Episode 58: Full Transcript

[00:00:00.05] Bobbie Collies: The way technology is impacting our industry and how fast things are changing, we really need to start looking at where is the puck going. If I'm going to steal from Gretzky here, let's start to skate where the puck is going, not where it is today.

[00:00:14.73] Christen Kelly: Welcome to the Insurance Technology Podcast. In this episode, we're going to be focusing on some of our past guests who really want to impact the future of insurance. This first clip, we're going to talk about something that is hot within the industry right now - artificial intelligence.

[00:00:33.69] Elad Tsur: I always knew I'm going to bridge my passion. The AI hold no practically changing domain to the not practically changing domain, slow-moving world of insurance. Joined hands with few experts from insurance and other team from the previous startup experts on AI to bridge that gap between the AI and insurance.

[00:01:05.70] We've had a thesis. There will be a core platform of five layers of AI, and on top of it, many different insurance use cases. Those five layers I mention them in very, very briefly. You have the sources layer, the first layer, which is the data layer, crawling the web, getting raw data points, matching it with the business, images, videos, audio files, maps, governmental databases, social reviews, website texts, rating site, et cetera, all of that, one layer, and you need lots of AI for that.

[00:01:39.57] Crawling is simple. Crawling the right stuff and matching it, extracting and matching it with the business, understanding that the picture has the business in it, speaks about the business, shows the business. This is not trivial.

[00:01:54.19] Second layer out of the five layers is what we call the knowledge layer. There, we're extracting information about the business from the raw data. Think about a selfie that someone took in a bar besides the beautiful person at the center of the image. In the background, you have sprinklers in the ceiling, you have the material of the floor, the width of the aisles.

[00:02:15.42] You know how many bartenders, what type of atmosphere is there, the ratios between beer bottles and wine glasses, to dishes and plates on the table. Everything that we have been requested to do or that we believe to be correlated with something the carriers are looking for. For example, when we've opened the company, we weren't looking at littered exit signs in those images. Now we do because carriers asked for it. Whether the exit sign above the emergency exit door is littered or not, what's the intensity of the light in that place?

[00:02:44.82] Now, we're looking at all of those and more. That's the second layer. As carriers, don't yet care about that intermediate step, about that knowledge. They care about underwriting insights. That's the third layer of AI, which we call the wisdom layer. We have the data, the knowledge, now the wisdom, it's the underwriting insights themselves.

[00:03:09.16] We're using machine learning models to predict the underwriting insights based on all of those intermediate insights that we've created out of the raw data. It's a gold in, gold out concept. We're taking let's say that we have the percentage of liquor revenue out of the total turnover of a place as a training data set.

[00:03:31.63] Reid Holzworth: Orlando is in the wisdom layer, if you will. That's where you're bringing in third party data sources. The carriers directly saying, hey, these types of customers, we have these types of losses, blah, blah, blah, bringing in all the claims data, things like that. It learns basically whatever you want it to learn.

[00:03:51.48] Elad Tsur: Exactly. On that layer, every layer of AI you need the training data set. Either we get it from the carriers from our customers or we're building it ourselves. We can do both. In this example they get, we've built it ourselves. We're able to get an audited set of bars and restaurants in this example with their percentage of liquor turnover out of the total turnover, how much revenue they get from liquor, from alcoholic beverages.

[00:04:17.71] Let's say that we have it for 2,000 businesses. Then you can train an AI model to predict it for those 2,0000 businesses, and then extrapolate it to any other probe. You give it to any other business out there. We have that for hundreds, actually today, more than thousands underwriting insights. That's the wisdom there.

[00:04:37.75] The fourth layer is what we call the decisions layer. At that point, we're training AI models to mimic the underwriter. Think about your AI underwriter, pair underwriting. In coding, you have pair programming, like two coders sitting and coding together. You pair AI underwriter. Looking at stuff way before you look because it's much faster, and make a decision like you would make on 80% of the intake, on 80% of the cases. Because the anomalies are closer to art and you do want those anomalies to be reviewed by very experienced underwriters.

[00:05:19.05] We're capturing those decisions and being able to mimic them in 80% of the cases. If you're trying to build rules to make that decision, let's say, I don't want full service restaurants, I do want catering services. And then you have a catering service which has a full service restaurant in it, but small one. Is it in or out? It can be captured by a simple rule. You need to train an AI model to find [INAUDIBLE] those cases and mimic that. That's the decisions there.

[00:05:55.32] The fifth layer of AI is the actions layer. Study this data into knowledge, into wisdom, into decisions, now actions. Example, you're missing an information, let's say, lost funds, or missing the revenues of the business in the input.

