From Chatbots to Claims Agents: Inside an AI-First Insurance Startup

**[00:00:00]** Hello everyone. My name is Arando Prescareno and welcome to the web talk show. Today with us is Juan Garcia from Tuyo. Welcome Juan. How are you? >> I'm all right. I'm all right. Thank you for having me. >> Oh, it's a pleasure. I want to talk about Tuyo and what it means because from what I understand you built the Spain's first AI native insurer. So what does that mean? Let's get started with that. >> Well well that's that's the only latest iteration from the company. We've been in market since 2021 because we just realized that there was Disney in the market that there was this customer segment from 25 to 55 give or take. They're not uh very strict uh frontiers. Um we figure that that this segment was underserved from current offerings because they current insurance offerings at least in Spain but probably almost everywhere in the world are just very cookie cutter. So you just have this same product doesn't matter the customer doesn't matter the age segment. you're going to miss status is this is what you have. You either buy it or you don't. And the on boarding process is the same for every customer. And and it just doesn't work for this segment. When that's what we started to see because now we're used to just buying online. I mean, we consume content through Netflix. We purchase our stuff on Amazon. We even buy our groceries on online, right? So, so when you actually have to call um a call center or you actually have to go to to a store to have just just have your homeowners and we started the homeowners uh insurance piece of insurance policy uh sold to you. It just didn't make much sense for **[00:02:00]** us. So, so we started looking into into this customer segment and and then we realized that it was a problem both from the customer side but also from the industry side. um we started looking at the profitability of different age cohorts and what you what you would look at is that older cohorts 60 plus they're pretty profitable for companies and I'm sure this is the same everywhere in the world these are customers they see price as a proxy for quality these are customer I mean we all see price to a given to a given extent as as a proxy for quality we these are customers that don't use the product as much they are not as careful to see um what the product entails is they're more trusting from brands whereas if you look at the younger goes we are learning a lot and I'm say we because I'm obviously one of those you are as well um we areing a lot we tend to browse and compare on price comparison websites when we're looking at insurance and and that put pressure on prices because you're going to charge as much as all the cohorts um we also since we're comparing a lot we we tend to look at the trust pilots of the world and to look at reviews and and then we tend to build financial literacy through comparison. So we're pretty aware of uh what the policy in this case entails and and what are different coverages and what's covered and what's not. So since you know more of what you're purchasing then you also know more of of what it's covered and you use the you tend to use the product more the product um prince races larger right so this **[00:04:00]** also puts pressure on your bottom line because you you have to pay out more more more claims which is what the products for don't get me wrong but still you get >> this result that you get one customer segment that is not happy by the way you're treating them and you get and you get no incentives to invest in this in this in this customer because well the reality is it's not profitable for you because it puts pressure on your top line and your bottom line. So, so we found this this great problem to work at and that's how how we built Toyo because we had this insight that if you give them uh self-service deal based platform uh to buy and and manage and and just do everything related to their basic insurance needs and the basic I mean we started the homeowners then term life because obviously it's very close to homeowners through mortgages and everything else then we we look we looked at pet health which is a a growing very growing business line and then we are looking at auto and probably travel by the end of the year. So, so yeah, this is this is us and and since we we when we were building our our diesel first platform, then all the NAI about a couple of years ago, which seems like ages ago, but then GPT only became really known a couple years ago. >> Then it was the new and best technology that came over. Then we we started thinking, how can we do better with this technology? And and and that's that's where we're here today, just showing up and our message to the world. That is great because I have seen that experience where shopping **[00:06:00]** for insurance wherever you are. It's like, okay, so I need something for the car or for homeowners and I have to call someone and so I have to wait on hold or have someone call me at an maybe inconvenient time and then to see if I like the person or not and then you go through the whole process and and they have to go and check something and get back to you and you don't know if they really checked all the choices or they just gave you the ones that were easiest, right? So I like the online approach I used. I remember when we were renting, we I found Lemonade or I heard about Lemonade over here. I'm like, "That sounds interesting." And I went in and everything was done in an app or something and it was so quick and so easy that I expect that's sort of the type of experience that you're giving your users and as a user it's very welcome. >> Oh, for sure. It's um Lemonade is obviously one of the inspiration that we took to bring to bring that type of experience to SP Spanish market but in southern European market and and um and yeah it's one of those things that you don't even know how much you are supposed to pay for a for a piece of insurance right there's not even there there's even there isn't even any transparency in the market so that's an additional issue with with the product right it's just like why I cannot know what's supposed to be cost um before I even talk with anyone but it's I