LMND | Lemonade Inc. | earnings call 2026-09-15
Verbatim transcript (source: faster-whisper-small.en)

an absolute pleasure to be here with you all and to talk a little bit about our industry and where AI and Inchore Tech in general is going. Now to gain some perspective, I'd like to start off with a really zoomed out overview. We're going to take a whistle-stop tour of risk pooling throughout the ages and we're going to start literally with a pre-human era, go through the incarnations of risk pooling throughout the major events in human history and then start looking at the post-human incarnation. How are we going to see AI doing more and more in the insurance space with humans helping less and less? So whistle-stop tour, we start with our closest relatives, the bonobos and the chimps and we'll find that there's many behaviors built into the DNA of primates quite literally that involve risk pooling. Various behaviors, reciprocal behaviors, food-sharing behaviors and as best we can tell ancient humans, prehistoric humans did exactly the same thing. Every extant hunter-gatherer society today does it but it really is only upon the arrival of the agricultural revolution that something that begins to look like insurance emerges. Today we would call it assurance. Four thousand years ago in Babylonia today's Iraq the Hammurabi Code governs loans for trade, maritime trade, allowing you to charge a premium in return for debt forgiveness in the event that your ships go down at sea and this arrangement survives into the Greek period, the Roman period and then the next major revolution, the commercial revolution, the 14th century sees the arrival of what we today recognize as an insurance contract, a stand-alone contract, not an addendum to some other commercial arrangement where throughout Europe we begin to see things that begin to look like insurance companies and legal contracts as I say that are about risk transference in the modern sense of the word. I say the modern sense of the word because the modern complaint about insurance companies emerges concurrently with the emergence of the first insurance contract. So already in the 14th century we have Francesco writing to his wife to complain about insurance companies. It is sweet to them to take the monies but when disaster comes it is otherwise. Each man draws his rump back and strives not to pay. This age old complaint about insurance companies. I'm not entirely sure what drawing your rump back means but I don't think it's a compliment. And if we zoom forward to today you look at the crowd source dictionary, the Urban Dictionary and the number one most voted definition of insurance almost 20 years in a row is a business that involves selling people promises to pay later that are never fulfilled. Not that much has changed in these 700 years. Like all good dictionaries the Urban Dictionary will also show you the usage of the word. GOMA paid flood insurance premiums for years but then the insurer decided to go out of business when the flood came because management had spent all the premium on hookers and private jets. If you want to know the pulse of the people and what they think about insurance just go over to the Urban Dictionary and see how they define it. So something that looks like modern insurance has been around for quite a while but in an important way the pre-modern era was profoundly different. The legal structure may have been similar but the mathematical scientific actuarial basis of what we recognize as insurance was entirely missing. In 1540 King Henry VIII not too far from here decided to raise some money by selling annuities and he was willing to pay 14% for life but he charged the same amount regardless of the age of the purchaser. All pricing and insurance was basically just supply and demand of commercial transactions with no scientific or mathematical basis and that changed in the scientific revolution. So I'm flying through different kind of major revolutions but of course 100 years after Henry VIII 1654 Pascal Fermat started corresponding and that gets us to modern probability theory but newly comes out with the law of large numbers and truly modern insurance. Not in the legal sense but in the mathematical sense emerges during the 17th century and a lot of companies that are still around today not too far from here, Lloyd's of London, Aviva, others around the world emerge on the basis of everything that was developed during the scientific revolution. So it was called the revolution for good reason because not a single insurance entity collective company that predated the scientific revolution survived it. Not a single one. Conversely many of today's major brands have survived and thrived for hundreds of years for centuries on the foundations of the scientific revolution. It really was quite a revolution but I put it to you that longevity should not be confused with immortality because a new revolution is afoot. It's gathering steam and it's rendering incumbents structurally disadvantaged. It doesn't mean they can't cause correct. It doesn't mean they can't harness all the goodness that's what these two days are all about but it does put them wrong-footed going in to the current revolution. I'm generalizing and this is a self-selecting crowd so it may not apply to