Leveraging the Power of Automation for CPA Firms
00:00:00 Reid Colson: Welcome everyone to our LinkedIn Live event today where we’re going to be exploring AI in the insurance industry. We’ve got over 75 of you that registered today. So it’s pretty clear to us that AI is a topic of interest as the insurance landscape evolves. AI is rapidly changing the way folks are underwriting, assessing risk, personalizing experiences and streamlining operations. And in this session today, we’ll hear from two industry experts to get their perspectives on the application of AI and insurance. And I’m sure knowing these two, it will be a lively discussion. I’m Reid Colson, I’m facilitating the conversation today.
00:00:51 Reid Colson: Now let’s meet our panelists. I’ll begin with Rob if you could introduce yourself and then let’s turn it over to Lloyd.
00:01:15 Rob Reynolds: Sure, thanks, Reid. Yeah, my name is Rob Reynolds. I lead data and analytics here at WR Berkeley. Berkeley is a large multinational insurer, a P&C insurer. I’ve been here for a while, been in insurance for a long time and just happy to have an opportunity to talk about something as interesting as this. Lloyd, take it away.
00:01:36 Lloyd Schulz: All right. Thank you, Rob. Good day, everybody. My name is Lloyd Schultz. I’m the chief technology officer at Markel. Markel is a Fortune 500 specialty commercial carrier based here in Richmond, Virginia with global operations. We operate everywhere from Vancouver to Singapore and recently opened three offices in Australia, which will speak to some of the complexities that we’ll talk to when we start speaking about data and AI. I’ve been with Markel for about eight years. I spent my entire career in IT, most of that time as a data engineer. So I bring a more technical bent to the conversation and I’m looking forward to getting into some of the questions today. Back to you, Reid.
00:02:41 Reid Colson: Yeah, on behalf of the audience here. Thank you both for your time. I think they’re going to be very interested in what you have to say. I know you both have had long careers in and around this space and have a lot of experiences to share. So I’m really going to start with some general observations. Right. Again, you guys have seen all sorts of hype cycles around things. With respect to data, AI feels initially like it was going down the same path, but now it sort of feels a little different. It’s in the news everywhere. It’s LLMs to coding assistance, robots, self-driving cars. There are companies making tremendous headway with AI, others are beginning to experiment and some are still at the starting gate. So I’d love to get your perspective on what is and isn’t working with respect to AI and insurance. So my first question would be to each of you. Where do you see insurance companies getting it right when it comes to using their data to develop or deploy AI solutions? Maybe we’ll start with Lloyd since I let Rob go first on the intros.
00:03:34 Lloyd Schulz: When I think about AI, I actually think about a three-part Venn diagram – your core data, the AI aspects of how you leverage that data, and automation…I think if you focus too myopically on one area without considering the others, you’re going to be just that – you’re going to be near-sighted and you’re not going to get the full value of all the data you have available to you…But from my perspective, when it comes to how do we get it right? Where do you start? This is going to sound like an age-old adage, but it really comes with thinking about the value that you can derive, that you can give back to the business in terms of whether it’s generating revenue or reducing expense. Be very specific early about the use cases that you go after and be really conscious not to try to boil the ocean. There’s a lot of hype right now around generative AI specifically. Many of us have been using AI and machine learning for many years and just over the last couple of years, generative AI has really sprung up. So I would say start small, focus on delivering specific business value. And the last component is making sure you’ve got business buy-in from all levels of the organization, whether from starting at the executives, all the way down to the performer levels. If data and AI are seen as something that another group does or that centralized data team does, you’re not going to get the full benefit and value or the support that you need to make it successful. So I touched on a number of things there but start small, think about the value that you’re adding and make sure you’ve got business buy-in to me. Those are really the critical elements that will let you be successful as you start to deploy a new solution.
00:04:57 Reid Colson: Yeah, some great thoughts. Rob, what are your thoughts on where you see insurance companies getting it right?