think it's the only real market that you have anything like like this happening right like it's online it's products are **[00:08:00]** not online really so it's just yeah for us was yeah for us was very surprising >> it is it is very surprising it's like well insurance in general right healthcare and all that as well but the it's yeah when we talk about this I we find it very ridiculous that you can't just give me a price or I can just look up a price that makes no business sense why Why would I? And so I really like uh from your website it it says like in Spanish I see verio right like I want to see my price and I click it and then it'll start this I guess it'll start like a flow, right? It'll ask me some questions and things like that and then eventually I expect I would get some sort of pricing upfront, right? I like that as a customer. What >> how how can you manage that? Like how can a company actually manage something in an industry where everyone else is doing it differently? How were you able to jump that hurdle to be able to offer something that nobody else practically was offering? >> Well, so there's a few things that that we will we were doing here and one of them was as I said is trusting that people know what they're purchasing. Um there is there was there was this mantra in the industry at least in Spain when we started that that said they used to say that um industry is not bought sold right so there there is this mantra that that to buy a particular piece of of insurance you just need to go through brokers and and you need all this all these complex products and you can put you can put coverage **[00:10:00]** you can take out coverage and and we started from the thinking that I mean if you need a harmonous policy it doesn't have to be that complex. You just have two flavors at most um with all the traditional coverages that someone uh may want. You explain it very carefully and very transparently and you give price and that's it. Um and there will be customers that won't be for you and that's okay because if you are designing for this particular in our case age segment of 25 to 55 which is not really an eight segment it's just as I said before the proxy for an online behavior then you know what you can show them. you know how to show them or at least you're continuously iterating, show them and then you show the price and you hope for the best and that they like your product and they end up purchasing it. But you that that's that's the way any online business works, right? So I think we brought >> this uh product based kind of thinking to insurance that at least in Spain wasn't wasn't very successful before because what you would found is that you would have the same offline type of product just put online and that doesn't work because there's too many options too takes a lot to to purchase a lot a lot of questions and and we are another thing that we that we brought in we we simplified the on boarding process just pulling data from different public data sources. So you have to be you have to be a little a little bit creative and a little bit forward thinking or at least has come from a technology background like like a product tech type of business **[00:12:00]** just to implement that in insurance because I mean in the end it just it seems very obvious but it wasn't really at the time as it wasn't when Lemon is started for example in the US. So, so you coming nowhere from a tech side of uh a product product based tech tech side of of the industry. They just applied insurance and it seems very easy to improve on what was going on before. But you had to do it right. >> And of course, but also how I mean you have to have the creativity, the knowledge in the industry of course, but also the daring to go into an industry that is so regulated and so controlled by a few specific participants. Yeah, one of one of my co-founders was Maria, he he says all the time that um you have to be an entrepreneur, you have to be very courageous, but you also have to be a bit naive >> just to you have to to have courage to just jump into trying to solve problems that nobody has solved before. But you also have to be naive >> to just not be aware of the actual complexities of it all. You just try just go in there and then you just stumble into the different difficulties as as you go along and you try to solve them. But if you're looking at if you have complete knowledge in of what it's going to entail in the next two years, you maybe don't start. So that's why you see a lot of at least from us, none of us we were aware of insurance, but we weren't from the insurance um industry. I was I I was working in in tech before and I built **[00:14:00]** an insurance business for a telecommunications company when I was working when I was working there. So I just realized that there was a lot of issues but I I didn't have that in-depth knowledge of why things had to be the way they were. And Jose said he was consulting for for insurance and and assist our third co-ounder the as he says the the geeky of the three he was working as as a model modeling and and and in data but we didn't have a complete end to end view that someone that's worked in insurance his whole life would have. So some of the issues that we were that we enc encountered along the the way we just didn't even imagine them right and that's why we started because if not maybe we have wouldn't have >> I like that I like the concept of needing to be a little bit naive. We were touching on that on a recent episode where we're talking about AI in the enterprise and in business and things like that and where it is great sometimes to let the AI to put it in the same context be naive in the sense of we were talking about the example of having a let's say an outreach campaign I create I was working on an outreach campaign and it was going through businesses in a certain city and things like that, right? If I see the list, I might see a specific museum or something that I like and be I don't know, maybe I won't send that one, right? But the AI doesn't care because it's being naive. So, it'll just go through the whole list. And so, that happened and it was eye opening for me. We talked about **[00:16:00]** it in the other episode