you but overall incumbents find that many of the things that they regarded as assets reveal themselves to actually be liabilities. You have 10,000, 20,000, 30,000 brokers or agents. You thought when he took the job that that would be a huge asset and then suddenly in the era of the direct-to-consumer app-based distribution you're wondering if that's not an albatross around your neck. If that's not creating channel conflict, sales conflict, cost structures that don't make sense in the direct-to-consumer era. Your culture, your management teams were handpicked and groomed for legacy preservation whereas what you really need is an engineering culture and a team hell-bent on business transformation. Not an easy transformation to make. And then there's the tech. You've been building tech since the 1980s or 70s but you find that now in the era of AI it's hard to cash in and to truly harness everything that's coming out because you've got decades of legacy that are built up. Your systems behave more like a black hole devouring time and cash than that black box that makes everything faster, cheaper, smoother. Everybody in the room will have a dozen anecdotes to tell I'll share with you one. I was talking to an IT professional at one of the major insurance companies and he was bemoaning the difficulties that they were having implementing a very simple system, Salesforce.com. Should have taken days, a couple of weeks, it took them two years and he had a litany of issues that made it so difficult to integrate such a simple technology into their tech stack. And as he's listening the morph he says and then we had to integrate with JPL and I go blank. There's plenty of acronyms that I'm not familiar with. He says, oh sorry, Jim's programming language. Stay here, oh no, no at the back there. Jim's programming language. So this company had an IT guy called Jim in the 70s or 80s and he did amazing foundational work for the company based on his own unique language, Jim's programming language. Jim died but his legacy lives on. You get the point. Which brings us to what I see and everyone will have their own take on this as the InsureTech manifesto in its purest form. There are variants on the theme but certainly this is what animates me, what animates the company that I work at, Lemonade. And it is quite simple that companies built in the AI era on an AI and digital substrate will be able to delight consumers, collapse costs and quantify risk like never before. Welcome to that revolution. Now if we set aside the manifestos and the slogans and the playful or pugnacious banter, we can go a little deeper and understand why they're strategic at a structural level, I say what I'm saying. This is a screen capture of a slide from Lemonade's very first presentation. And naturally because I work at Lemonade, I'll be talking about Lemonade but it's a stand-in for any number of great InsureTech companies. The very first slide. Back in 2015 we founded the company. This is the first slide that I showed our board when it first convened and our management team when it first convened. This is the first bullet of the first slide and it lays out our strategy and you can see that a lot has changed. We hadn't discovered the color pink yet. We thought that it was clever to call ourselves a peer-to-peer insurance company. Some things have changed, some things have not. A lot of people are discovering AI for the first time since November 2022 with the emergence of chat GPT but AI was really the founding thesis of Lemonade 10 years ago. And you can see here we talk about insurance powered by artificial intelligence rather than artificial delays. And we talk about software that can settle claims in seconds at the touch of the screen at untouchable rates. Very much what still animates Lemonade and a lot of the InsureTech sector 10 years later. Let's dive deep. Why? Why is technology and particularly artificial intelligence such a game changer, such a sea change, not an increment but a disruption, a step change. I see it as coming down to two things. Companies built on an AI substrate enjoy a structural advantage and a competitive advantage and those two rarely come together with the strength and the force and the ferocity that they do in this space. So let's talk about the structural advantage. Companies like ours have a single proprietary system built in-house by engineers who work at Lemonade to understand the code, to make changes in real time, dozens of times a day. That's the rate of progress and change. But more to the point, I mean a single proprietary system allows us to garner and deploy insights and data at every single touch point with the customers. We are a young company. We operate in Europe and in the United States. I can say now there is not a single insurance company that I know of, including those that have been around for one or two or three hundred years, that we would trade data sets with if we could. When you build it right, the depth of the insights that you glean compensate for the fact that you haven't been around for quite so long. And it allows you to take insights from marketing and apply them in sales and apply them in claims and apply them in pricing and do all of that in real time. And you can trust that Apropos JPL with what happens across the