00:05:04 Rob Reynolds: Yeah, I mean, I still feel like insurance companies have a lot of opportunity to just leverage the data that’s available to them. Forget about AI for a second, right? Like there’s still lots and lots and lots of opportunity out there to make better use of what we have kind of just laying around an organization. But I think if you continue to keep your eye on leveraging the data you’ve got available and then you incorporate AI as an additional tool in support of that process and data, I think you’ll see a compounding effect from those two things coming together. I don’t think we’ve completely tapped out the opportunity that data offers our organizations in this industry anyway yet. And I think the other thing that I would say is AI, while a huge opportunity, I think it does have an opportunity also to be a bit of a distraction if we’re not careful. So I think there is something to be said for trying to craft what it is you’re going after and go after it with a real focus and desire to get the results and make sure that you’re not just running in 1000 different directions all at once because I do think there is some risk that this becomes a bit of a distraction. But generally I think it’s going to be hugely helpful.
00:07:36 Reid Colson: Maybe I’d ask, what do you see in some early wins in terms of use cases that you’ve either seen locally or seen from your peers?
00:07:57 Rob Reynolds: Yeah, I mean, we just were talking before we started right about any instance where you have a laborious task that you can make easier from the use of AI sorting through files…where they haven’t been categorized or they’re not appropriately documented, they’re not appropriately tagged so that you can quickly reference them as part of a process or ancillary to a process. Those are all use cases that offer a ton of opportunity. And I think applying AI is going to be about solving really specific problems within a process or use case…If you can figure out how to say, hey, I normally would page through this 100 page document to try to elicit out of it some piece of information or some insight into a risk or a claim or what have you, you can use AI to do that much more quickly. That’s an enabler that I think can be applied to so many different facets of our business.
00:09:37 Lloyd Schulz: Yeah, I like what Rob said there. So one of the things that we’ve been saying internally is AI is just another tool in our toolbelt. It’s not necessarily going to solve all the problems and we have to focus on the data, right? You’ll get garbage in garbage out. If you don’t have good data, you won’t get good outcomes…From a use case perspective, very similar to what Rob said, we’ve kind of been bifurcating it into two categories…The horizontal things are those operational efficiency gains – hey, can I save each associate in the organization a couple hours a week and that adds up over time and there’s meaningful operational savings…So that’s kind of the horizontal. We consider that a little bit of the low hanging fruit…The more vertical use cases, that’s where you go from kind of the walking into running…Things like intelligent document processing for submissions – can we scan submissions and look for criteria that matches our appetite, can we look at reinsurance exclusions on reinsurance contracts and see if incoming new business might be immediately excluded or included within our appetite, things like that…Within Markel, we’re trying to balance across both of those – can we start with the crawling operational efficiency? Can we look at a series of use cases where we get similar solutions for multiple problems and implement more of those vertical solutions that I mentioned?
00:12:15 Reid Colson: You both alluded to this a little bit. The places where there are challenges or pitfalls or things that maybe companies are not progressing, you talked about maybe not having good data and not having good use cases. Quickly, what do you guys see as the key challenges that are preventing insurance companies from really leveraging their data well and taking advantage of AI?
00:12:53 Rob Reynolds: Yeah, I mean, I think there are a lot of challenges, right? I think there’s the challenges to change that exist for almost anything that would apply to data. And while I think AI is exciting, I think it’s also at times could be threatening. And I think you see that in the broader marketplace, not just insurance, just in terms of how corporations are trying to figure out how to apply it. So I think from my perspective, you’ve got incumbent processes, incumbent biases and beliefs and ways of doing things that are going to be upset and that makes perfect sense, right? There’s going to be resistance to that, some of it valid as well. You know, I think there’s going to be a lot of questions out there that companies certainly in insurance where you’re heavily regulated, there’s going to be a lot of questions around well we could do this, but should we do that? Right. And those kinds of questions are important to ask and make sure you’re applying the right kind of rigor to that decision for your organization to ensure that the outcome you get is what you’d want for your organization, your partners and your customers and shareholders. And that can be complicated with a new technology like this that is so powerful and could be profoundly transformative. So I do think there’s going to be some barriers along the way and there’s stuff to figure out that is not really dissimilar but does apply probably more heavily applied to AI just because the nature of the change is so remarkable potentially.