where I start getting replies from some of these organizations and I said, "What? I sent I didn't like the thing sent, right?" But I it was very good to see. Wow. Yeah. Sometimes you stop yourself because you know things and you shouldn't especially being an entrepreneur. Oh for sure. For sure. That's the actual same example. I mean we're not I don't think we are more intelligent than anybody else. I just think we were courageous and naive in 2021 when we started. That's that's basically it. That's some also starting anything to do with health in 2021 is audacious to say the least. >> Yeah. Yeah. Yeah. I remember vividly how my mom told me like so we are not outside of COVID and you're telling me you are starting a business online and we don't even know when we're even going to be I mean we were still um within our homes when we we decided to start it. So I was like, "Oh yeah, well yeah, I mean we may as well." >> That's amazing. And you serve Spain specifically or you mentioned Southern Europe. >> We we serve currently Spain. Um personalized as I said before, we are looking to expand into Southern Europe. We see a gap >> in southern Europe that we didn't see in Northern countries in within Europe. If you look at the UK, you look at the Netherlands, you look at the Nordics or or Switzerland or Germany. um they are there are companies similar to us competition I mean there so there's the traditional players like in Spain or southern Europe but there's also a new breed of players um in all those different comp the all those different countries because the access **[00:18:00]** to capital is better it's larger um in the insurance uh business as part of GDP is a larger chunk uh I think the UK if I'm not if I'm not wrong it was like double the penetration from Spain So it's just they're more developed in terms of of actual purchasing insurance and and when you look at southern countries we we're just not um and it has to do with with uh the number of policies but also the prices and the premiums. So, so it's just like what what if you look from an Australia standpoint, you wouldn't start by this the countries in around the Mediterranean. But for us, it's an opportunity because we are from one of those countries. And then you look at Portugal, you look at France, you look at Italy or or Greece. I mean, for us there are great opportunities because there are only the traditional players, right? So the competition is going to weigh less. So So we think even if we think insurance is not a winner takes all type of market, >> Mhm. It's always better just to have less compos competition. >> Right. Right. >> So we will be we will be looking into expanding starting from the second half of this year probably. >> Very nice. What does it take to run an insurance company? Are most players the insurance company or are some aggregators of sorts where they they grab from multiple companies sort of as a broker to get things like just for the listeners and myself we have no idea how that industry works. So if you could tell us that would be as well. >> Right. Right. We we so we so we do a bit of everything. We are not a **[00:20:00]** broker. We are what's called an agency here in Spain and and in Europe. there's this regulated figure that you can you can basically do as much as you want or or you are a able to negotiate with a with a rinser um from the value chain and we we're lucky to have great partners in aliens and um CISA uh to do almost everything basically because we we believe that to be to to build our model you have we have to do everything because it is a closed loop from claims towards marketing and and underwriting. So it's just like having the whole view of the business helps us to improve it non-stop in a closed loop type of fashion. >> Um so what we are is we do all that but the balance sheet it's it's basically our partners so we structure it financially. So um so we pay them a fee obviously to use the their balance sheet but um but that's what we are that's how Lemonade started as well. So this this is a a figure that I think it's worldwide just to >> further um further innovation. So how it started as figures that when when an insurance company didn't know a particular potential risk for example and one of my partners always talks about um um weighing um whailing boats. Yes. they would basically go to someone that knew wells and they knew all the figures, all the numbers and they could with their under underwriters could figure out how to price that risk and that person would run the business for them. So it's basically the same. It's that that's how I started and we've we're using them rather than just to price and run a particular risk. We're using **[00:22:00]** them to just improve on the way we price and risk a particular customer segment. That's very interesting and I see that like you can see here for example there's Tesla has Tesla insurance for the cars right but if you get a policy someone gets a policy with them you'll realize that once the policy is actually generated regardless of you interacting with Tesla the company Tesla insurance they are underwriting it via someone else right or there's there's layers to it right and so that makes sense like you are providing the face the experience to the person and maybe even the uh management of the account etc. But at the end of the day, the the big investment of of like who handles all that is one of the the major players in the space is that >> it depends. It depends. It depends. It's it is um and as I said, you can negotiate as as much to to absorb as much as of the business you want. Some some players will just only distribute. Some players will design the product and distribute. Some players just like us, we we do we do the full end to end. We just just design the product. We distribute it. We we manage all your claims and and we even do the customer service, right? So, we do end to end. It's just um we the only thing we don't do is it's we are not a fully licensed insurer because >> for companies is more like ours. It doesn't make financial sense. So this is I mean the the the return profile of of a full insurer because you have to a business spend you have to have this this liquid money just resting on on a bank **[00:24:00]** on a bank account just to make sure that you can you can actually handle claims if if something weird happens or too or massive happens. Mhm. >> So that just having if you raise for example I don't know € 10 million and then you have to to just keep on a separate account €4 million to just ensure that you have the liquidity which makes sense from a regulation standpoint to make sure that you are able to cover all the different risks for a startup it doesn't make much sense and that's why Lemonet started this way we started this way and and a lot of other companies hippo in the US as well started this way so there's a lot of companies start this way and then when it when you get past the critical mass then you it makes sense to to just to internalize everything and just to build up your own balance sheet and your structure financially. >> That makes sense. Okay. So you you can go that route. So if you want to as a company that's working like yours, you can then decide okay we have the critical mass it makes sense now to become a full-fledged or something like that. Super interesting. What is the what we briefly touched on AI before. How does AI fit into the picture here in the insurance space? >> Right. Um, so we've done a few things. We started like as everybody else and I hope and I think it's mostly everyone not not just in insuranceances. Everyone everyone starts dealing with AI as a matter of perceiving cost to self efficiency. It's just efficiency. This is like you how can you reduce operating costs for your business? And and we did that as **[00:26:00]** well. We started with a what we would call now a text agent and at that point in 2028 in 2024 we call a chatbot. It's just now it just do way more things than than just a traditional chatbot. But then we had this um sorry in 2023 we had this epiphany in a way in 2024 when we just looked at the cost structure of insurance you see that cost to serve like the operating costs of the business is just um 10% of your cost base. If you look at marketing, look at investment cost together. If you look at marketing and claims, it's 85% of your business of your cost base. Sorry. So if so even if you can reduce your operating cost by 50%. You're only getting what five percentage points of profit margin. I mean, and that's given that you're able to reduce 50% of a of an already streamlined business that's been running for a long time, right? So when you're looking at that 88 85% that's um marketing and and claims then you start thinking differently about about u the way to deploy AI because if you just can improve um on that just 10% you you learn more you're even you're you're even improving on that five percentage points for your for profit margin. So, so we're started to look at that way and and we figured out that instead of chasing pure cost to ser efficiency, we should be looking at how to use AI to make better decisions and and for us making better decisions is just improving on the allocation of that 88 85% of our of our of our cost structure and and and we're looking at three levers. We're looking at growing more efficiently. How we **[00:28:00]** do how we can invest in marketing better >> then underwriting is smarter. How can we get all these different pieces of data and we can use it to just price better and operate better? Because it's not only pricing. You don't there's still thing there's things we do with underwriting that doesn't always reflect in price. It's just uh we we flag customers and for all sorts of things and and then we use use them when we manage claims and then how we can manage claims more efficiently. So more effectively, sorry, not efficiently. And that's how we started looking into this these things and we we've built an agent for each one of those those levers or not not only agents as as we use all other AI technologies and and that's what we think it's neat that we were able to figure that bolting AI into an existing uh insurance business process wouldn't work or or at least it wouldn't work to the extent that we wanted it to work. We we had to re re-engineer how our own company would work and and it was a bit traumatic even if we were small at the time. We just we just changed how we thought about the business because then AI is just not >> an additional piece that you put to make to make your business run better. AI is how we run our business now. It's just it's just core to the way we think about our business and and we start thinking and when you start thinking about that way then you figure out that all the things that you can do now will just give you superpowers. That's very true. What I really like about the AI movement is that it's not **[00:30:00]** here to replace people or anything like that like some some doomsday people say. It's a tool to help you make those people that you have be better, right? be more efficient such that it will give you more revenue to get more people to do more things better and not have people doing repeatable redundant tasks that they shouldn't be doing anyway. At least that's my take on it. But I really like the fact that now like you were saying instead of fitting yourself into a box of a software that somebody built for an industry 50 years ago, you can now mold the software to you, right? or the tool or the AI so that it helps you run your business better, more efficiently and help the whole the whole ecosystem, right? And I would like to hear some examples on your side, but I can see it clearly on the software side where just day-to-day activities like somebody can call in with a bug, right? And they can say, "Oh, this thing isn't doing this other maybe it's a course platform." and they say, "Oh, this person was supposed to be assigned this course, but it's not doing it." Um, we don't know why. We're trying to replicate it. Before it would take a whole debug process. Now, we can open an agent in the codebase, say, "Hey, for this person, this is happening. What's going on?" The thing will scan the whole code, look at that person's profile, find the fields that are making that mismatch, and say, "Hey, you forgot this edge case. It should have this there." Two seconds, literally, you have the solution. And whether you wanted to fix it