John Hele & Greg Framke: Portage AI: The hard 20%: where AI could create the most value in insurance (420)

Welcome to this week's It's Tech podcast. I'm joined today by Greg Framke, who's the CEO of Portage and John Heal, who's the executive chairman. Welcome both. Thank you, Robin. Appreciate you having us. We've been doing this podcast for about 10 years and I'm accustomed to seeing fresh faced disrupts the start-ups are going to change the world of insurance and I'm going to be a bit brave early on. Am I the youngest person on this podcast? 66. Maybe. Maybe. It's close. That's good enough for me. That's part of what I want to talk to you both about today because you've been successful in business. You've built businesses before. He are, again, launching a new business port in order to understand why you're doing it. We should talk about your antecedents, what your past looks like. John, why don't we start with you? Sure, so I'm originally a life-actually startup but did a wide variety of roles. I was an investment banker doing M&A transactions. I was a chief risk officer, a CFO, a publicly traded large insurance company, some banks. I've been a president and COO as well. So really a wide range of jobs and a good amount of technology. Greg, what's your background? I was fresh faced and changing the world about 25 years ago when I went out to Silicon Valley and spent 13 years with a company that did disrupt the trading in brokerage and somewhat the banking industries. I had my turn at that but I think by your question we have seen some things and I think we've seen some things that have worked exceedingly well and we've seen things that we've learned from. So we kind of bring all of that approach to our present endeavor. But I've been a technologist, my whole career, Wall Street, like I mentioned out in California, and have been in and around and involved with the insurance industry for about 10 years in a couple of capacities. I didn't know how to do it diplomatically. I hope I got away with it. But you guys worked together before, haven't you? There you are co-founding. What were you working on together before? I joined a firm resolution life and we had bought a large life insurance company in Australia and Zeeland as well as one in the US and we had to do a large technology transformation in that to extract all the processing from the other company and set up independent operations. Historically, resolution had always outsourced us to other people, but we decided with the advancements in technology that we could do it ourselves and do it more effectively. So it was a very large project. It was $400 million but the save 150 million a year and we did it on time to budget to spec over three and a half four years. And the real reason we were able to do it was I happened to find Greg to join me and help me in this whole effort and I'll turn over to Greg to give his side of the project. Yeah, mutual admiration society. I met John and Sir Clive right before COVID lockdown and had enough experience with insurance, life insurance, large kind of multinational life insurance is to appreciate the challenge. And like John said, we borrowed a playbook from my years in California. And so I think the experience of these really large scale transformations, which included a number of AI projects and I've personally been involved in AI for 25 years now in some form or another. But particularly with the insurance companies that resolution was buying around the world, we did a number of AI projects together. And John and I were coincidentally leaving resolution around the same time. I basically given ourselves a four year run to get this project done and we did. And that experience around AI and the advancements in the technology and the successes that we had at resolution brought us together to say, let's bring that to the industry at large. So your wise, there's a little bit of mutual admiration going on. You understand AI tech and its potential. But why go again? Is this driven by the desire to work together? Is it because you're inherently entrepreneurial and you don't feel as if you've quite had enough of it yet? Or can you just not face a life on the golf course? What's the motivation? Probably a bit of all those. However, we believe firmly that AI is the biggest technology chain to our lifetimes. And we've been around for a few years, as you've already said, but it's going to dwarf every other thing that we've seen before. And we were quite convinced of it, so even from several years ago, when I retired the first time in 2018, I took a summer course because I had a garden leave last summer. A master's course from Columbia University, in AI, to understand the mathematics behind how AI works. And what's happened since that course, I took in 2018 and now is just simply breathtaking. And we can do things now that took a much longer when we did our projects at resolution. So we're very excited about it. I think it's a combination of all those things, but we really wanted to be part of this fourth industrial revolution and really help companies do it because we were able to implement AI at scale and in production and such that we can do for people. I'm not sure I can improve upon that answer. I, from my point of view, I'm still learning. I'm still very intellectually curious, but more for me, my whole career has been about applying technology to problems and driving value. I love technology. I can get excited about technology like any other technologist, but