Renters and homeowners insurance carrier Lemonade burst onto the scene last September with plans to “Uberize” the insurance industry – that is, to leverage the breadth of digital technologies, including big data, smartphones, and machine learning, to cater to consumers’ need for instantaneous service online. To support its business, the company built a network of artificial intelligence bots, named after real-life employees, which manage the claims, enrollment and customer service experience via its mobile app.
Lemonade CEO and co-founder Daniel Schreiber sat down recently with Digital Insurance’s associate editor Danni Santana to discuss how its machines are learning, the company’s its business model, and how lemonade plans to maintain a culture of digital innovation. What follows is an edited version of their conversation,
Digital Insurance: How would you describe Lemonade?
Dan Schreiber: Lemonade is an insurance company built on artificial intelligence and behavioral economics. Those are the two pillars of our business.
DI: Let’s start with artificial intelligence. How does AI make the insurance experience at Lemonade unique?
Schreiber: Our bot “Mia” can sell an unlimited number of policies concurrently any time of night from any device. And the idea that you can file a claim without filing a piece a paper, without speaking to a human being, by a few taps on a phone and recording a short video of yourself so that it is all over in three or four minutes, is a pretty radical departure from the incumbent experience. But we’re not able to do it across the board yet. Our AI is still learning.
DI: What does a “learning” AI look like?
Schreiber: All claims start with our bot “Jim”, but Jim doesn’t have the confidence yet or permission to make determinations on all claims. Some claims we trust the AI to make the determination; it pays it and does everything. But it’s still got its training wheels on. Sometimes it says “I think I should pay this, but let me get ‘real’ Jim to look over my shoulder”. But it’s a leaning system and it will get better and that’s what I meant by across the board, getting to a greater percentage of the claims being paid that way.
DI: Do you have a general idea of when you’ll “get there” with AI?
Schreiber: Not sure there’s a real “there,” there. It’s just a general process of improvement. We have human beings that produce reports on what questions are commonly asked and rather than having them repeat the answers to those questions, we train our system to be able to answer those questions directly. There will always be a new frontier. It’s just a question of transferring knowledge from humans to machines for processes which are repeatable.
DI: What does this require from a technology-stack perspective?
Schreiber: It’s a combination of server-side technologies — all the rating engines and back-office tools our team has in order to track policies — and front-end technologies, which are focused on the user experience. You have an end-user experience that is delightful and instantaneous. But behind that, to make a simple experience, you have to have very complex technology. You want to hide that complexity, obviously.
DI: Do you worry about not having a personal touch customers may want?
Schreiber: We get a lot of inquiries that are answered by human beings. We have a phone number you can call and we also have chat through the app if you’re asking a question the AI hasn’t been trained on. But the short answer to your question is no. I think the target audience for us doesn’t look for a human touch. Admittedly, there are customers who don’t view the world that way and they can go to Liberty Mutual, Allstate or State Farm. There are loads of people who will give them that human touch.
DI: When Lemonade launched, there was a lot of emphasis on the idea of “peer-to-peer” insurance. Is that where the “behavioral economics” comes in?
Schreiber: We’re actually using that term [peer-to-peer] less. Basically, when you buy Lemonade insurance, we ask you to choose a charity that is near and dear to your heart. We create a group, in a sense, of peers who are defined by their common cause. The “peer” element is using the member’s premiums to pay claims and if there is money left over, it is given to their common cause. We had expected it to be something that would help people understand the model, but it generated more questions than answers. We decided if it’s not helpful, we won’t ram it down anybody’s throat -- we’re not hooked up on one title or descriptor.
DI: Lemonade received investments from a number of venture capital firms before launch. What business areas did those funds go to?
Schreiber: We are using the money to hire the talent and build out the AI infrastructure. We’re a small company, but the talent pool is deep. The technologists that we’ve hired in Tel Aviv are some of the best and brightest and have just done extraordinary work. Some of it is also for regulatory compliance. Regulators require us to have a reasonably healthy balance sheet. So we needed some money in the bank that we don’t spend.
DI: You previously have said that Lemonade is a different kind of carrier because it is not as bureaucratic as other legacy insurers. As Lemonade grows in staff and stature, how will it keep its innovative culture?
Schreiber: You do see tech companies becoming institutionalized after a while. It’s not something that keeps me up right now. But we do have a term we use internally called the “black hole”—the idea that there are gravitational forces upon us to make us conventional. And we have a culture where we call each other out when we have some black hole thinking going on.