itself or just fix it yourself, it doesn't matter. It gave you **[00:32:00]** the solution in two seconds. So that's like in my space, that's how much better you're doing with AI. Can you give us some examples of how AI can help in in your side? >> So So cloud is amazing. That's probably what you were getting to and and yeah, we use it and and some of the stuff they can they can do. It's just incredible. So basically we do kind of the same not the same thing but similar process you describe it just for our business processes. So for example I think claims is a great example then from claims which is where we started then we can move to marketing or underwriting. So claims um so when you think about a claim it doesn't really follow straight line. It's just um it branches out depending on the coverage the severity within a single coverage. If there are fraud signals, which you can look at, there are third parties, if they're expert, if they're repairers, there's this whole context of things that can be happening at the same time that if you were looking at automating claims with previous technologies, it would build this crazy crazy amount very complex tree because that's what that's what previous technologies would build. They would build a tree from A to B to C. The neat thing about AI is that you don't really have to limit yourself to a tree because um what GNI in particular um deals with very well is nondetermin the nondeterministic cases and and I mean a claim is just as I said before is just it's just a crystal clear case of a nondeterministic business process. So what we when we're playing with it, what we're looking at, so what you would think about in **[00:34:00]** a claim, even if it's not a tree, what you could see are stages, different stages. So you can you get different inputs and you get you can you can use for outputs and jai is is something that does this type of task very well. So you give you give that the tool the the machine the agent in this case we we call it Watson because I mean the helper of Serons because he's obviously investigating with this mag with this with this little little magnet um all a given claim. So you feed all the all the all these context um from a given claim and and we we use it in in jointly with our claim process. So to file a K in Tuio, you have this deal process. It's self-service. Again, you don't have to call us. You can call us, but you don't need to. So 85% of our customers just file it digitally. You just take the video or you take some pictures and you can you can even send us a a voice note about what's happening and explain everything and you get all these different files and and if it's a if it's if it's something that you got stolen, then you send us the the police report. So there are all these different um things that you can send us. And once we get all that, then you you just file your claim. So we we make Watson go through all of that through the videos, through your explanation, through different pieces of data that you've sent us. And then then and there it pulls also from a customer like the context of the customer. How long he's he's been a customer from us, all the different interactions that through **[00:36:00]** through phone or through through chat. Uh if he if if there if there is any payment extent that he hasn't uh been able to fulfill. So you get all these different data and then you get uh for example atmospheric data if this are water damage and and you get the customer navigation patterns things that you wouldn't think they're relevant but maybe they are. So you get all these different pieces of information and this is something that GNAI technology I mean LLM do very well they go through all that and we have these manuals on how we think at every stage uh claim should be treated. So he goes through all that he gets a manual and he give us recommendations of next steps. So we basically built a next best action machine and and if your listeners are aware of marketing this is something that has been built for a long time for marketing but we've built that as a core principle of how we build uh deal with claims. So we got this Nestnet action machine and then he proposes actions with a confidence level on and we we decide automating on how confidence the machine is with these recommendations and um se there's several recommendations like for example uh updating reserves reserves is the the pro projection that you think a given claim is going to cost you. So this is something that doesn't have any customer impact. So to just update that it's just you just need a very low level of confidence. So you can do it very easily >> for and for a rejection for example we just don't do that. We just don't we just don't automate rejections ever. And this is a matter of principle this >> a **[00:38:00]** customer when when you have to unfortunately reject this claim because it's not covered or because whatever. Um the claim more often than not does exist. It's just that it is not covered because of the product is purchased. So he is very vulnerable and you know you want to be humane and you want to be empathetic in the way you deal with the given customer. So this is something that for example we will never automate um or at least as long as I run product in in this company. Um but there are other that we we can automate with different levels of of confidence. And so yeah we get all these different at every stage of the of the of the um of the claim we get uh all these inputs we treat them and then we have these outputs. Some of them we have to have a human in the loop to just approve them. Some of them just going to be approved automatically. This this obviously evolves with time because we get better and and the machine gets better and we just get this feedback loop from our current adjusters that review uh the recommendations like oh this is not what I thought it should be doing. So we just um we can reject the uh recommendations from from Watson and and they have to write down why so we can feed that back in the feedback loop. Um so yeah we instead of building like a tree in a way of we used to be talking about automation we're building we're building agents NBA machines that they actually advise us how