Insurtech Insights 2025—Adina Eckstein & Tom Hulme Fireside Chat

I have so many thoughts about everything you just said right now. And thank you so much, everybody, for coming here. I love how you've self-selected to talk about, I think, one of the most important topics, which is thinking about the people element when it comes to incorporating AI. And you're right, the pace of acceleration is just insane. It's almost like the time between innovation just completely collapses. You don't get any chance to get used to it and get good at it and develop skills. The thing that you need to do is learn how to change. And indeed, Lemonade, I actually was in a panel yesterday, I was watching a panel yesterday, I heard somebody say, 'Yeah, AI has been around for two years now.' And I was like, 'Okay.' So, Lemonade's been using AI in its technology for almost 10 years now, and it's built on a complete digital substrate with a vertically integrated system at the core of its business. We have a singular system for our marketing, for our claims, for our customer support. We've been using AI in all of our functions since the outset. So, generative AI is definitely a big boost. It provides unparalleled opportunities for us to increase the impact. But there's always been AI. There's been fraud detection and image recognition and machine learning. So now we have natural language that enables natural language as an input and an output to those models which just accelerates those uses. So it kind of slides in naturally into Lemonade's tech stack. And you also talked about how these companies are becoming huge within seconds. And we also think about that sort of revenue over employee as a good base for measuring the efficiency and the pace of scale of a company. Just at our last investor day, we outlined a 10x strategy and we expect every single employee at Lemonade to be generating $4 million of revenue based on just their own input and we're already starting to see the results of that. We more than doubled our business in the last three years and our employee base has stayed basically flat.

It's definitely a make-or-break time. There's unparalleled opportunities but also so many companies will just simply not be able to reinvent themselves for the AI age. Many legacy companies will kind of get confined to a certain efficiency zone and as they try to scale they just see diminishing returns. There are so many companies that the way they scaled over time is through series of mergers and acquisitions and then they have multiple systems that are sort of bolted onto one another and they don't speak to each other. So like the data structures just don't work. So even if you manage to put AI on a certain element of your system, it's only going to be confined to a certain zone. So definitely the technical debt is going to be a big thing and in some of the data sets that I can see with our legacy insurers and in other industries it's just going to be a garbage in, garbage out situation based on the structure of your data. But I think the cultural point is also very much overlooked. Culture can be a competitive advantage and leaders don't think about that enough. And in fact, if you have a bad culture, it can be a strategic disadvantage. If the incentive structure within your organization isn't one that's going to incentivize and motivate individuals to argue themselves out of the job, even if you do have a great technology stack, you're just going to get stuck. And I like to think of it almost like an organ transplant. If you need to transplant an organ such as AI or generative AI into a body, which is a company, it's a technology stack, it's a set of people, and you don't have a DNA match between those two objects, the body is just going to reject that organ. It's not going to be incorporated. And you need to think about the DNA of your company and the DNA needs to be one of technology, or you think about those waves of constant innovation and individuals need to not fear the wave. They need to be exhilarated by it. They need to want to be surfers. And so the rate limiter is the worst of the people, the process, or the culture in adopting new technologies.

