Machine Learning’s Role In Insurance

INTERVIEW

On episode 34 of the InsureTechGeek podcast, talking about machine leaning’s role in insurance with Zehra Cataltepe from TAZI. 

The InsureTechGeek podcast, powered by JBKnowledge, is all about technology that is transforming and disrupting the insurance world. We will be interviewing guests and doing deep dives into specific technologies that we see changing the industry. We are taking you on a journey through insurance tech, so enjoy the ride and geek out! Back in a second. 

JAMES: Right. Another day. The hurricanes have passed in Texas for now. We have clear blue skies finally, after a week of rain here in central Texas, and it is a gorgeous Friday here as we record this. It is Friday, September 25th, 2020. That is right. This dumpster fire of a year is getting close to being over. We are almost in the 4th quarter of 2020. Definitely not the year any of us thought it was going to be. We had to modify all of our one-year goals and do all kinds of adjusting, but that is okay. We will survive. And with me of course as always, my cohost, the illustrious Rob Galbraith. Rob, how is it going?  

ROB: It is going, man. I guess that is the answer every week, these days. So good to see you again, James.  

JAMES: Good to see you too. We are glad to have with us from the beautiful state of California, a little smoky state. Everything probably has a smoky smell right now. All those forest fires over there. It has been quite a wild ride the last few weeks over in California. With us from California, Zehra Cataltepe from Tazi. Zehra, so good to see you on the show today. 

ZEHRA: Good to see you guys too. Very happy to be here. And I look forward to more sunny skies here, and blue skies in California, obviously.  

JAMES: Yeah, no more orange skies please, it looked like the apocalypse sometimes when the smoke was in the sky, of course, everybody on the internet saw those pictures. You are in the Bay area, is that correct?  

ZEHRA: Yes. I am in Los Gatos and we have seen some red suns. The first time I could take the blue-sky picture, I shared it with everyone. Look the sky is blue today. 

JAMES: Yeah. 

ZEHRA: For the last couple of days, getting out, seeing the blue sky, thanking God, and returning to our lives. We did not know it was such a blessing before.  

JAMES: Exactly. It is a blessing, and you are in such a beautiful part; I mean Los Gatos is like right there at the gateway to get down to Santa Cruz and to go down to the coast and see Monterey and Carmel. I absolutely love that part of the country. My great aunt lived in Carmel most of her life, and I would visit her in Carmel by the sea. We would go to the Monterey Bay aquarium and it is such a beautiful place. 

ZEHRA: Definitely. I totally agree. I have never been there until a couple of months ago and there are just lovely little towns here and there. And then Carmel is on the most famous ones, but every one of them is unique and definitely worth seeing. And there is always almost a 10° difference between inland and the beach. The beach is warmer during winters. Did you know that? 

JAMES: Yeah, the ocean moderates the temperatures there. It is wild, the closer you live to the ocean, you get a more moderate temperature, but it is still not exactly an ocean you would want to go for a swim in without a dry suit or a wet suit on because it is a little cold there. California has been in the news a lot lately for forest fires and all kinds of things. But this week it was in the news for something interesting and impacts insurance. I am going to start with this, and we are going to jump into the show because I feel like it is so timely. The governor of California, Gavin Newsom announced that by 2035, all new cars have to be electric. No more combustion engines allowed in the state of California 15 years from now. And then they said that the larger vehicles will be 2045. So, 15 years for the passenger cars and then another 25 years for the larger cars. California has 38 million people. That is a pretty large consumer market, so it has a big impact on the plans that large companies make. Zehra this had to be a topic of frequent discussion there in California.  

ZEHRA: Yeah, but my question is why wait that long to be honest with you. I often think we do not realize that we do not have to impact the environment this much. I think they are even later than where we should have been, and I see a very similar trend in AI. Just because we have TPUs, people think that we can just irresponsibly create thousands of models that we do not need, and we should just not do that. We should just be responsible for the environment. And COVID has shown us that we can live by consuming less, we can live by polluting this. And I hope we do electric cars even earlier.  