industry. This is GEICO, one of the United States' foremost insurance companies and their owner a year ago standing on stage and saying that GEICO has more than 600 systems that don't really talk to one another. And that is a structural advantage. It's very, very difficult for large corporations, willing and keen as they are, to reinvent the very chassis upon which they are built while the car is moving. There are almost no examples in any industry of any incumbency managing to pull that off. 600 systems down to one that really doesn't happen. It's almost unprecedented. And that's why it's a structural advantage. Legacy becomes a pain. It becomes a drag. It becomes something that holds you back. But more to the point, that structural advantage begets a strategic advantage, a competitive advantage. And that is because AI is better than humans at the two things that matter in insurance. Precision or accuracy and automation. And by precision, I mean the ability to quantify risk. That's the core of what we are about. In the early days of companies like ours, a lot of people paid us the backhanded compliment of saying the great up branding, they've got a cool app, entirely missing the fact that that cool app was garnering and generating and utilizing data in ways that change the 90% of the iceberg that has submerged and that is invisible. It's not about, certainly not only about the great experience, this is about the data architecture that yields and the insurance results that that ultimately gets to. And insurance, remember, is pretty unique in the following sense. AI has been transformative for industry after industry after industry. Look at what Spotify has done to the music industry or Amazon to the retail industry and how they harness AI to get you just the song or just the product you want. But almost every industry other than insurance, almost every industry other than insurance, is ultimately selling you either atoms or bytes. When Spotify guesses that you want to listen to Bohemian Rhapsody right now and it plays a few just at the right time, it's using AI, but the song that you listen to is unchanged since Freddie recorded it in 1975. It is the same byte. AI, machine learning, doesn't change the song. And when you need a toaster and you go into Amazon and it recommends a particular Russell Hobbs toaster and you get it and the box that arrives the next day has the same atoms and had you got off your chair, gone to John Lewis and bought it there. It changes everything around the experience, around the product, but not the product itself. Insurance is different. Insurance is selling probability theory. It is monetizing statistics. It is an algorithm once you strip away everything else, that's what you're left with. And boy oh boy can AI and machine learning transform algorithms. That translates into loss ratio, into ability to price risk, and the ability to get dominion over how you price, the ability to de-average, to break up monolithic groups and break them again and break them again and break them again. And then you get to the expense ratio. And automation is just phenomenal at collapsing costs while delighting consumers. Lemonade has been deploying and we share a lot of our statistics about this more and more and more automation. And I can tell you as an empirical fact, our customer satisfaction scores go up every time we move something from humans to AI. Every time. While collapsing costs. While increasing response times. So you collapse the cost to serve not at the expense of customer experience, which is what the trade-off used to look like, but to the delight of consumers. In guys in our industry there's an expense ratio, there's a loss ratio, rather unimaginably they're called the combined ratio, and that is the whole kitten caboodle. And AI transforms both of them more than anything else. Now I think that for in short-text that translates into untouchable rates, but frankly that's a choice. What I've shown you or claimed is that having the right AI architecture can let you price risk better and serve customers at a lower cost, but you could decide to keep your costs, your prices rather on par with the industry and just have higher margins. That's the trade-off that you get to make. But I think for in short-text the choice is always going to be in favor of lowering costs and capturing market share and growing. And I want to talk you through why I say that. There's our profit frontier, which in theory it doesn't matter. You could charge x and you'll have this much profit and you can charge 25% less and you'll have the exact same profit in theory at any given point at every any given scale. It's much of a muchness whether you optimize for growth or you optimize for gross margin. But while that's true at any given scale, we are not indifferent to the scale itself. As you grow, as in short-text grow, the profit frontier itself shifts. And if this was the potential profit pool at a certain scale, this is its potential at a higher scale. Scale moves the frontier and this too is different for incumbents than it is for in short-text. Because when you're built on a human substrate, you do get advantages to scale, but they diminish and taper off over time. And the inverse is true when you're built on a digital substrate. You