So think about the skills that we have today. I don't know if you're an actuary, so you crunch numbers. I know it's much more complicated than what I just said now, but let's say that AI is going to do that a little bit better or is doing that a little bit better in the future. If you're a claims adjuster, you're looking at documents, you're corresponding with customers, you're making claim determinations. So tomorrow you're probably going to be training AI models, supervising AI models, maybe conversing with non-AI parts of the world. So I think that the main thing is not to get too hung up on your job description. You were hired as a claims adjuster today. That's great. But I would actually like you to move on and become an AI trainer. And your job is now to supervise the activity of the AI, observe it, extracting the data from the police report and going back to the customer and saying, 'Actually, I'm missing something.' And looking at the policy language, give it a nudge if it's doing something wrong. But that's completely different. And if you have a problem with that, if you're going to come back and say to me, 'But that's not what's written in my job description,' then maybe this isn't the right place for you, because I'm never going to make a commitment that the job that you were hired to do is the job that you're going to have at the end of your journey here at Lemonade or in any of the other companies. So, what we're starting to hire for is personality traits. We're looking for people who are curious, who are early adopters in technology, who are resilient, who are adaptable. Your job is going to constantly change, but then it's my job as a manager, as head of people, as head of operations to make sure that I'm transparent with you about what's going to happen and make sure that the changes that we're applying to your job, to your role, to your department, to how we do things as an organization isn't met with fear, uncertainty, and to help you be part of that process. And I think that is the key sort of thinking about how to specifically go about incorporating it and doing it. And maybe I'll give a little bit of an example around that. Way back in Q1 2023, right? So it's just post-ChatGPT launch, we decided to run a hackathon at Lemonade. I'm sure most of you are familiar with that term. It's a merge between hacking and a marathon. We're talking about marathons again. And we asked people to submit topics. What would you like us to hack? And then the activity itself was two days trying to actually work on something, multifunctional teams, and get it into production. And we had the legal team suggesting to hack their way out of reviewing legal documents and getting their job done on their behalf. And the finance team thinking about the invoice approval process. And the CX team was thinking about let's use AI voice for IVR. At the time, we didn't even have the voice capabilities that we had today that we just launched into production last week, by the way. And they were all arguing themselves out of the jobs. We then brought many of those features into production. The winning teams actually got their ideas to be fulfilled and to make impact at Lemonade. But you have to have them be part of the creation. If I'm building a tool that's going to replace your current job, I want you, the domain expert, to be part of scoping that. I want you to help design that. I want you to help test that. I want you to use the low-code version of that. It's not us engineers, we're going to do that for you and you can go sit on the side. You're part of the creation of this thing. And then it creates FOMO around it. There's only 10 places for CX trainers. 10 places. Who wants that job? Who wants that job? A CX trainer presses pass, fail, pass, fail. But everybody's lining up because it's the most amazing thing that's happening in the company right now. And you glorify it and you give those people recognition. Sorry, I went on for a bit.

I can't think of a single function in insurance that is not going to be, is not already completely revolutionized with AI. I really cannot think of one and all those fields that you just mentioned are examples of such. Also, once upon a time in a far away land, I heard somebody tell me that I need to make a choice between having a good experience for customers or having an efficient experience for customers. You have to choose between automation and a better experience. You have to choose between automation and a compliant experience. And we've got years of data that just bust those myths one by one. As our automation rate increases in claims and in CX, our NPS and CSAT scores go up. We only have 3% of our complaints on claims on claims that were managed automatically. All the rest are from manually handled claims. So I think maybe the specific example to answer your question is I'll talk about claims because I think that we're in insurance. We know that insurance is a high interaction business. So automating interactions is the thing that's going to have the highest impact and call centers and CX is a very obvious implementation of that and claims is a less obvious implementation but it's got a much higher impact or higher potential if you do manage to crack that. In our pet business, we use automation AI in every single part of claims. We use it for the FNOL. We use it for assessing and reviewing and diagnosing medical records for pre-existing conditions. We use it for customer communications. We use it for coverage determination. We use it for payouts end to end. And in that business in the last 5 years, the business grew 20 times. Our loss ratios decreased by 40 points and our cost for handling claims decreased by 70% simultaneously. And if we go back to that myth about the experience, our NPS in pet claims is above 75 and our customer satisfaction scores are 4.8. You really do get it all with automation. And I'll give you a sub-example of the claims and some people and some of the sessions I've been looking at have been talking about catastrophes and obviously we recently just had the catastrophe of the LA fires that was big in our business in the US. By using AI we were able to solve hundreds of claims within the first few hours of the catastrophe having occurred. And still maintained an NPS of 91 and a CSAT score of just shy of 4.9. Speed is what customers want. Fair is what customers want. Non-bias is what customers want and technology really gives you that. But AI, and I could go on, I could talk for another five minutes or maybe five hours about AI and pricing and in underwriting and maybe we'll jump into one of those examples in a moment.

We don't use generative image generation for our marketing materials per se. We've got very particular illustrations, very pedantic illustrations that are part of our brand story. And actually we haven't found that AI has been accurate enough for the image generation itself. But we definitely use AI in marketing in our LTV predictions in assessing how good campaigns are performing and segmentation, churn predictions etc. But not so much in the actual image creation itself. It's going to be interesting. We're starting to see tailored videos sent to customers for example after the experience to remind them a year to build brand loyalty. It'll be a fun place to observe.