JAMES: Yeah, it would be nice. Certainly, there are some use cases that the law of unintended consequences will drive some challenging facts, like people who drive several hundred miles a day, there is not an electric car on the market that will work for them. And there has to be some accommodation for that. And I imagine the value of used combustion in engine cars in California will go up because of the people that need those, but there are going to be all kinds of interesting things that occur. I agree. I typically like to let the market play out, but in some cases, you have to nudge the market along to move on to its next step in the evolution. And I love electric vehicles. I think they are fantastic. They are absolutely amazing. I am watching a great series right now on Apple TV. I promise we are going to get talking about InsureTech. This impacts InsureTech because electric cars typically carry with them automation packages that will drive automotive insurance rates down. And we can talk about that in a minute, cause that is directly tied to machine learning and AI. 

Great series on TV right now on Apple TV Plus, with Ewan McGregor, the famous Scottish actor, who of course from Star Wars and all kinds of other movies. He is riding with his buddy. He did 2 other around the world journeys, but this one is from the Southern tip of Argentina and Chile and Tierra del Fuego all the way up to Los Angeles on 2 electric motorcycles that are prototype electric motorcycles, Harley Davidson built just for the journey. They had 2 Rivian’s, which are all-electric truck that is built in Michigan, so Rivian delivered VIN 1 and VIN 2 for this movie. And Harley Davidson delivered VIN 1 and VIN 2 of their motorcycle. And they to Milwaukee and they went to Detroit. And they picked up the vehicles and they shipped them down to Tierra del Fuego. They installed charging points along the way, and they are doing all–electric, all the way from Tierra del Fuego, all the way to Los Angeles. 

ZEHRA: Cool. 

JAMES: Yeah, it is super cool. They ran into some pretty big problems in parts of Argentina and then the document how they dealt with it. A really fascinating view of the future of electric vehicles, but we talk a lot about electric and automated vehicles because it is going to have a massive impact on auto insurance.  

Before we get into that, remember that you can subscribe to the InsureTechGeek podcast by texting GeekOut to 66866. Make sure you never miss an episode. You can subscribe to our weekly emailer and we will send you the show notes. We transcribe those, and then we send you the link to the video. Cause we are video now, by the way, if you have not watched the video and then we have our audio episodes.  

Back to you Zehra, we are going to talk about Tazi, AI, we are going to talk about how artificial intelligence and machine learning is impacting the insurance space. Before that, I want to have a brief discussion about you. You have an interesting background. You grew up in rural Turkey and you definitely are very well educated. Went to B. Sc Computer Engineering at Bilkent University, and then you went and got your M.S and Ph.D. in Computer Science from Caltech. Just walk me through growing up in rural Turkey, a family of farmers. What led you to computer science and what led you to Caltech and InsureTech?  

ZEHRA: I had a father and mother who just believed in me. Whatever your income level is if you just let your children be and if you trust them, they will do just so many great things. I believe in that. They always told me, do what you like to do. And if you fail, we are here. So that confidence was just so great. And I chose computer engineering. I could have been a doctor; I just could not memorize stuff. I have such a horrible memory. My mom wanted me to be a doctor and to help people, but I just said I cannot do it. I cannot memorize stuff. I decided to become a computer engineer. I chose AI during my B. Sc project years because I had great role models who loved what they were doing.  

And I wanted to have an AI that was like a human. At those times there were Japanese generation projects where people were using rule-based models, to mimic humans, et cetera, and then I came to Caltech. I worked with one of the pioneers in machine learning, and his love for AI was also so inspiring. I have worked in the industry, I moved to Academia, and then we established Tazi. And with Tazi, the machine learning solution that we provide, it is such a perfect fit for the insurance industry. We chose insurance for that reason. And I can talk more about that later. 

JAMES:  Yeah. Well, we can chat about it now. Tell me what Tazi does. What exactly would you say that it does? It is an AI. Okay. Well, AI is a super, giant, broad, massive umbrella that can apply to just about anything in computer science at this point, right? 

ZEHRA: Yes. To anything. 

JAMES:  Yeah. And we are not talking about general AI, you are not developing general AI. You are using specific forms of AI, specifically machine learning and other techniques, to solve specific problems in insurance. Cause general AI, for those of you who do not know the difference between general AI and specific AI, general AI is, we are thinking like Hal from 2001, A Space Odyssey, right? You have a sentient computer that is functionally self-aware, it can learn, grow, think like a human. That is not what we are talking about.  

And whenever we talk about AI, by many people’s estimates, 20 to 30 years in the future, as far as we know it has not happened yet. And what we are dealing with right now is specific forms of AI, like natural language processing, image recognition, speech processing, text to speech, speech to text, you name it. Sentiment analysis is common, reading free text. There is a lot of different forms of specific AI. And machine learning, for our listeners, is when you teach a machine to learn, rather than explicitly programming every potential outcome.  