get accelerating returns to scale. The more you grow, the more profitable you become, which is why certainly Lemonade and I think many others will choose increasingly to cash in that cost structure advantage for lower price, greater growth and ultimately bigger profit pools as we get through, as we push that frontier. And if I connect the dots and just kind of outline the emerging strategy as I see it, is that AI leadership begets you to gifts that keep on giving. Precision leadership and automation leadership. Certainly for us, we choose to cash those in as price leadership, which in turn begets growth leadership, which ultimately gets to profit leadership. It's a long game, but this is the game that we at least are playing. And as you grow, you get those feedback loops. You see that frontier moving. The cost structures remain a constant. More and more of the gross profit drops to the bottom line, and your precision keeps growing as you're able to break up monolithic groups as data accumulates. Now, this was our thesis 10 years ago when we founded the company. This is ostensibly what I told the team on that day, that critical day, that first meeting. I think it held sway. I think it held water. We managed to get investors and employees to buy into it, but it was unproven. It was a hypothesis. Today, and this probably wouldn't have been true as little as 12 or 18 months ago, we and other InsureTex can talk about actual results that begin to reflect this hypothesis. And in the coming years, I think you'll see that compounding in powerful ways. Let me show you lemonade results in the last four quarters. And as I say, I talk about lemonade for obvious reasons, but you can look at root and you can look at HIPAA and look at any number of other InsureTex and see not altogether different kind of trajectories. This is our loss ratio over the last four quarters. Our actual gross loss ratio improved more dramatically than this, but I'm tapering it by showing you the trailing 12-month loss ratio in order to take out seasonality and other kind of events. This is a trailing 12-month loss ratio. And you can see the 10-point drop. This was actually a ninth quarter in a row of an improving TTM loss ratio. So this is not a one-off. This is not a two-off or four-off. This has been a constant. Now, insurance companies know how to improve loss ratios. I remember early on being told that you grow into excellence, being told that if it grows like a weed, it probably is all these kind of truisms that the industry lives by. But we've been growing pretty quickly concurrently with crushing our loss ratios. 26 percent growth. In fact, we've been accelerating growth over the course of the last five quarters. Every quarter has grown more than the quarter before it, and we have guided to continued acceleration in this year 25 and continued acceleration into the next year 2026 as well. And that's already a bit of an unusual pattern. Rapidly falling loss ratio is rapidly expanding book of business. If I layer on top of that some of the cost structures, you'll see just how unusual some of the financial results coming out of InsureTech are. That's our headcount. We grew 26 percent and we shrank our headcount. Fortunately, no rounds of layoffs. This is natural attrition. But we're able to do increasingly more with increasingly less because of AI. We have fewer employees now than we had a year ago, and a year ago we had fewer employees than a year before that. This is not a one-off thing. You grow at 26 percent going into the 30s. You need more intelligence at work in your company. You have many more customers. But it doesn't have to be human intelligence. And you see that in other cost structures. This is our cost per claim over the same one-year period. Our cost per claim has dropped by almost a third over just four quarters. And you layer on top all of these and you say, shrinking headcount, collapsing costs, improving loss ratios, growing book of business, our gross profit trebled during this time. And you start to see that the hypothesis that was just that only a few years ago has already got some receipts, some actuals, some evidence-based, rather than just a thesis that kind of held water. Where does this all take us? Well, there's an ancient Chinese proverb, if you don't change direction, you'll end up where you're going. And the question of where do we end up depends very much on where you're facing. Where are you going? I said earlier, this is a self-selecting group. I imagine everybody in the room is tech forward. But not everybody in the industry is and not everybody in your company is. I want to demonstrate this by showing you an ad that State Farm, the largest insurance company in the United States, ran in 2018. This is an attack ad on Lemonade that says so much. 2018 Lemonade was doing less than $100 million in premium. We have more than 10x in the intervening years. They were more than a thousand times our size. And this is the ad that they run with two notable stars from the NBA. This happens all the time. I'll help take care of it. No, I'll take care of it. What's that? The budget insurance companies are building these cheap knockoff robots to compete with us. How may I assist you find gentlemen? It's kind of freaky. All right, look at his face. His