are more clear use cases. The appetite is still there because I think we can see the commercial benefits of AI. If we can do more for the customer, we can make the business more efficient. Those are two of the big goals that companies are focuses on.
Corey: Absolutely. I read where all state insurance is partnering with AWS to deliver an omnichannel, customer focused platform. It’s going to deliver things like Chat, text, email, voice all together for customer interactions. Do you think with a big player in the market doing that kind of thing, that’s going to cause the rest industry to move more rapidly?
Alan: Yes, that could really happen. Interestingly USAA is doing something similar with Google. Insurance companies have feared the idea that if Amazon were to jump in and be a competitor, and disrupt that whole distribution, sales area, that would really spell the end of work for the ages, so there is a bit of fear. But I think if you can work with a player like an Amazon, or a company that’s already figured out how to unlock this whole omnichannel experience and do it quite well, it would be a faster way. So, I do think other companies would jump on board. Today companies are faced with finding parts and pieces and trying to stitch them together and also trying to retrofit them into their way of doing business. So, if it does emerge that an AWS could help accelerate those things, you’d have to take a strong look at that. The other thing about insurance companies especially in the auto lines is they are very common in a lot of ways, and they do tend to follow each other. So, if one or two companies jump in that direction, I think others will follow suit. I do think that will likely happen.
Corey: Almost like any brand-new technology, when it first comes out you see a lot of custom development. You see some companies initially jump into it and try to custom build it themselves, and then almost immediately you start seeing companies like Guidewire start embedding those technologies into their systems. We have talked about photo estimation - It looks seamless, it doesn’t even look like AI, although you know in the back-end AI is being used. My guess is you’re going to see a lot of that over the next 2 or 3 years, embedded AI technologies into platforms that insurance companies are already using, and so it will natively happen just automatically.
Alan: Yes, that makes sense. You can already start to see evidence of that’s how the InsureTech market place is shaping up. I think we have a way to go. But if you look at the maturity cycle over the last few years its gone from endless numbers of InsureTechs and startups that are out there, but if you watch the funding cycles it’s become more pronounced the bigger and the more successful areas, and then you are seeing AWS, you’re seeing google. You’re even seeing Guidewire that was focused just a few years ago on systems deployment more involved with the InsureTech ecosystem. So just even 2 years ago there was a probably long line to get integration done with them. And now they’ve created a whole ecosystem around InsureTech. Salesforce, I understand has jumped into the mix and is looking at their solutions. It’s only natural that this next step is the bigger keep getting bigger, and keep connecting things. It’s probably better for the insurance industry instead of having to deal with a number of one-off and small entrepreneurs and trying to bring them to life, vs working with an established organization that can bring that to you. I think you will continue to see that happen in the marketplace.
Corey: That’s exciting stuff, that’s great. It’s coming a long way rapidly.
Alan: It really is. That’s the other thing to think about. Innovation for insurance companies is not something that’s been around a long time. Insurance companies haven’t been known for their research and development. It’s roughly a 5 to 10-year story at this stage. So, when you think of all that is happening and all that has happened in the last 5, 6, 7 years, it’s pretty amazing. I think that appetite is only going to continue especially now where companies are stepping back and saying large workforces they’re distributed; do I need as many people and do I need them in a facility and I think that’s just a further extension to how they think about the customers. I actually think we’ll see things accelerate in the next couple of years.
Corey: That’s all good stuff. Alan, I want to thank you for coming in. Great, great discussion. You are actually, officially our number one podcast ever.
Alan: I’m delighted to be here today. I’m happy to know that I’m number one. At least I hold that space until you have the next call. So, thanks Corey, it was great talking with you, some terrific questions, and hopefully we’ll do it again in the future.
Corey: Thanks for listening to DayCast where we talk about how AI is really being used in business. For more AI content go to dayhuffgroup.com. if you have a great story on how your company is using AI, and would like to be featured on DayCast send us a message from our website. Until next month keep delivering great AI capabilities.