So, Arthur Lee Samuel, in the 1950s and 1960s was one of the pioneers of machine learning (and he was one of many people who worked on this), but he taught a machine how to play Checkers against itself. And then it used those games to learn how to win. And then it played the world Checkers champion and won. And those are the seeds of machine learning. Walk me through, what Tazi does. What specific forms of AI it uses and what platforms it sits on. Let us get technical. 

ZEHRA:  Sure. Okay. We established Tazi in 2005 with my cofounders and my husband. So, I told him, I do not want to establish a company with you because I want you to be my husband. And we might have fights when we establish this company. For the last 5 years, we did not have any fights, but we created this great technology that was inspired by the problem of adoption. So, think of COVID 19. Machine learning systems, as you said James, they learn from data. And the machine learning systems that we are talking about in insurance, they are not just differentiating apples from oranges or cats from dogs.  

They are trying to differentiate risky customers from not risky customers, the loyal customer versus the customer who is about to leave you. So, those are changing concepts. We realized that. People use a lot of data science resources, business resources to have those report solutions, and solutions could not live long because life changed, and it took as much effort to update a solution as to create one. So, we decided to establish Tazi as a self-maintaining AI system. A system that is deployed and can learn from the data, in many ways automatically, without needing lots of cranking.  

JAMES: So, by that you mean, not a ton of human intervention in the way of hinting, having people hint and trained. And that is what machine learning algorithms have become famous for is, you have hundreds of people in offshore offices, constantly hinting all these systems and then supplementing them. That is not what we are talking about when you refer to AutoML.  

ZEHRA: Yes. AutoML concept, hinting is good. And I will tell you why it is good. AutoML platforms are for data scientists. They are aiming to make data science tasks easier. What we realized after having this continuous learning in technology and deploying it in customers, we realized that it was just not enough to rock as data scientists, because insurance is such a complex business. I am not a data scientist person. I am just a technical person who is learning insurance, so I might make mistakes. Correct me. Rob is here is an insurance expert and you are here. So, I will make mistakes, okay.  

But what we realized in insurance was, as a data scientist, you might find a solution, but that solution might not be serving your return on the investment. It might not be serving the business KPIs. And the business KPIs change. Today I might want more customers, tomorrow I might want less risky customers, or I might want customers who want to use my new product. So, while those KPIs change, and while the data that the customers produced and what they want to change, it was impossible to have an AI system that worked only with data scientists.  

So, our customers started asking for, hey, I want to understand how this AI works. And these people are insurance professionals. We are talking about underwriters, pricing experts, marketing experts, not necessarily data scientists. They asked for explanations and on a Sunday morning, I came off with this (and we are using some open source tools) and I was able to interleave our continuous machine learning with explain–ability. And I still remember that morning, I am like, hey guys, look, it is so beautiful. I do not have time. I would love to show it to you so you can use our explanation model. It looks like Iron Man. 

JAMES: Nice.  

ZEHRA: We are turning every insurance professional into Iron human, in the sense that they are empowered because they can understand machine learning. And then we realized after the explanation came, hey, I understand this. I know I can make it better because. Because, you are using these features, these variables in your model, but I know the other variables that can contribute. And then, you are making mistakes here, you should not include Porches in that risk model, for example. Then insurance professionals saw how models worked.  

We realized that they could help in 3 different ways. Define very timely business KPI, and make sure that they are being followed by the models, monitor them. The second one is doing model creation, fast forward the process by pointing out what should be used in terms of data. And then the 3rd, to help the models learn better. So what you are describing was a negative thing, but what we are doing is a positive thing. For example, if I know that restaurants will be closed in a certain area in town, I might suggest some rebates or maybe even increases in insurance premiums, based on weather delivery is more common or not in that area. And models do not know that yet because there is no data generated yet for that. So, insurance professionals and underwriters can intervene with the models when there is no data. We call this batch machine learning, which is a state without insurance right now, and in machine learning, what we are doing is learning from today with continuous learning.  

And if you put those 30 years of an experienced insurance professional with the models, you get learning from the future. And that is just so exciting, like having that whole loop. Our vision is to have not just data scientists, but insurance professionals themselves using machine learning and being in the process so that models are useful, not just from the beginning, but all the time. And insurance professionals are the perfect match because their life depends on predictions. Because they had been predicting, they had been dealing with data and models all the time. They are the perfect fit.  