to deal with different with different claims and some of them are fully automated end to end because all the decisions can **[00:40:00]** be automated. for example, um, stove tops when they're broken. That's something we cover. And they're really easy because you have the video, you have the picture, you can just see if it's a real picture, if it's a real video, because fraud, it's on on pictures on video. We we can talk about that. Um, then you see the brand, for example, from the picture. You just figure out how the amount it cost, when it was bought, and then you just and then you just pay it back. And then it's really easy and really fast to do. And sometimes that same type of claim uh sto a stove top then then the picture is not as good and or or you you see something weird with the pictures or something that we find very often is that the picture was taken before the policy was purchased. >> Yes. >> Which is not I mean it's just um trying to like trick the system a little bit but that's this is something that we have to be wor I mean worry about. >> So maybe it rejects it or there's something weird. So even for next the same type of claim um we just we can deal with it differently because we've built this this agent as opposed to building an automation tree. Um and yeah that's a neat example of and that's how we started >> moving from efficiency towards better decisions. >> That's a great example for listeners. The reason you don't want sometimes you do need a decision tree. want things to be discreet and very specific and very practical and that's okay. The thing is managing it with as complexity grows becomes even more complex and then the system becomes a monster. So there **[00:42:00]** are certain scenarios where this makes a lot of sense like Juan was explaining because like the whole flow you mentioned. Imagine how much time it would have taken everyone in that loop to first get the pictures, look at them, check some of them, go through a checklist, send it to someone else, they go through another checklist, then they go through claims, then they review. Now they have to see the brand. They go back to the pictures, check the brand, they have to go online, find how much that is where now all that can be even the research to find out prices and even like recent prices if you want to be I don't know you want to make sure or whatever and that there's no spike or I don't know you can do a lot of things now and you what you were saying is very real and very true that now you have if you have that feedback loop the thing can learn more so it's not locked If there's a mistake and you reject something, you can either just reject it and it's a soft rejection or you can say depending on how you set it up, you can say, "Oh, learn from this. This should not have been this way." And then next time it won't it won't make the same mistake. >> Yeah, that's exactly right. One of the things that and this is not very popular but um the way we've built the process now you could even it's not tree based it's task based in a way which is as I said it's completely changes the way you think around these processes but also >> we always design with a human loop approach because we think there there are **[00:44:00]** things that we don't know I mean there's things that we don't know we >> so there's always going to be something new and and we will probably need to to figure out uh and people we still are better for that. Um but we can still we we you can think about the process now as as Watson is a master of the process and then the human in the loop is just another tool of the process. Um and it's and if you think about it just or at least we like to think about it it's just uh it changes completely the way you think about the process itself. >> That makes sense. And if you think about it that way, the nice thing is eventually if you need to automate it, since you're already seeing that as part of the flow, then it could be automated for some steps and then for not. I really like that you're doing that where you're doing it for human moral principle reasons that some things you just don't want to automate because it might do it well, but it might not. And the feeling of rejection for someone with a claim is something that nobody wants to go through. So, it's better for someone to call you, explain the situation, or just handle it in a more human way. And and I think that's another thing like if you're passing some of your work to these agents, to AI, to software, to whatever you want to call it, then that opens up more time for your human staff members to actually interact with the customers and not be burnt out by all the back and forth of like just going and data capture and doing menial tasks. they can **[00:46:00]** actually be focusing on servicing the client. >> Yeah. Uh you just have the person where it adds more value and and if you you definitely are not adding as much but you have more value dealing with the person than right just as you said you just need to go online just look all the different pieces of information you watch a video you need to go and and look at different models and you just need to figure out the amount of price that I mean and the the demonization that you were going to give to the person. I mean the value ad really for us is or at least the way we think about it this is in difficult complex cases and the human factor and that's where we think uh our our people are just um adding more value to business. >> So let's look at it from the other side. What do your customers think? What do people think of working with a company that is more upfront with things that gives self-service that maybe uses AI to give them better service? Have you had any feedback from your customers saying like, "Oh, this is a better experience than before, worse." >> We we do we do try to um quantify everything and one of the one of one of the things that kind of told us that we were in the right in the right path was we started to see we we work with Trust Pilot a lot and for those that don't know Pilot, it's like a review um website. Uh you can you can do a lot of of neat things with the integrating through the website and and all the reviews are validated. So, you know, they're