at the end of the day, it only matters what it can do. What can it do for you? What can it do for me? And I've been on both sides of that desk. And I do like John see that where we are today and where we're going, very much like early internet days. By the way, there's a lot of corollaries. The opportunities for companies to really change the way they're doing business and change the way they're thinking about business is pretty profound. And I want to be a part of that. And I want to help be part of that, frankly. We're 10 minutes in and we haven't talked about portage and what it is. So this would be a very good time for you to tell us about what portage does and what's the problem that's designed to solve. So we formed this company to bring AI to automate the most complicated processing problems that we can find and that we can solve for clients. And in a world of 80-20, where maybe 80% of the work is done, and look, most industries are not 80% of the way to Nirvana or automation, but we like the 20%. We like the stuff that has been either not economically viable to transform or impossible to because of technology just wasn't there. So we custom train models for clients on their data. We support those models when we put them into production. And based on the experiences that we've had with AI projects over time, we want to tackle a real problem with a real solution and do that in a way that demonstrates value, period. And that approach we will do with our capital. If we can't demonstrate value to a client, we'll shake hands, walk away, no harm, no foul. But because there's been so much hype around AI and so many failed attempts to get AI into production, we'd rather put our money where our mouth is and say, "Look, we'll do a minimum viable product. It's a production-ready model. It's trained on your data. We will demonstrate the value. If we can't solve that problem or prove that value, no problem. We'll take our toys and go home." But so far, we haven't had to do that. What further differentiates what we do is we know which AI tool to use for a particular application. And it's this a very important element. We can do it securely. We do it behind your firewall, the client's firewall, or in a secure client-rented cloud environment. And we don't access code or routines to go outside of that. And that gives us high accuracy. You won't drift. It won't make up new answers for you. It tells you when it can't solve it. And that's very important because if it's the client's core processing data, that 20%, they've got to be able to rely on it. It needs to be audited. It needs to be explainable. And there are AI tools that can do that. And there are some of the cat. And you need to know when to apply the right tool for your problem. And we guarantee our accuracy. As Greg said, we guarantee that you get value. I hear that. And I think that I share your view that this is AI and Genetic, don't have AI. I have this sort of game-changing capability. And then everybody did a whole series of POCs. But they did them on simple stuff. It's simple processes that they can automate. You've gone to the other end of the spectrum. And you're saying that everyone else can play in that space. But we are applying what we've learned over the years about AI. It's two of the most complicated things that you can do. We're going to throw these tools that you will data sets and tell you things that you need to know or didn't even know you needed to know. And we'll do that in a way where we'll build it and turn it over to you and only if it works and successful and builds value. Is that what you're telling me? Yes, about it in one, Robin. I'm quite good at working out what people do. It's just I never quite work out what to do with it. So give us an example of the kind of things that you either are doing or you can do for the insurance industry that forces that category? - For the insurance industry, there's multiple use cases. I think one of the things that attracted us to the London opportunity, which will come to maybe a little later, is the post-bind processing and commercial specialty reinsurance in Bermuda. Okay, same classes of business that we find in London, obviously. And this is just an area that has not been addressed. Pricing is kind of one side in the reinsurance world. There's various levels of technology that apply there. But when it comes to what happens after they have the business, it's still BPO's spreadsheets and a lot of manual labor to basically take that information, extract what you need, and making sure that it gets to where it needs to go. Now, that has a really great return for clients. AI in this case is just a better way of automating than maybe we could apply before. AI is just really good at extracting and harmonizing information and it's very economical at this point in time. So that's just one area that we've been going after. But we'll talk more about what we think the big picture is here. Because I think that, by itself, as a good ROI, we'll do those projects all day long. But that information, once it's structured and available for other software, could be AI, to look for patterns in, to help inform reinsurers. That becomes a lost ratio move. So you're taking kind of this operational data that is really just a pain for most of the companies to deal with, right? And turning it into more of an offensive weapon. Most companies aren't set up for that right now because just the fundamental data is just not in a form where it can really