And insurance is such an underserved sector in terms of AI, they are ready, and we are making data scientists more efficient and insurance experts involved, so that models are working all the time. And if they are not working, you know that they are not working, and you know ways to make them better. I am just so excited. Sorry about that. If I start talking about this, I will never stop. 

JAMES: Rob? 

ROB: Zehra it is great to see you again. I  wanted to know a little bit more about your background, coming from the world of Academia and then, you talked about, you wanted to make this practical. You wanted to make this accessible and you decided to go the route of a startup rather than go work at an insurance carrier. So maybe just talk a little bit more about how you made the transition and then, I am sure in Academia because you have such a foundational and theoretical knowledge and you are very steeped in AI. You have talked about the opportunities in insurance, so I am just curious a little bit more like why insurance versus, some other sectors, which of course, AI has applicability in a whole lot of different industries. 

ZEHRA: Very good question. I love being an Academician. My advisor used to tell us, hey, I am having so much fun and they are even paying me for this. I feel the same way, when I interact with my students, teases students doing research especially, however, when you do Academia, you are more or less in an experimentation and research phase. And as Tazi, we are involved in two European Union projects, one of them is on digital twins. The other one is explainable AI. That research always has to be with a startup because we are racing against, huge elephants. As billions of dollars of those companies are at our competitors. How can we survive against them?  

The research is the only way, we always have to be doing state-of-the-art research. Our advantage as a startup, as research enhanced, researched super–powered startup is because in our team we have people who have a research background and also applications background. So we can come up with an idea and bring that into being a part of our solution in a very short time. We can listen to our customers; we can identify what the problem is and how we can solve it as a product or using what kind of AI. And we can implement that in a very short time. And we can do that only because we have research as well as development. And the excitement of seeing something that you do is used by an insurance professional and getting that praise is just so amazing.  

Writing papers as an Academician, doing projects is just so awesome. And I have done very rewarding work. I have worked with students with ADHD and Autism, for example, I have done very interesting stuff, but having what you have done, being used by people to improve their business processes, to improve their understanding of their customers, and seeing that light is just so rewarding. So that is why we did this transition from Academia to being a startup. And I did work in huge companies. I had worked at Siemens Corporate Research. It is hard to see what you are thinking, being a part of your product and being used in a large institution.  

And the start startup gives you exactly that. You can see the problem and solve it in a week or so, wow! This is super empowering. I just cannot exchange that with anything else right now. In terms of insurance, it took us such a long time to find insurance. We started with telematics actually. We also thought about the gaming industry, like computer games, and then it was just hard to develop business and to find real business value there. With insurance, on the other hand, we found a great traction. So, we have two customers in insurance, two European international insurance companies. And for them, they signed up right after COVID, and we have been able to increase their revenues because of our customer retention solution. 

So there are such great traction and such a nice fit, in terms of the difficulty of the business and enhance the need to include business in the solutions. And also, the continuous change, enhance the need for a self-adapting model. For that reason, it is a great fit, but insurance is the first industry that we are solving problems in. In finance, for example, manufacturing, these are all very interesting. And health care. These are all very interesting other verticals, but we are a startup. We have to concentrate. And in insurance, we found good traction and a very nice fit, so we want to keep doing this. 

JAMES: Yeah. Zehra, let us break it down into simple terms. An insurance company signs up with you. You have auto insurance, profitability, and growth, you have profitability and growth detection, and you have claims prediction. And then you have life insurance claims prediction. Those are the three main service lines you have listed on your website at least. Walk me through, I am an auto insurance company and I call up Tazi and I say, please come help me. What is it going to look like for them to work with you and your tools, and then what is the result for them? Let us just talk about outcomes. What is the problem they are experiencing that leads them to talk to you and what is the outcome when they are done?  

ZEHRA: We did more than 30 use cases, and we are listing all our use cases on our website. And those are use cases with customers. But right now, we are concentrating on especially auto insurance and for auto insurance, customer loyalty. We are talking with especially mid-level insurance companies. And there are several pain points. One of them is customer retention. So, they also have problems with customer retention, especially post COVID. Unfortunately, COVID is not paused, so during COVID, they see lots of customer retention problems. We have an initial conversation. We show them our demo, and then and business always has to be included in the process. We cannot just be talking with the BI teams. So, we had this 3 x 3 solution, if your data is machine learning ready in three weeks you have a solution, not just a toy model. We have a solution in three weeks where you can see the results. There is predictability, and you can also see that there is a return on investment, because the return on investment requires some a prescription.  