00:14:20 Lloyd Schulz: Yeah, I’ll add on to that a little bit. There’s a number of different dimensions here, right? Some of them are your traditional data challenges – legacy technology, data silos, any of those things that we’ve been grappling with in many cases for years apply in equal measure to the use of AI. When I said before, if you don’t have good data, good access to data, it’s going to be hard to drive incremental value from the use of AI and generative AI…But when you move beyond that, data governance and data quality is very, very important. Making sure you understand where your data comes from, how it’s being used, how do you do data definitions? I kind of mentioned at the lead in, Markel is a very complex industry, right? We often operate globally, we probably have 300 products that we sell across the globe, all very different. So when you start to think about things like, well, how do I use the data? How do I manage it? That data quality and data governance is critical to any success that you’re going to drive…I would say above and beyond that, when we start thinking specifically about generative AI, it’s not just what could you do, it’s what should you do. It’s the responsible use of AI, it’s the governance, the thinking about how it can benefit your customers as well as your employees and your shareholders and making sure you’re doing it in a responsive way. Again, there’s a lot of hype here. There’s some scare around the use of generative AI, especially when it comes to voice and video. But if we think about how we can use that in a productive way, the what not just what could we do, but what should we do? That’s an important element as well. And then lastly, I think just the talent angle as well. This is definitely a new tool in the tool belt. It’s a change in the paradigm for how we can interact with technology and data. And we have to really be thoughtful about how we bring our staff along and give them the skills they need to be successful in this new arena.
00:16:40 Reid Colson: I’m going to shift gears a little bit and let’s look ahead…If we look ahead and you think specifically about AI and insurance, how do you see over the next 3 to 5 years AI impacting insurance generally and maybe specifically in certain areas?
00:18:03 Lloyd Schulz: AI is going to be pervasive. It’s going to show up in many different ways and it can be hard to predict how it’s going to happen over the next 3 to 5 years. But I think we have some inclination of how things are evolving…I think as we look at the way AI can play in that bigger tool chain of automation and improved use of data, I think there’s some kind of obvious use cases that are showing up that we’re already starting to experiment with and actually roll out full production solutions…It’s the understanding, we work in a very complex industry. I mentioned hundreds of different products and so many different forms of submissions that we can take in when somebody’s applying for new insurance coverage. So using AI to help interrogate and summarize and create consistent views of documents and inputs is really critical because that’s one of the challenging parts of the process…Things like intelligent document processing and then using that automation focus to then do appropriate routing to your appropriate other downstream systems or users, underwriters or claims handlers, what have you. So I think there’s definitely something around helping us understand and ingest documents more effectively…Another critical set of use cases that we’ve seen evolving is really around enhanced search and summarize which generative AI is really good at. And there’s a lot of tools and capabilities in the industry that can help us with that where we need to pour over a series of legacy documents or in many cases, a large number of documents and ferret out information about what’s the appropriate historical policy that may match a new submission that’s come in. How do we summarize that information to make it more readily usable by the underwriters or by the claims handlers? So there’s things that we can do to gain efficiencies along the way I mentioned before, matching our appetite, looking for exclusions, things like that are kind of clear early use cases that are building a lot of momentum.
00:19:42 Rob Reynolds: Yeah, and I agree with all of those. So rather than pile on, I think there’s also an element of what we’ll see in the marketplace that AI will create some dynamics that allow companies to be not only more efficient, but I think barriers to entry will potentially fall. And so I could see large companies doing business or competing with smaller companies as maybe aggressively as maybe they did larger companies. So if I have now the use of AI and I’ve removed a competitive advantage from one of the market leaders or something like that in some niche or what have you. How does an organization that was built around AI compete differently than maybe organizations that have it in terms of any number of different dimensions? So there’s, I think there’s that element that will be interesting to see how that plays out. I do think that speaks to, I think as an industry, we have to get more comfortable with the idea of being flexible and fast personally. And so that first point is a reason to be more flexible and fast, another tactical kind of reason to be flexible and fast other than it’s good to be flexible and fast is I don’t think the use cases are completely nailed down yet in terms of what AI will enable. So I think we are still figuring things out in that space. So we have to be able to be flexible in terms of how we pivot to one technology or another.