help machines understand patterns. But we're convinced and will be happy to work with any client who would like to look at that because we think that's where it's going. And we're going to be there with the clients who want to work with us to do that. - The challenge has been, historically, that you've got all this data in various formats. But there's different language use for the same data elements often between different suppliers, brokers or clients themselves. So to come up with that common language, the common ontology is called, has been a great deal for work and effort. But AI is very good at this now. And finding the terms that mean the same thing and checking them and double checking them. And when an insurer can get all that data and all those elements into a common ontology, that's when the real power will come forward. And this has been, I think every insurer's had a data cleansing project. Thars, I don't think anyone's ever finished them, but you can do this now. And we're really good at this. That's the starting point. And we think there's a massive opportunity there. And then as Greg said, the real end game is when you have that and have all that history and all the detailed history not summarized. 'Cause you want to get all that detail, all the claim reports, all the detailed 100 page claim reports, all the key data elements extracted, stored, then you can start applying very powerful predictive analytic tools to help you improve your live racing. - Maybe that should be your strapped line, your mission will finish for you, your data sorting out projects that you started 20 years ago, spend tens of millions on and never quite finished up. So that tells me that the more complicated the better from your point of view. So specialty market, UK, given how much of it we have over here, Bermuda, US. Is that going to be where you focus your attention? - Yeah, those three markets are our focus right now. - And then in June this year, I noticed that you bought the ingestion capability of insure wave. Is that now part of your tool set? And tell us about why you bought that and then where it fits within the people proposition now. - Yeah, so we did have an opportunity to acquire this business from insure wave. When we were doing the work in Bermuda around kind of post-bind Bortaro processing for the reinsure, we understood that from a pricing point of view, 80% of the information is very similar. So 80% of the data was a lot of overlap. So we always knew we would most likely go to pricing from the post-bind Bortaro processing. When this opportunity came up to acquire certain key assets of this business, including a client for pricing, it just gave us the other side of the coin, right? So to the two sides of the coin pricing and then post-bind processing. So for us, it was complimentary and we are also intending to use the people and the technology and the assets in London to market and sell into Bermuda. We like being outside of the US when it comes to these markets from a data point of view and it just makes a lot of sense we have a lot of expertise in London around particularly commercial specialty, both the technology people, the AI people, as well as the data analysts who need to finish off what the models do because there are no models in the world that are going to be 99.9% accurate every single time. So there's a little bit of human element involved and the client in London, we deliver this as a service for pricing submission. So when they get an output file, it is a finished product for them that many times will go straight through. I think the market is likely moving to tighter and tighter timing SLAs when it comes to pricing submissions 'cause that's the real value and that attracted me as well. It's fine to have expense savings or expense ratio impact. But for pricing, the timing SLAs really intrigue me because we think we can also improve upon what we're already doing for all clients in the London market. That's the strategic rationale behind the acquisition. John, I don't know if you wanted to add anything to that. - Oh, it's a great opportunity. We got some talented people as Greg said and a happy client, which is always good. And we are improving and adding in new lines of business for them right now and also more functionality. So then applying and fine tuning the AI core regimes to increase the accuracy. Instech all sorts is back. After the popularity of our all sorts event this summer, we're returning to the old library at Lloyd's on 20th October for another evening bringing the Instech community together. All sorts is designed to be a little different from our usual events. There are no lengthy presentations. Instead, we'll have 45 minutes of market news, member updates, introductions and short pitches followed by plenty of time to catch up with familiar faces and meet new ones for across the insurance ecosystem. And particularly relevant to this episode we'll be welcoming Portage AI as one of the newest members of the Instech community. CEO Greg Framkey will be joining us on stage to introduce Portage and share more about the company with the network. This event is open to insurers, brokers, MGS and Instech corporate members with just 150 places available. Join us at Lloyd's on 20th October register now, Instech.co. - The industry, particularly the soft market gets itself a bit tied up in saving money, operation and efficiency. Whereas I can't help thinking that the ROI on most of this stuff is around equality of