I cannot just sit down and cry if my customers are leaving me, I have to have some set of actions. What can you do if your customer is leaving you? , Well, you can see if you can give some rebates, as most insurance companies are doing right now, or you can see if you have new products to offer, you can see if there is a mismatch between the agency and you can also take a look and see what your competition is doing and see if you can offer even new and better products. We can keep the pressure of the customers with the customer retention model and then the actions that can be taken, can be taken by the upsell, cross–sell models, price, and assistant models, and so on. So, in three months you would have a deployed solution with the ROIs in your hand and that is what we are offering. 

JAMES: And you have a deployed solution that helps you make better decisions on claims when they get intake? You have an auto claim that is intake and you can set the reserve better? Is that the end outcome, is you are going to do a better job on reserve setting? 

ZEHRA: Well right now, we are using claim risk prediction as a step in our customer loyalty solution. Our claim risk predictor, it just aims to say whether Zehra is a risky customer or not. So only if the risk is less, you should be offering stuff to your customer. Because if the customer is a high-risk customer, then maybe you should rethink what you need to offer. I will give you an example in terms of customer loyalty. If the customer is going to leave you, but other things that you can do, I will ask Rob, because you guys are the insurance experts more than me. So based on a different state, if you are in Michigan or if you are in California, there are different things that you can offer to your customers if the customer is going to leave. So, what are they? You cannot always give a reduction, you cannot always give a rebate, right? Or you cannot reduce the prices because some of the things are regulated, but the customer retention model tells you why the customer is leaving you. Is it the price? Is it an experience with the product or is it the agency? Those are things that we have seen so far as the root causes of why the person is leaving you. There might be other things that you will see, but those seem to be quite prominent right now. Does it answer your question, James?  

JAMES: Yes, absolutely it does. We are trying to understand because I think this can be mystical for some people when they are trying to understand exactly what value it adds and why should they use it. It is not just for marketing. Like there is a real practical return in value and utilizing machine learning to do things that humans just do not have the time to process. Rob your thoughts? 

ROB: Yeah. And just to add to that. This problem is very interesting, particularly in customer loyalty. Based on previous experience, I have learned a few things over the years. Number one, the people that complain, are not always the people that cancel, which is very interesting. And so oftentimes companies will spend a lot of time trying to help those that complained and take care of them. They do not want them to leave so they will threaten to cancel, but actually, those are not always the people that cancel. Sometimes the people that cancel, just walk away and you never get a chance to talk to them, they never complained, they are just gone one day. And you do not know that. And knowing that those are two different profiles, that the people that complain are not necessarily the people that walk out the door, I think is important to know.  

The second part is, to Zehra’s point, this is very crudely defined, but you have loyal people that will be with you unless A, you price them out, or you raise their premium enough and it does not just raise it but raise it by a certain amount. Generally, 10% or more I think is what I have seen, or they have a bad claims experience. And unless you do anything to drive them away, they are going to be with you because they do not want to swap their insurance. They are generally happy, whatever. Then you have people that are always looking to save a buck. They are constantly churning. For example, when I worked at USAA, one of the things that were tough to get your mind right around was, we lost the most members in auto insurance to Geico. And we were like, Geico is a tough competitor. The Government employee insurance company, they were started by a former USAA employee way back in 1936 so that was alarming. And then if you say, well, who are we taking the most customers for? The answer was also Geico.  

And what you started realizing is that, oh, there are people that churn, that switch every six months and go from one company to another and you can spend a lot of money chasing these people. We know customer acquisition costs are super high. It costs six times as much to acquire a new customer, if not more than it is to retain an existing customer. And when you start chasing these people that are going to churn anyway, you end up spending a lot of money, whereas those loyal people, you take them for granted. And again, those are the people that may be walking out the door tomorrow that you do not even recognize. So, understanding the nuances of those patterns, being able to leverage the AI that Zehra is talking about, I think does have tremendous value.  

So, Zehra, we met in Istanbul last year, at the Global InsureTech Summit and you were part of the InsureTech hub. We were one of the startups that presented there. And then obviously you have moved from Turkey to California. Earlier this year we even talked about, where is a good place for you to locate, and obviously you have a history in California, which is great. And I know that you are also a part of the Alchemist Accelerator, in Silicon Valley. So maybe you can just talk a little bit about that journey, the startup community in Turkey There are several great startups are there. Why you decided to target the US market. You have talked a little bit about it, your success in the EU market. So just tell us a little bit about that journey over the last 18 months.  