[00:06:12.03] Automatically, send an email to the agent, so the broker. I'm missing the revenues of that business. Use ChatGPT for that. Using ChatGPT automatically to the agent. I'm missing the revenues, can you please provide it for that business? Use ChatGPT to parse what you just received from the agents and continue the process.

[00:06:34.36] That fifth layer, the actions layer, using AI to just save you mundane tasks that you would do yourself is the fifth layer of AI platform. We're building that from scratch from the beginning, that way with those five layers, and started building the insurance offerings on top of that platform, such as stress reprocessing, such as lead generation, which you and I talked about in our last meeting.

[00:07:01.22] Generating lists of leads that already comply with all of your eligibility criteria to be sent to the distribution. Before writing, you will promise 100% submission to quote in this case. All of those solutions are being built on top of that five layers of AI platform that we built.

[00:07:22.96] Christen Kelly: I really like the way that Elad defines artificial intelligence, and makes it easy to understand its impact and what he and Planck are doing in the industry. It'll be exciting to see what they do in the future. With our next segment, we switch gears from a technology startup to an investor lens and what companies they see as being successful, and ones they're willing to put their money behind in the future.

[00:07:49.75] Jesse Wedler: I think that there's so many point solutions that are interesting, but I think the biggest question is the broader AI question. I think the reason why this is relevant, granted it's totally overhyped right now, we're in the pink hype cycle. But I think the reason why it's interesting and relevant for insurance is, first of all, in terms of the legacy code migration and just maintaining legacy systems, there's a lot of engineering efficiency that comes from AI assisting the engineering teams.

[00:08:16.01] We talk to teams, mostly outside of insurance, I haven't seen this in insurance. But teams outside of insurance where tasks that used to take an engineer two weeks to do. They can now do it in three hours using some copilot or assistance from AI.

[00:08:31.13] Wow, that can have a huge impact in the back end where customers and the insurers won't really see it. But then you started thinking about the front end use cases, and the way people interact with their policy, the way people interact with their carrier, with their broker. There's a lot of just new opportunities for customer engagement that we haven't even started to scratch the surface on.

[00:08:51.65] You can start thinking about underwriting, you start thinking about claims, talking a couple early stage companies where you're still very early, seed stage, where they're building an underwriting assistant. Hey, let me take this corpus of data that you just got for some commercial policy. I'll synthesize it through a large language model. Depending on the parameters you care about, I'll search for these words. I'll tell you what is a potential red flag in this business.

[00:09:16.67] The person is still going to be human underwritten, but it's going to be AI assisted. I think that's an important thing. As teams are thinking about how to apply AI to this space, carriers don't want AI to go replace the secret sauce of their business. Just like if you talk to the legal profession. There's a bunch of AI going after legal tech right now. Lawyers don't want to be told how to do their job, but they definitely want something that's going to make them a lot more efficient in finding that needle in the haystack and going through a bunch of information.

[00:09:45.41] I think that's where there's a bunch of opportunity, but it'll for sure create a new hype cycle and a wave of craziness and funding probably within Insurtech. I think the question is, what are the models that actually have long term staying power where there's just clear ROI?

[00:10:03.11] Reid Holzworth: No, absolutely. This whole AI thing, man, is really interesting. It is like so overhyped right now, but for good reason. I think it's funny. I had a buddy who called me yesterday. Freaking guy's blowing me up. He's like, dude, I got the best business idea ever. Trace commas. Trace commas. He's saying all this. It's funny. If you know that, that's [INAUDIBLE].

[00:10:28.43] Jesse Wedler: From Silicon Valley.

[00:10:29.48] Reid Holzworth: Silicone Valley.

[00:10:30.71] Jesse Wedler: On the AI thing, I think it is super interesting because it's an equal playing field where everyone's starting from the same spot, and no one's figured it out right now. I think if you're an incumbent, you've got to go experiment with it. You got to figure out what the use cases are so you're not flat footed with it. But I also think it's so interesting with the Silicon Valley lens, all these early stage AI investors are like, aren't you worried that AI, people are going to build and replicate businesses like Applied or other great software businesses?

[00:11:01.16] No. Maybe you'd be able to replicate interfaces quickly using that to replicate the domain expertise and the connectivity. I don't worry about disruption from AI. I think it's really just a question like, what are the use cases that are actually going to work in this industry and who should be the companies to build and own that?