service, speed of service, better underwriting, better decision making, which is where the real value lies. So you bought it's your way June, it's now October. You give us a progress report. How's the business doing? Got any clients, anything you can talk to us about? - Sure, I think given summer, everywhere, it's given us actually a reasonable amount of time to get ourselves organized. We are looking to kick off a delegated authority processing project, even as we're speaking now. Because of our experience in Bermuda, we've trained our models on this type of information. And we believe that we can bring the same level of accuracy and timeliness to the delegated authority processing that we're bringing to pricing. So we're kicking that off. That's a really important project of stay tuned. We'll want to be talking about that very publicly as soon as we can. But we've been organizing for that. And in addition to that, I'm very happy to say that we have hired a leader for our endeavor in London, Ben Howe. And we're very excited to have him on board and looking forward to working together. - How many of you got now? It's time to work for us. We brought over 24 people split between, it's about equal parts technology, a smaller AI team, some data analysts who are really skilled in this product set, commercial specialty. I really know the product well, some product people. And now Ben. - In the 15 minutes since I've got to know you, I get the impression that you haven't come in to sort out the insurance industry's board row processing problems that you have a vision for AI that goes way beyond that. What do you see as the potential of this technology in terms of what it's going to do to the insurance industry? - Let me give an example. The game called Go. Go is one of the most complex games in the world. It's three dimensional chess. And a few years ago, an AI trained system was playing the world champion. And early in the game, it did a move that no one had seen before in that period of the game. It's called Move 37. And it took back. the world champion, he had to think about it for 15 minutes before making his next move. And the AI agent won, that'd be the world champion. And this move had been seen before, but later in the game, no one had ever done it early in the game. In fact, when the computer played it, they thought it made a mistake and was going to lose. And it did it because it could analyze years and years of all the games ever played. And it thought, hey, what if we did this move here instead of there? And that's when you have your data and these predictive models can go through and analyze the data in much more detail than we've almost ever been able to do as humans. That can be a breakthrough. And when an insurer has their data well organized with a good ontology, has all the data there, these tools, these predictive models can then be applied to that to help find that move in 37. Something that you may have missed for the positive, or maybe the negative, that price better, maybe you could lower your price if this data was there, then that you do that or you may stop doing something. But actually, we use mathematical tools that are over a hundred years old. The math has been around since the '50s and '60s and '70s, but the ability to get the math is now here. And I think it'll be one of the most exciting times ever in our industry. And that's what we want to help people with. I hear that. And I think I started a business at the dawn of the internet with my vision of how insurance was going to change. And I was that fresh-based change the world disruptor. And then absolutely nothing happened for about 50 years. So my next question is, I could buy that vision, and I get excited about it, but when's it going to happen? Is this going to happen in three years, five years, ten years, or was it going to be like, invest in one of my businesses and wait 15 years for anything happens? That's the question of the fourth industrial revolution. And what we've seen already is it's coming faster than the first, second, and third industrial revolutions. The changes happening are quite breathtaking. If you step back and think about it, where we were only five years ago or ten years ago. And in industrial revolutions, it is a long time to have it widely used across the industry. But there are people who take advantage of it and who make it through the ups and downs. And they're always ups and downs. And we'll have one coming here in Amsterdam, sure. The few that make it through totally transform the world and Amazon's Apple with phones with them and think about a phone in 2000 when the internet was coming out. So all that has come together and it does accelerate. And in hindsight, it's fast, sometimes it feels slow when you're going through it. But there will be some players, I think, in three years, the five years time that will be able to increase market share very profitably by applying these techniques. And that's who we plan to help. I'm probably more optimistic about the timescale for something like this happening. It used to be there was a course of action that would get all your data organized first and then go get value from it. I don't think that's the course of action anymore. You can get good returns from doing structuring and extracting and organizing data the way we've been talking about. And I think those stand on their own two feet. I think there are means to this end though. I think those are the means to the bigger picture end. And