true. Um, **[00:48:00]** for us as insurance is a trust business, then we we thought you we better get as many reviews as we could. Um, we started thinking we started seeing people actually give give us good reviews with Leia. Leia was the that initial text agent that we deployed over WhatsApp. WhatsApp WhatsApp >> and we started seeing reviews like saying like, "Oh, thank you, Leia, because you you helped us very well. You told me perfectly what my coverage were." and and we started seeing more and more um those types of of reviews which was were surprising because people were thinking that it was a person that was out while was replying to them. So that was like oh this is something and then we started uh looking at the NPS uh from the conversation that where Leia was able to finish and where you needed to have a person >> and we saw 15 points top from Leia only and Leia plus person and obviously there's some some bias there because the Leia plus person type conversations tend to be more difficult topics or or >> rejections or things like that but we never thought we would have a larger NP PS and and that's I mean 15 points it's a lot. Uh I mean for those for those that know about NPS and yeah 15 points is a lot. So so we we those two those two things in in in conjunction we we saw it kind of tell told us oh we were on to something here and and we were very careful at the very beginning on which where to deploy and which things it could do and which things it couldn't. For example, we didn't deploy voice because we didn't think the technology was **[00:50:00]** there. So that's that's that was somewhere humans still had the advantage and we are deploying it now because the technologies were more advanced and and life kit life kit and >> 11 labs are amazing amazing companies amazing technologies um >> but also there there there were several things that we didn't want to deal with uh in 2023 when we started building this text agent we didn't want it to do anything with payments for example if someone had an issue with the payment it would go straight it would identify that and go straight to a person because I mean there there isn't anything more critical for a person that is money well family and stuff but at least dealing with an insurance company so that was something that we didn't want to deal at the beginning >> claims if there was a question about a claim it would also also be something that we would forward to humans and you have to be very I don't want to say delicate but you have to be very purpose purposeful on where you deploy the technology and and it has to serve of a business initiative not not just deploying technology for the sake of deploying technology which which is a risk that you can actually incur when with with AI because it does so many things so well that you maybe just deploy for the for the sake of deploying it and it's just not improving as much as it should and you you shouldn't be working on something you should be working on the more on the things that are going to impact your business the most and and I mean with this technology can do anything so yeah there that's that's another layer where the **[00:52:00]** human brain is still still at least for the moment does better than than the technology. >> That's that's a great point. You don't just want to put in new technology or AI or anything like that into a company for the sake of it. You want to like Juan said have a good valid business reason or or customer reason or whatever. But it has to be a real reason and a real outcome that you're looking for before you do it. But once you do it and if it's the right thing to do then efficiencies could definitely grow and then customer experience too and I I imagine in your industry it's very similar as well but I see in in particularly like healthcare I know you're not specifically in healthcare but you have this waiting room experience or even on the phone hold experience where I want to ask something or maybe I want to make an appointment and I'm sitting there 30 minutes on the phone until someone answers the phone. And so when technologies come in and allow me to either self-service book via an app or call a voice agent or use a chatbot like you were saying and help me just do the action that I wanted to do quickly instead of waiting in line and then also have the staff focus on the important things like actually talking to people about claims about other stuff instead of scheduling. that all that is really helpful in that space at least from my experience as a as a user as a customer for sure. One of the things that we realized when we started looking at at implementing that text text uh based agent is that 85 and we could and this is this **[00:54:00]** this was through at several months we could automate 85% of our customer inquiries over WhatsApp. >> Wow. >> But that's only and this is not because we tried to overreach. It's because most of the questions were again I'm getting the same type of questions. It's like oh is this covered in I have issues with that one. Do you do this? Do you do that? Uh how do I use the app? How do I purchase? Are you uh safe? Who are who do you have behind uh as a reinsurer? So what you could see is that there were no I mean we had done a decent job in terms of the actual product. Um, but there were questions about the coverages and and the very detailed, nuanced type of of uh of offering and you just go that and then you could you could have the people our people our customer service agents just deal with, oh, I couldn't pay this. Can you change my credit card? I had this issue with my bank and it just rejected the policy. Can you do can you help me? And all these different things that were really important and were really delicate for the customer. So you could have your people on that 15% of the of the actually inquiries and then 85% just because they were let's say more automatable inquiries you can just have the agent deal with them. >> Exactly. Yes. >> And it's exactly it's exactly in the process you described. It's just it's just on a different on through a different um channel in a way. >> Yes. Yes. I love it. I I wish more companies would would go into were just talking this morning with a with