nose. His ears. He's got terrible eyes. Well, not so still. I never had a girlfriend. Guys, regardless of the looks, one thing is certain. These bots don't have the compassion of a real State Farm agent. I have compassion. We're cool. Okay. Absolutely love that ad. We did something that to my best of my knowledge has never been done before we downloaded it and we started paying to have it promoted on Google. We pushed the ad. We loved the way it delineated a clear line with them being anti-technology and us pro-technology. We loved how ridiculous in our eyes it made them look for mocking bots that were already at that time outperforming their own agents in terms of compassion and customer satisfaction. They didn't like that very much. They got Google to take it down very quickly. But it was fun and it got some coverage and it shows the first reason. Don't mess with the company one thousandth your size. There's no upside in it. You are upscaling us. You are diminishing yourselves. Not a wise strategy, but beyond it just being not smart marketing. It was factually wrong. This is a review from 2018 the same year. Pregenerative AI. The cheerless customer service representative at Liberty Mutual had about half the personality of lemonade chatbots. Back then, lemonade chatbots were plenty empathic and they only got more. And that brings me to the third reason it was wrong. It totally missed the trajectory. It totally misunderstood where we're going and it's not good enough to look at what stuff is available today on the floor outside. We have to skate to where the puck is going. We have to figure out where we need to be. So where are things going? I'll show you a quick video, a longer version of this was tweeted out not too long ago. This is us moving from Maya the bot chatting through texts and emails to chatting through voice. This is now live, but this was a demo of one of our engineers showing this and we shared a longer version of this video. Do note some of the framework in which he's working. This is live to consumers already because I'll come back to it later. I have a pet policy with you, but I'm not sure for which pet. Can you please help me out? Of course, Shai. You currently have an active policy for your cat named Princess Donut who's a Persian. Let me know if you need any more details. Yeah. What is my monthly premium please? Your monthly premium is $47.69. With fees, the total comes to $49.69. If there's anything else you need, just let me know. Yeah, I do. I have a new cat and I would like it insured with lemonade as well. Can you please help me get a new quote? You can start getting a new quote by visiting the link I've sent you via SMS. Let me know if you need any help. No, my God. That has been very helpful. I don't need any other thing. I'm glad I could help. Have a wonderful day. Okay. So once the call is ended, we can immediately review the conversation in our conversations list page. You can see it appears immediately here. Let's click it and I can see the full transcript of the conversation. I can listen to the entire audio of the conversation and see the different segments when the voice assistant spoke, when I spoke, etc. I can also see the different events that occurred during the conversation. For example, do you remember when I asked if I can get a new quote? So Maya sent a query to our system and got that response. We can review it here. Once the response arrived, Maya decided that she can send an SMS with a link to get a quote. Okay. You can begin to see, and I'll come back to this in a minute, it's not enough to have voice recognition systems paying $20 to GPT a month doesn't do it. You have to have a single system that can integrate, that can send quotes, that can look up. You have to be able to control the whole thing in order for that seamless experience to materialize. And as I say, this is now live. Let me tell you why I think this goes in a not too distant future. We've gone from text to voice. That's live. I think we're not too distant from having video. The next few videos I'm going to show you are entirely AI generated, but they're indistinguishable from normal human interactions. So Maya can have video conferences. Jim Abbotte can do the same. This woman has never existed. This office never existed. The people behind her don't exist and never did. But it's not just about customer facing. This is about internal transformations, about crushing those costs, holding headcounts steady even as you grow across every facet of the business. And to do that, we envision a future, and again, not a distant future, where a lot of our colleagues become AIs. This actuary, who can get on a Zoom call at any moment, 24-7, has access to all of Lemonade's historical data and actual data in real time, will cost about three bucks a month. She'll interact with anybody that she needs to interact. She'll give guidance. She'll be able to have multiple Zoom calls concurrently. You get the idea. And we're already using AI in our engineering. Our engineers are using AI to output far more code than a human can. But we're not too far away from having an AI boss for engineers. And again, this is AI generated. This doesn't exist. But you get a sense of how we think a lot of code reviews are going to take place in the next few