I saw some people in the crowd here from my HSBC days. I'm not here representing HSBC, but I do appreciate that Lemonade has had the good sense of being founded in the digital era. Hence, have the ability to be a growth company sitting on a growth stack. So, I have the liberty to be able to say to my team, don't fear for your jobs. Every single one of you is going to create much, much more impact. I just want you to adjust. I'm well aware that that message of change your skills, AI is not going to take that job is a much more difficult message to land when you're in a shrink-to-excellence situation. And frankly, given my experience in some of my previous companies, I don't think that fixing things on the peripheral or adding on an AI tech that's just going to slide in on the side is going to get the job done. I think most companies that end up disrupting themselves successfully are companies that are not run well. Companies that are run well, that have good margins, that are doing well, that have a good distribution system going on, I think that the only way that they're really going to succeed in this next phase of the revolution is by completely starting from scratch, which is what we did at HSBC.

So I kind of referred to it in what I previously said, but it's just you have to go all in. You have to go all in. Don't hire a chief AI officer. Have him or her build an army of people. Don't acquire $20 ChatGPT licenses for your team and then say we did the AI. It's simply not going to work like that. You have to go big, you have to go bold. I don't know if that's by acquiring another brand and letting it start to shine. Because, you know, we all know about the innovator's dilemma and cannibalizing your business and all that's going to happen. It can't be an add-on. It has to be the core.

So, I really do think it's going to be used everywhere. We're doing some crazy, crazy wild stuff in HR. And I say that with all seriousness. We actually built an in-house generative AI-based performance management system that we use internally. So, our internal bot, Kooper, will go and reach out to employees, collect 360 feedback. If you don't answer right away, it'll bug you until you do. And then it will take all of that feedback and compose a set of hashtag-style performance, what we call superpowers and kryptonite. And it will synthesize all the feedback into a summary that the manager can go and review. And that actually makes the performance review less biased, less recency bias, more comprehensive. And it means that people are getting more rapid cycles of feedback and performing at their best. Now, that's a wacko example because I think that there's nothing that AI cannot do internally in an organization. We're using AI in procurement. We're using AI in finance. We're using AI in compliance. And then what we create, if we're going back to the original point about the DNA, is we create a bunch of what we call makers, our employees that are constantly surrounded by AI. Everywhere they look, AI is doing something for them, creating some sort of superpower dynamic where they can do so much more. And they're harnessing these AIs to make them perform much, much better. So they can't help but becoming innovators themselves. And that's this life cycle, this beautiful flywheel of actually creating a DNA of technology culture. And then they create the next set of ideas that argue themselves out of their new job and that just continues. So actually if you build a surrounding that just feels like technology and the technology is great, there isn't a single area that you cannot do that in and we're already doing that today.

That's a really good question. I would say given, let's say the company was starting today, given the rapid pace of technology, I would say it's like you're on a fast train and you kind of, if you manage to establish a company these days and raise money and have product-market fit, you're probably on the train. And it's kind of all about just hanging on to that train. Attracting the type of talent that is going to want to be curious, that's not obsessed with army building. I really do think it's all part of the talent and the corporate culture that you build, which is a culture of collaboration and curiosity and trying out new ideas and not being afraid to push things to production and to fail and to course correct, etc.

Well, in the future, in the next coming years, Lemonade is going to 10x its business. It's going to be offering any single product in any single market. Globalization is going to be less of an issue, localization is going to be less of an issue. We're going to be offering insurance products wherever we can offer good margins because that's where our efficiencies will shine. Internally I think we'll probably have some sort of multi-agent system where we'll have like the finance agent and the claims agent and the CX agent and maybe there'll be some sort of orchestrator layer and we'll still have humans and our org chart will look like there'll be AI, human, AI, human. Some of our AIs will be managing humans and our humans will be managing other AIs and our other AIs will be nurtured and onboarded and fired and rehired. And the humans will kind of help interact with the non-AI parts of the world, whether it's regulation or things that you can touch. I don't know what else is not going to be in the silicon sphere at some point. But we'll definitely be still around creating value for our shareholders and our customers and our employees and we'll be having fun and we'll be wearing sunscreen.