00:21:58 Reid Colson: So with that, maybe what I’ll ask is the last question for you guys and then we’ll see if we have anything from our audience. What’s your advice? What do insurance companies need to be doing now to prepare for the changes that AI is going to bring? How are they going to be successful in this new environment?
00:22:36 Lloyd Schulz: These are probably tried and true – AI is just another tool that’s available to us. So you could probably apply my thinking here across a number of dimensions. But I think in general, you have to invest in talent. I think there’s a talent war, we all recognize that you have to invest in your people. You have to grow them, give them the skills and opportunities they need to be successful so they can feel that you’re investing in them but you’re also investing obviously in the growth of the company and the ability to use these new tools effectively. I absolutely agree with something Rob said a minute ago and it sounds like a tagline, but I think we all have to foster that culture of innovation and be willing to test and learn, move quickly. I think Rob maybe used the word agility, not everything is going to work and you don’t have to bet the farm on every idea. You have to go try it out, see if it works, see if you can, then if it does, how can you multiply, how can you grow those things? So you have to have a culture of innovation, you have to invest in your talent…And honestly just stay informed. This is a marketplace that’s changing very, very quickly. New models are coming out every day. Every vendor is infusing generative AI into the tools that they offer. So having that market scan capability to understand what’s already available to you with the tools that you have, what’s coming that might enhance your experience or be able to deliver better business value. The market will continue to change very, very rapidly. I think we’re still in early days here. So having an eye on those changes will be very important.
00:24:10 Rob Reynolds: Yeah, and I think as well, we’re, this is really early days and that’s why I was kind of giving this some thought, right? Is I don’t know if we are completely appreciative yet of how AI is going to transform what we do or how we do it. So I think there’s something to be said like as we look into the future and as we kind of plan, we have to almost acknowledge the fact that what we do today will not be what we’re doing two years from now would be my guess, right? Like that may not be right. But I suspect that the pace of change in this space will continue to move so rapidly that so too will the evolution of thinking and application of the technology to what we do. I think the other piece is we have to be thoughtful. We run the risk of being myopically focused on the outcome that I’m trying to achieve for my function, my area, my line of business, my department, my segment, whatever it is. And we have to think not only like, hey, this kind of goes back to what Lloyd and I were talking about – it’s like, you know, should we versus can we. Yeah, we can do this thing. If we do that thing, it might be good for me, it might be good most immediately in my future. But what are the long term effects of that? Like what does that maybe mean to my business if I were to apply this technology or this approach or this use case in a very wide sort of way, not always are those two things completely in concert. So like if I make a solution that targets a certain problem that is kind of right in front of me today, that solution may look different in the long term if you scale it across an organization and I may have unintended consequences that I didn’t anticipate if I just looked at it through my lens. So I think there’s an element of be mindful of what could be coming, be mindful of what’s beyond your purview and think about the implications of the solutions that you’re trying to implement in your organization and revisit it from time to time, be thoughtful about actually forcing yourself to say, look, I thought this was a good idea yesterday. Today I might feel somewhat differently, just doing that check in either on an individualized basis or strategically as an organization I think is important and will become more so given the pace of change.
00:26:31 Reid Colson: Yeah, that’s great. I appreciate it guys. We’re coming up close to it here. I think maybe what I would summarize from what you both said, companies need to be ready with a few things to be successful right? There just needs to be a sound ethical policy for using AI. You need to have good data management practices in place, strong use cases, good executive support, you need to be flexible and experimental, revisit your ideas to validate that you’re still on the right track and making sure that you’re investing in talent…selfishly, what I would add to that is that getting there doesn’t have to be a solo journey and there’s lots of talented partners like Udig that can help accelerate your progress. It looks like we do have a couple of questions here. How do you guys think about the risks of an AI mediated process is categorically different than a fully human process? Is there liability from a bad decision of AI treated differently than it is with a human?