years. I'm not talking decades out. I'm talking about before the decade is out. And that will be true in engineering, and that will be true in actuarial, and that will be true in compliance and legal, where the AI will not only be able to review contracts at the speed of light and look at for compliance errors in real time and do three months' work in three seconds, but we'll be able to interface with humans with his colleagues using the interface that we humans expect and want. In fact, I put it to you that there is not a theoretical or an empirical thing, limit to what AI can do relative to humans. We know of no reason to believe that there is anything of commercial value that we can do with neurons running on protein that cannot be simulated by neurons running on silicon. Nothing. And if that's not your experience yet, then underline the word yet. It's coming. It's coming fast, and much of it is already present. And that is true customer-facing, actuarial, legal, financial, every single facet of the business, if you're architected for it, and it's not just architecture in the classic sense. And here I come to my kind of parting tips, if you like. I talked about, in short, in the purest sense and the lemonade sense, if you like, of an insurance full-stack, vertically integrated company. But as I say, there are flavors and ways in which InsureTech makes sense for you. Even if you don't work at such a company, everybody should be in tech forward. My three tips go as follows. One, you have to commit wholeheartedly, unreservedly, damn the torpedoes, burn the bridges to AI. When Netflix started streaming videos, it was crap. It was unwatchable. You much prefer to have them or Blockbuster give you a DVD and plug it into the machine than stream. But what Hastings understood is that bandwidth is going to zero. It's going to be infinitely big, and it will be ostensibly free. And he built a business, a stunningly successful business, based on that insight, even though when he launched it wasn't quite there. You have to understand where things are going, how quickly they are going there, and commit wholeheartedly. Because if you hesitate, if you defend, if you do more, if you use countervailing forces inside your company to slow down, you will have your lunch eaten for you. Which brings me to my second point, and that's a cultural one. And honestly, one of the world's leaders on this is Adina Exine, who's going to be presenting at 10.35 in the purple room tomorrow. I would not miss that. Culture is more important than technology when it comes to technology. I think that many companies will absolutely fail at the AI test, and they'll fail at it for the same reason that most acquisitions fail. On paper, it all makes sense, but there isn't a DNA match, and there's organ rejection. And there's good reason for organ rejection. Some of the things that I just showed you would be daunting and scary to cultures that have not been prepared for it. It's not enough just to write a big check to some, integrate, and expect it to work. It will fail. So, preparing your DNA, preparing your senior leadership for a new world where having more employees is not, or is no longer, the telltale sign that you're a powerful professional, but to the contrary. Shrinking your organization, boasting about the fact that you're doing more with less. What a deep cultural change that requires, and the same is true every other place. People being threatened by technology, etc. The single biggest obstacle. And the third one is AIs are people too. And what I mean by that, it's very tempting to knock up a demo and just show you what GPT can do for 20 bucks a month. What you saw before with the voice implementation, and we don't have time to delay on it, is profound and deep. Because the only way for AI to be successful in your organization, beyond being a tool that people use to write things and query, to truly transform an organization, is if you prepare for it. AIs are like people in that, if you bring it on board, it needs to be trained. You need a platform upon which it can be trained. It needs feedback. It needs to be managed. It needs people looking over at children saying, yes or no. And if you haven't built that platform, the AI adoption will come to nout. You need to be able to fire AIs. And as new models come out, you need a layer of abstraction that allows you to plug in and plug out and displace and replace and have multiple AIs solving different problems. You need to build a kind of infrastructure that you have already built for your people. You have entire HR departments that do nothing bad. You have feedback cycles. You have managerial reviews. You have so many tools that you use to get your people to do what you need them to do. But when it comes to AIs, people think, oh, we will give it to the IT department. We will pay for membership or let them use the free one. That won't get you very far either. And I end up by going back to that, ensure tech manifesto that I kind of provocatively threw out there. But I do think that wholesale, massive, deep adoption of AI will allow companies to delight consumers, crush costs and price risk in a way that is absolutely unavailable for those who don't make those commitments. And this is just the beginning. Thank you so much.