ZEHRA: Sure. By the way, I just love California so much. We always wanted to be here, but I guess there is always a time for things, and this was the time. Silicon Valley is such a great place to be as a startup and if you can do it, you should be in Silicon Valley as a startup, you should not be in Chicago, you should not be in New York, you should be in Silicon Valley. I am so sorry.  

JAMES: You should be in Texas Zehra! I hate to break it to you. 

ZEHRA: I spent a couple of months in Chicago. Come on guys, from the mountains, this whole valley is pulsing. Even though there is COVID.  

JAMES: Yeah. 

ZEHRA: You are just like, oh, I am a product manager. I have done installations to companies globally like you can just talk with one or two people and you will find the team in such a short time in Silicon Valley. No. And being a startup, you need a great team. And it is all about human power because we had a product, but the product happened because we worked with great people, and Silicon Valley has the best resources in terms of professionals who know how to create a product, how to sell a product, how to make a product useful. So, for that reason, I am in California. Robbie, Danielle, those people, I just do not know how to thank them because in the middle COVID, because everybody was home, it was just so hard to reach out to people, to investors, to customers and their mentors. The connections involved. We have access to almost 30,000 people who are great experts in very particular things. And I can have a meeting with them in just a week. Within a week, they will give some time so that is just such a huge opportunity, I could not find it anywhere else. Like nowhere could I find that so that is why we are here.  

And we are seeing lots of interest in insurance companies, especially the mid-tier insurance companies. Because, while the big guys are saying, I have hundreds of millions of people’s data, (and I guess some of the customers might not be present anymore) but I have a hundred years of data, I have done 150 PhDs in data science. And they still need continuous learning because they do not have it yet. But the meteor companies, they are bleeding out like. They need a solution yesterday and we can deliver that in such a short time. And now our deal is clear. Now give us three weeks. If you do not see anything, money–back guarantee. And I am learning how to talk that talk as an Academician, so I am learning. But there is a need and there are such agile and wise insurance professionals out there who know their customers, who know the local customers. This is like, Amazon versus your local shop. And they do know what the customer needs, and once we had that matching, I am sure we are going to create great value.  

JAMES: Bold words from a new Californian. We will have to challenge you with the sheer number of defections to Austin, Texas from Silicon Valley, but that is okay.  

ZEHRA: I have friends in Austin, Texas. I have friends in Arizona, and I love those places too. Sorry, I did not mention those places.  

JAMES: Look, I love going to Silicon Valley and it is wild cause you are just driving. Even just driving up and down the highway, you are like, oh my gosh. It is literally every tech company that you work with. 

ZEHRA: It is like being an RDM concert. You can take it a little bit to go to the cost, get some rest, and then come back.  

JAMES: So Zehra, you are focusing on the insurance business, you have auto and life. Let us talk about two years from now. What is the big stuff you want to work on in the next couple of years? Let us finish up by talking about the future. 

ZEHRA: Sure. So, first of all, insurance will be disrupted. It is being disrupted, divert things like UBI, connected mobility and those were things that people talked about and we were just thinking they would be here one day, and they are here much faster than they would have been if COVID did not happen. I think the insurance industry will be disrupted and we are going to have more interesting data, and we are going to be facing a lot of different issues that we have not faced before. Our claims process will be different. Our customers will require us to insure at different times and in different places, and even under different scenarios. Like, think of Uber drivers‘ insurance. Think of the scooters.  

We will be able to compute risk for a person who rents a scooter somewhere and then drop stock at some other place. And hopefully, COVID will be over and tourists will be back in New York. Another tourist will pick up that scooter, and we want to ensure that tourists as well. We will need to ensure people under very interesting scenarios with different vehicles and with different people. We would need to create new insurance solutions for insurance, and I am hoping to be a part of that. I believe our ability to make AI or machine learning a commodity product, you know there are so many engines that are in this room right now, they are doing their work and I do not even know their presence. But somebody put them there. The insurance professionals will find the pain points in their processes, And our aim to be able to help them answer their questions in a very short time, as opposed to months, in weeks, in days, in hours, we are going to be able to create solutions for the predictive problems and they are going to help make better decisions. 