like John mentioned, we're already talking to a couple of companies who are already left to the end and are just looking for a way to get there. So that's kind of the conversations we're having with a couple of clients, particularly in Bermuda. So that I'm clear and the audience is clear, you build this stuff. The world is jam-packed full of consultants and advisors who are standing up in intake events telling everybody what it is they need to do, but it's shorter to people actually doing it to learn this. You're doing it. You're not telling people what they need to do. No, we're doing this. We're bringing the solution in. We're working directly with a client. We support the models. We keep them from drifting. We integrate with other applications, which we prefer to do. We're not a platform play, we're not looking to replace your systems, but we want to be that engine that's receiving and feeding the applications it needs to apply. Now we are also technology nimble, I guess is the best way to put it. So I think given our careers and our histories, particularly what's happening in AI right now, betting all your marbles on AI technology is very difficult to do. So we will license core AI models from really deep resource company and then train those models on client data and then do everything that we've talked about. And that allows us a certain degree of freedom when it comes to advances in tech. And it also gives us a very deep bench to draw from without having to hire 300 people and try to bet that we know where this is all going. That's a good question, and just to be clear, these are our solutions and our products and we're not consultants, certainly from that point of view. I want to do everything, because if you use parlance occasionally, I'm thinking particularly about integrations and other tasks, you'd want to focus on the areas where you're experts. Yeah, we do and we can. We want to stay to what we're good at and what we're here to do. We've been in the game for a while, where we need to bring our own partners. We have some really good partners that can round it out. What do you think needs to be? You mentioned funding earlier, you used to do well-funded, did you have to go off and deal with some repacious VC or PE? Greg and I are funding it right now with help of some close friends, but we're very well-funded and we know all the VCs basically in the world. So at the appropriate time, we will go and get more capital. But as I said, we're taking this very patiently and we know this takes a long time. The sales cycles are long dealing with companies and we're here for the long haul. Look, I think it's fascinating. I think you've got a fascinating story. You bring a different approach. I really like the pricing model and the kind of booth before you're forced to buy. I think that's incredibly attractive. Speaking of somebody staring down the barrel of retirement, it's nice to speak to people who have real experience and still got the passion to go with it because there's nothing better than people who know how the world and industry and tech works and can bring both wisdom and that knowledge. I wish you the best of luck. We always give people on our podcast the chance to leave the audience with one busy insight. The one thing you would like the audience to take away, would you both like? Let me go first. That AI is, as we said at the beginning, we believe is the biggest technology change of our lifetimes. It is the fourth ingest of revolution and companies and people need to start using it and get on it and really not just playing with it, but really thinking about how can this really help my business? It's much more than your co-pilot helping you start through your mail. So I really encourage every listener to start really thinking about it, but demand value from the use of the tool. Greg, I will just say that I foresee a day when we will be processing these kinds of complicated submissions in the time it took to record this podcast. So I would welcome a challenge from any potential clients and we will put our capital where our mouth is, we will agree to do a project that can go into production only if it delivers value, period. That's what gets us up in the morning and that's what we care about, delivering value. So take us up on it. I'm more than happy to engage in any complicated conversation folks would like to have. Well, that's a really good place to leave it. It's been a pleasure chatting with you both. You have become members of insight almost immediately on starting your business, but thank you for trusting us and thank you for working with us. You're the one who is going to bring something for the industry to think about. We really look forward to working with you. Good luck. Any insurance companies out there, there's the challenge, Greg and John are ready to help you with your most difficult AI stuff. Hello, Instec, or either of them on LinkedIn. Nice to talk to you both and see you soon. Good luck with everything. Thank you, Robin. Thank you very much, Robin. Really appreciate it. Thank you. Well, if you've made it this far, I'm pretty sure you found that as interesting as I did. The Instec podcast comes out every Sunday morning, where we spotlight the latest news, leading voices and freshest updates across insurance that you need to know about. If you would like to take part in these conversations, head to www.instec.co to find out how you can join our network and be a part of the insurance intelligence for the curious.