00:27:45 Rob Reynolds: I’ll give you my two cents. I think it depends on how you phrase that. I think errors are errors regardless of who makes them or how they’re made. However, I would also say the scale of error is different on a human basis versus potentially an AI basis. So, if you systematize, I don’t know, that’s probably not a word. But if you systematically make an error and AI is making that error across a large base, your scope of impact is much, much broader. So I would actually, I think that’s the can versus should consideration here. I also think that specifically in the insurance industry don’t kid yourself for a second that there aren’t people out there looking to capitalize on those errors and actively waiting for that to occur. And that’s just the nature of business, frankly, not just insurance. So those would be my thoughts around that topic anyway.
00:28:41 Reid Colson: And maybe the next one is good for Lloyd. It’s a little bit more technical. How do you see carriers handling external data requirements and ownership in order to build solid depth and breadth for AI use cases including generative AI? So really, it’s kind of about the third party data ownership and requirements.
00:29:10 Lloyd Schulz: Yeah, that is a very good question. So for us, I think, and we didn’t mention it earlier in the conversation, but I think having access to external data, third party data and having the appropriate infrastructure to be able to support that is a critical piece of the journey that can help you augment your AI use cases and I’ll answer this question and kind of touch a little bit on what Rob said before, everything that we’re doing and who knows how things evolve over the next 3 to 5 years to Rob’s prior point. But we consider everything human in the loop. We do not want to be turning AI loose on anything and allowing it to make decisions for itself. So we want to make sure that we’re using all the tools, data and capabilities that we have at our benefit to simplify and help the end user with whatever process they’re already executing upon. So I don’t know if I touched fully on the third party data question, but we kind of think about that as just part of the bigger puzzle.
00:29:51 Reid Colson: Maybe last question here that I’ve got from the audience is how would you define where you are in your AI journeys? You’re setting your vision, evaluating use cases, you’re engineering or you’re just accelerating and operating responsibly.
00:30:00 Rob Reynolds: Yeah, I mean, we’re early still. So we’re figuring out how to apply various tools and technology to use cases and the like we want to enable our business, we want to ensure that it is human in the loop. As Lloyd said, I think that’s the right way to start. And we’re trying to make sure that we’re doing so in a way that is pretty conservative and pretty controlled, but also taking advantage of the opportunity that it does present.
00:30:50 Lloyd Schulz: Yeah, relatively similar. I think that we’ve been at this for just over a year now. So we have made some headway here. I would say we are kind of between two and three, but it really just depends on the specific use case. We’ve gathered a series of use cases and try to summarize those into different categories. We have been engineering for some time and building those things that we think are showing value. And then starting to look for opportunities to scale. It really just depends on the different aspects of the tooling or the use case. But to Rob’s point, I think for all of us, whether it’s the insurance industry or more broadly, this is still a new paradigm. We’re still early days. Nobody’s too far behind yet and it’s a good time to get involved, learn and figure out how you can apply these solutions to help within your own personal context or that of your company.
00:31:33 Reid Colson: Gotcha. Well, thanks so much again. I want to thank you guys for your time. We really appreciate you taking time out of your day to share your thoughts. I think it’s been very insightful and I’ve enjoyed the conversation. It’s good seeing you guys again. We had, it looks like over 125 people actually viewing the event. So quite a bit more than we actually even expected yesterday. And for folks who joined us, if anybody’s interested in going deeper learning more, you can feel free to reach out to me via LinkedIn or you can hit our website at udig.com. And you know, thanks again for your time and Rob Lloyd, really appreciate it. Thanks so much.
00:31:55 Rob Reynolds: Thank you guys. Thank you. Take care. Bye everybody.