[00:11:22.45] I'm sure there'll be some really exciting new upstarts. I think distribution will be the key question for them back to what you were saying because whether you're selling carriers or agents, you got to stand up in big sales team. That's where the incumbents are going to have a lot of advantage and so everyone's trying to figure it out.

[00:11:41.95] Reid Holzworth: Stand up a big sales team and where everybody fails going after the incumbents is migrating all the data and migrating everything that that business has today. All their workflows, everything that they do. Everybody thinks if I just bring out to the street this new shiny thing and it's cool and it's and it's smart and it's better than whatever, it's a long road to get people to actually move over on that.

[00:12:10.77] I like what you said, though. You're right. It's a level playing field. Like I said to my buddy, it's like, dude, everybody's after this right now because it's not super expensive to go spin up a team to go get after this. It's like when data started to become a big thing. It's always been a thing. But it's gotten bigger. Building a big data science team is no joke, man. But this stuff and we're taking existing engineers really smart people. They are, dude, they know it. They're getting after it totally.

[00:12:44.94] The big incumbents, they have resources to do so and their customers are demanding an answer on this. They're doing it too because they're a little bit afraid like, Oh, what does this mean for us? We got to start figuring this out and have an answer for it.

[00:12:59.73] Jesse Wedler: It's a mix of fear and opportunity, which I think is healthy. But it's going to be a fun time to see what actually percolates, and ends up being good use cases.

[00:13:10.17] Christen Kelly: For this last segment, we shift gears again to someone who is in what would be considered a traditional piece of insurance - the carrier. But how do they need to change in order to stay successful?

[00:13:23.82] Bobbie Collies: The way technology is impacting our industry and how fast things are changing, we really need to start looking at where is the puck going. If I'm going to steal from Gretzky here, let's start to skate where the puck is going, not where it is today. Just painting that picture of-- you talked to Allen and Mike about APIs and having that be the building block of our industry moving forward. I'm trying to paint that picture for both agents and carriers alike just to see what's going to be over the horizon. We're going to look a lot different five years from now, a lot different five years.

[00:13:59.54] Reid Holzworth: It's so true and I truly believe that, and I've said this a lot within these podcasts. We are really going through a transition within our industry in a lot of ways. It's so true. With the way that technology is just growing and it's being implemented, and just technology, in general, how easy it is to build things these days with all these various platforms that do different things, low code, no code, blah, blah, blah, blah, blah. But just really many, many, many people being involved in modern technology within our industry is pushing it.

[00:14:36.36] If you look back at 2015, like, what? I remember in 2015 and 2014 literally explaining to people, training people on what Salesforce was. They had no idea. They're like, wait, what is this thing? I don't know anything about this. That was 2015. Not that long ago. Salesforce has been around for a long time. The API, they don't even know what it stands for. I mean, a lot of people still don't. They have no idea. Now, this stuff's really starting to happen. What are your thoughts around all that?

[00:15:06.77] Bobbie Collies: I read a couple of things. I think to your point, more folks in our industry are trying to educate themselves and understand it because it's become a buzzword and something that's become more real versus visionary. The other piece, too, and I feel like a broken record when I say this, but the pandemic really forced our industry into being more digital.

[00:15:35.14] I also think it accelerated the rate at which customers want to interact with every single thing that they do from a digital perspective. As we look at changing consumer expectations, they just want to do things differently. I looked at getting a mortgage not too long ago and went on better. I was able to not only get a pre-approval letter within 15 minutes, but I also got five insurance quotes from different companies on this house that I might buy some time.

[00:16:03.58] It wasn't even a retail agent involved in that situation. And that's what our industry, and I truly believe in the AI channel, but we have to catch up. As a non-insurance person going through that better process, I wouldn't have known what kind of coverages I needed, which is why I believe in having the agents involved. But they need to insert themselves into these digital purchasing processes in order to remain relevant moving forward, especially on the stuff that's more commoditized, like personal lines in some small commercial.

[00:16:37.38] I define Insurtech as where does technology intersect with the insurance industry in ways that are going to improve our efficiencies and effectiveness moving forward and furthermore impact the customer and agency experience in very positive ways. Because we've had technology and insurance for a really long time, like green screens and entering it like that. That was technology we used to do our business. But when we really think about Insurtech, to me, it's what are the technology pieces that are going to revolutionize the way we do our work.

[00:17:15.98] Christen Kelly: There you have it. Three different perspectives of what the future holds for Insurtech. As with all of our best ofs, if you want to hear the full episodes, simply go to insurtechpod.com and search through our episodes. Reid and I will be back soon with new guests. We hope that you continue to join us for the Insurance Technology Podcast.