a well Mike he's **[00:56:00]** he's on the show a lot also. We do a lot of live streams together sometimes and we're talking with someone else about the the fact that AI space is moving so fast and if you are in the space you feel like oh I I'm falling behind like there's so many things happening and like have to go ahead but then you just step out of it a little bit and nobody knows anything like the rest of the world has no idea. There is yes you hear about Chad GPT and other things but I think there's still a lot of value to be gained for businesses and organizations to just start start playing around with it and seeing how it can actually come in and help your same employees be more efficient or do things even reduce human error and things like that. Like humans are amazing and our brains are the best thing that has ever existed. Still we cannot replicate it. So why are we using it for menial tasks, right? We want we want our brain to be used for more advanced tasks, more complex tasks, more human tasks, uh things that require empathy and things like that. And then the rest the machine can do, right? So at least that's my my point of view. >> Yeah, for sure. um we have a lifetime of context and and the the different LLMs even though the the context windows are larger you still have they're still limited >> so much that's an excellent point they're they're extremely limited compared to what we can do right so this is amazing what Alex I mean the the the whole reason we do this show is to make it a possibility for people to understand an industry **[00:58:00]** like your Swan where specifically an industry like insurance we are all outside we have no idea what's going on it's like just a black box and having someone like you come on I I really appreciate because it is a way for us to start seeing oh okay so there's people behind this and that's what I like to see there's always people behind any company even the companies you hate there's always people there so if you hate a company typically it's because an experience you had with a specific speific person, but not with the company itself. And there's so many moving parts. And I always like talking to people in different industry. And this one is one that can be delicate because people either love their insurance or hate their insurance or or they're there some just really don't care. They just buy like you were saying, they just buy the product because I'm used to I've done it for 60 years, right? But getting an insight, this has been really helpful, at least for me. I hope for the listeners as well, Juan, for you to explain really what goes on behind the scenes and how companies like yours are making the whole experience better and also I expect even more affordable for people. >> Yeah, it it is definitely more affordable. We are this one of those things that when we stated it most the people that listen to us are just surprised. We are we tend to be in on average uh 15 to 20% cheaper than the competition >> given a similar set of coverages but we are operating and this is this is from home owners only and because it's it's the way the one that we've done the longest and **[01:00:00]** I think the one that's more stable. >> We are running it at three times the average profit margin that competitors are running in Spain even with higher prices. Um and this is all this is all because we are doing all these different things and we are designing for this particular customer segment with in mind. Um and yeah and sometimes we just have to say no I mean we don't we don't say it but just sometimes we you just don't find in us the actual experience that you want and we it happens to us all the time that someone calls and and and asks can you can you do this quote for me? Um my address is blah blah blah and whatever XYZ and it's like no no we don't do that. um you have to do it through the on boarding process because it gives us I mean that's the way we set up our business and and yeah sometimes you just this concept of firing your customers and and yeah we we unfortunately have to do that because we are not for everyone and we are happy not to be I mean we're not happy to have to we would love to be able to serve everybody but we just don't because we do this explicit um um decisions we take this decisions on on how we run our business and our product and >> and yeah And that's that's what B enabled us to to serve a customer segment that is not profitable for almost every company and to run it >> three to four times more profitably than anybody else. So yeah, there's that. >> That's amazing. And I can actually attest to what you're saying because I'm reading through your Trust **[01:02:00]** Pilot and I laughed while you were saying it because one of the reviews I saw somebody was saying like they couldn't they left a good review regardless. They said they couldn't get coverage because of their occupation. something like maybe whatever. Um, but it said regardless the the question, it's in Spanish, but regardless, the questions have been perfectly covered by Leia, right? So, so there were they were happy with the service even though for their specific occupation or whatever, uh, it wouldn't have worked out from what you're saying, right? Makes sense. But but they still had a good experience with the product and they they even talked about the name of the agent which is >> there you are. >> That's amazing. Juan, thank you so much for joining us today. If people want to learn I know we have listeners in Spain which which I found very interesting last year. I had I had no idea until someone reached out and that was a lot of fun. So, if someone wants to learn more about Touyo and how it works and maybe where they can get some coverage and things like that, where can they find you? What's the best place to find content or or the about your company? >> Yeah, our website tuyo.com tu tuiio.com and our socials which are suo in Spanish. Um, so yeah, they can find us there or just uh look to online. There's there's a I think there's a coding agency in Mexico. So, we're not that. We're the insurance company in southern Spain or in southern Europe. Sorry. >> Excellent. Excellent. Well, thank you so much for coming on the show and we'll