Another dream I have is enterprise–wide AI for insurance. Lots of data mean privacy and security issues with data. People do not want the data to be shared. The problem with UBI and telematics is if you get my data, how do I know that it is not going to be used against me, but for me. We want to have sharable models as opposed to data, across different insurance business units for sales and marketing, for finance, for underwriting, there will be different models and those models will talk with each other as opposed to data. And that is a huge vision. This will mean communication within the insurance company, without being scared of who shares this about data. Because models are much easier to explain. They have much less risk than sharing data because, with the models, they know what they do. If you have explanations we can model, great. And not only within insurance companies. This is just breaking down the silos, like having much clearer communication for the benefit of the company, not just for your department. For the benefit of the whole company. Sharing models is my vision, and another vision is to have users and know what they are sharing. 

So, if I have my sensor data on my car, maybe I do not want the sensor data to be shared. I just want some predictions about my behavior to be shared. And I have control over what is shared with me, about me, with the insurance company. Things like H computing and having more lightweight AI, more responsible AI. These are all things that we are addressing right now. We have the technology to address it. Continuous learning and explainable, AI, and human involvement, those three technology advances, we lead to huge things. And those are just some of them that I can think of right now, but people like you guys, I am sure you have more motivation than I do. I am just learning insurance and you guys are disrupting insurance.  

JAMES: We are all in the same boat together right now, but it has been a really, great discussion. We do need to move on to our news and Rob’s got some good news stories. I have got a couple and, so Zehra, just sticks with us for a second. We got to cover this and then wrap up. Rob, what do you have? And by the way, Zehra, thank you so much. Awesome conversation. Really, enjoyed this. And we are going to just have a quick chat with you about what is going on in the news. Rob, what do you have going on with this week? 

ROB: Yeah, I am really excited about your vision for the futures Zehra and what you guys got going on at Tazi. Thanks for joining us. So yeah, a couple of quick news hits from the week. Probably the biggest news item this week was the $250 million Series D for Next Insurance. Congrats, the Guy Goldstein and the folks over there. Big news in Crunchbase and other outlets. And there was some talk about following in the footsteps of hippo and talking about the Lemonade IPO. We have seen this movie before but yet another success story for InsureTech. So definitely check out Next Insurance. 

And then, there was a report in the Insurance Journal this week, that was neither confirmed nor denied by representatives at MetLife, that they are looking to sell their P&C auto and home insurance business. And this is a big deal. Obviously, MetLife is big in the life insurance space, as you might imagine. They recently acquired a pet insurance company. I do not remember exactly how it works at the beginning of this year. It is interesting. Hang in there, you are introducing pet insurance and you are potentially looking to divest your auto and home insurance book. If that happens, it would definitely be a big deal in the marketplace cause, they are not a small player by any means.  

JAMES: Yeah. And on my side, since you mentioned Lemonade, their stock price has dropped pretty precipitously from their peak. It is interesting to see what is happening. Of course, tech in general has taken a hit. So, you might have to factor that out. Their stock peaked at $85 a share and it is currently trading at $50 right now. And so, they are down about 35% to 40%, depending on the day from their peak stock valuation. Of course, some of this has to do with financial guidance they are issuing, and they are of course finding the brutal reality of being a publicly held company. There is a focus on this thing called profitability, which is really important to the public markets in particular and difficult times. Now, I would say you probably need to consider that a lot of tech has been hit fairly hard. And they may just be getting lumped in with the rest of tech. I think investors are pretty savvy with Lemonade and understand that there is going to be a much bigger play coming in the future. 

I am also in the news cycle, of course, Root Insurance was tipped by Reuters to be prepping an IPO. I am a big fan of Reuters news, and so you will see some more of these. And what is going to happen is we are going to continue to see some market validation on private equity values a company is at versus what the public markets are going to be willing to value an InsureTech company at. And that is an important cause that is going to backflow into the private equity markets on the valuation deals. If the public deals are not supported, then you can imagine valuations will come down in the private market.  

Also, in lemonade news. Lemonade is entering France. So just a little side note, that was reported on the 15th of September. That they planned to enter France. They are going to keep pushing for global expansion. Obviously, they have to keep pushing for revenue growth. So again, we are rooting for Lemonade. We want them to be successful. We are talking about this. Remember we are rooting for all InsureTechs to be successful. We want the evaluations to be high. We want the market to support it. We also want companies to be profitable, right? Profitability, especially once you go IPO. You have got to reach profitability fairly quickly or the market punishes you. Please understand the context of this that we want Lemonade to be successful. We want Root to be successful. Want these IPOs to go well because it is going to bode well for all the funding rounds that are occurring in the private market as well.  

So that it is just some interesting news around InsureTech in general because there are quite a few other companies that would like to IPO in the next year to year and a half and the market needs to support that. That is just a little smattering of news that I saw as it relates to InsureTech. Pretty good week, obviously in broader news. What isn’t it news these days? I  would almost call this like an over-hyped news cycle. Everything’s big news. There are no boring days. There are no slow days. It is a very interesting time. We have a Supreme court nomination going on. We have more police news. And then, obviously, there is a presidential election. Of course, there is a lot of local and state elections. And I want to remind you that a lot of state governors appoint insurance commissioners. And so, with political change in turmoil comes change and turmoil in the insurance markets too. Do not feel like you can just stick your head in the sand and ignore who is getting elected because that the people that regulate the insurance markets get appointed as a result of a lot of these elections. And that has a big impact on our ability to expand and write business and build technology and function in these markets. And so dynamic time right now. I think that is the nicest way to call it a dynamic time. But certainly, a lot to follow Rob.  

ROB: Absolutely. Yep. No doubt about it. The election will be here right around the corner and you are right. I am glad that you call that out, obviously ton of focus on the presidential election to understand them. But, yeah, those state and local elections matter. I know you were previously, on the city council there in College Station and yeah, it is interesting, definitely appointments, but there are also elected insurance commissioners. I know in California, for instance, has an elected insurance commissioner as well. So yeah, just encourage folks to get out there and vote and to make sure that you do your homework and researched all the races. Cause they all have an impact. And, obviously, new voting rules in terms of whether it is, mail or early voting, it depends on what state you are in and all that. I think you have to spend a little extra time on what the rules are and your location. It is crazy that, October is right around the corner. We are already in Q4. I keep telling people that, the year is slow, but cannot get over Zehra enough, but the days are going by fast. I think that is a reality for most of us and 2021 will be here before, it.  

JAMES: Yeah, absolutely.  

ROB: Zehra, what is your opinion? 

ZEHRA: Yeah. I did work on stock market prediction as an Academician and I know it is just such a hard thing to predict, but I want to comment on profitability. We have worked with a gentleman from Japan. Actually, our first insurance customer and he helped us so much in terms of developing our product. And he once told me, if you are going to do a fraud model forget that about their fraud model like you are not going to have labeled fraud data. You should look at your profitability model. We developed a profitability and growth micro-segmentation models for detection and also for prediction. So as an insurance company, you need to find out, while things are happening, not two or three months down the road. While things are happening, you need to figure out where are you exactly losing money. If you say to a company, oh, you are not profitable, you are not saying anything. 

I was once talking with a psychologist on autism and he said, if you tell a parent, your child has autism, there is not much that they can do. But if you tell them, your child has difficulty communicating, especially verbal communication, then they know what to do. Insurance is just like that. If machine learning can, our models can exactly pinpoint where you are losing money. And you are losing even more money in time. And when you know that you know where your problem is whether it is a pricing problem, whether it is a competition problem. I did think companies should look at where they are losing money, where they are seeing losses with AI. And try to find out solutions that will solve exactly the pain instead of just taking general measures. I just wanted to say that. And I am so happy with the investments, obviously, lots of investing is needed if you want to do all technology, and I hope that Lemonade and other guys all do well. Because I think insurance and finance, they are both very important industries for continuing our life, basically. So many things depend on them. So hopefully things will get better 

JAMES:  Awesome. Well, thank you for being here today. Zehra I appreciate you. And Rob as always great to see you.  

ROB: Likewise, James. 

ZEHRA:  Appreciated to be here. Thank you so much.  

JAMES: Thank you. This has been the InsureTechGeek podcast powered by JBKnowledge. It is all about technology that is transforming and disrupting the insurance world. I have been your host, James Benham, JamesBenham.com with co-host Rob Galbraith, EndOfInsurance.com. Thanks to Jim Greenlee, our Podcast Producer, Kara Dalton-Arro, our Creative Producer, and Adéle Waldeck, our Transcriptionist. 

Thank you for joining us today.

We are taking you on a journey through insurance tech, so enjoy the ride and geek out.

See you next time!