Making Insurance Fair: How Tuio Puts Customers First With AI ft. Juan Garcia, Tuio

Insurance was built to profit off confusion, and Tuio is proving it doesn't have to be that way. Juan Garcia and his co-founders rebuilt insurance from the ground up as a fully digital, AI-native company in Spain, and they're running 3x the industry's average profit margins while charging customers less. The secret isn't just slapping AI onto old processes. It's rethinking every layer of the business, from data infrastructure to claims handling to marketing, so that AI actually compounds in value. If you want to understand what it really looks like when a company is built around AI instead of bolting it on, this is the episode.

Transcript

0:08

hello everyone and welcome to another episode of the ai rebels podcast as always i am your co host spencer and i'm your other co host jacob and we have juan garcia here on with us we are very excited to have juan on with us he's the co founder co ceo of Tuio which is spain's first digital native insurance company which that just that phrase i don't know if i expect it ever come out of my mouth digital native insurance company that's big so thanks for thanks for coming on juan alright thank you for having me we are excited to hear everything you're working on cause i think this could change the entire industry of insurance so let's let's just let's jump into it if you could describe the mission statement of twio everything that you're working on what what's the big idea behind behind this so basically we we um when you have to describe the mission and it's not for an insurer or something some someone from the industry it's just from from a customer for example it's just like we say just we are reengineering insurance to make it fair digital and with a with a price point that that we believe is that we supposed to be so yeah in there we don't talk about we don't talk about ai we just let's talk about the things that really matter to customers which is having a fair price something easy to use and something that's fair for them so so that's basically what we what we're doing i like that focus on the the end result and the impact rather than the process um so what what was kind of the the the breaking point um for you when did you decide that you're gonna rebuild insurance and inject it with ai so there's there's three of us confound in tuyo jose maria assist and i and and it's funny because we i i met jose maria from my previous job and and we kind of hit it on and we we started discussing things and and one of the things that we discussed at some point because i was working in in orange i wasn't an insurer i'm i'm not originally an insurer or anywhere near from from insurance so i was i was telecommunications and technology before and one of the things that that i that i did before i was running the new ventures team in in spain for orange which is a and it's an telco company here in in southern europe in and in particular in spain and and one of the insight that i got by building orange insurance is how broken the whole insurance industry is it's not just about the process just the customer facing processes it's just like everything within insurance seemed to be very broken like everything seemed to be more difficult than it should be from launching an adventure like like building building from from the pieces of paper that a customer gets to even just the onboarding process it all seems very difficult and very cumbersome in a way so just like there has to be something here that can make it can be better through technology right that was the the original insight that i have and jose maria was used to working b c g consulting for insurers so he kind of had the same experience and and assist i see he's he's the geek of the three of us he he basically deals with everything data and everything um technology so so he got to a similar insight but from the technology from the actual let's say the actual inside of the industry so yeah so from different experience but kind of similar backgrounds we just got uh that inside that this is broken there's i'm sure there is something that we can do to make it better we didn't know at the time what was it but we just we just figured that there there must be something that we can do to make this better for the for the customers and for the industry and that's how we got inside from to you and then we started looking into stuff and then you realize oh there's there's a two sided problem here in the industry um for customers for a customer segment that's with between twenty five and fifty five can take so the thing is from the customer side they just don't they just i mean we because i'm i'm one of those we couldn't find an experience that we kind of like or we could equate to i mean we all consume content from netflix we all purchase stuff through amazon we even buy groceries on that yeah right but in insurance you just just had to actually call a call center and and be waiting for a long yeah for a long while just get your your policy made or you just had to go to an actual insurance broker to his office and just and just give him your information and he would basically come back to you in two days and like oh this is this is your quote maybe you can you can do your policy through that so that's a very strange experience in this day and age right it's like a black box it is very strange so and you get that and and from people fifty five and above it's just it's just the way they're used to doing things about us we're kind of spoiled in a way we just we just used to all these online businesses that just are kind of real time they're kind of you don't have to talk to anybody if you don't want to i mean you can call them or you can write them but you don't have to right i mean they don't force you to go through a physical experience you can do it on your own through a self service on boarding right and we we we couldn't find anything that was that was working at least in southern spain there was there were and there were some insights that they companies were doing some cool stuff in the us or or the uk and and we kind of got inspiration from that and and on the other side if you look at the industry and you look at the different customer segments in terms of age it's and obviously age and age cohorts are just not as crystal clear as i'm gonna make i'm gonna make it sound but it's kind of like when you look from fifty five and above you see i've got that they don't use the the these products as much because i mean they they just they're just confident they just want to have it as a kind of like as a black solution when they do when they do have an issue and and they're just very willing to pay for this service and and they are just i mean for them for everybody but for them specifically price is a very kind of quality signal in their marketing but when you look at the fifty five and below you look and you look at their behavior it's quite different and we we all we all do we we go to price comparison websites we browse and research a lot online we compare different coverages we're pretty well informed on the financial literacy on our on our on our age cohort it's just way more so what you get is that you get a lot of a lot of pressure on prices because we are comparing different companies different coverages different yeah different products but then you also get a lot of pressure in in margins because as you get more financial literacy and you know what you're buying you tend to use it right because i mean yeah you know what's there and there are a lot of coverage is like manitas what we call in spain when you when you have a plumber send it to you in that you don't have to have a claim just have like a handyman sent to your home and that's for example that's something that the the older cohorts don't really use but then in the younger cohorts we see use a lot we use we see use it a lot oh interesting so you see yeah when when you look at the industry you see is that the actual older cohorts are subsidizing the younger cohorts and you don't see any oh wow or at least when then you realize oh well why would they be investing in the younger horse because that's not their desired customer the desired customer is the one that don't mind paying a lot and don't use it a lot right just great great great top line and great bottom line yeah honestly so so we saw this problem that oh there's there there there is something here because there is this customer segment that are that are grossly underserved but also there's a it's a segment that probably is not as competed because i mean there's no incentive for a traditional insurance to compete for that segment because it's just a bad customer from the point of view of yeah right that's from a traditional insurance company yeah right and that's how we found this this segment that we thought oh maybe we can deliver something here and and that's that's what we've been executing from since twenty twenty one i like that i i really like businesses that start from a problem rather than like a a stated product for say right like there's some people who start a a company and i'm not talking down on them you know there's somebody who gets an idea for a product and they launch a company around that product um and and to me the the ones that start with a problem and then build a product to suit that problem are so much more compelling um and yeah and more interesting in a lot of ways right and it took and it took some time we didn't get we didn't quite get there at the very beginning because you have to look at the problem in different ways but one of the things that also gave us some insight is that in syria they just have this one uh go to market and one product and and they they try to make every customer just go through the same process and and one of the things we realized like oh maybe maybe there's something that we can do in in different segments and that's how we yeah we started doing research and then we just found and now it seems very obvious but i mean in twenty twenty one we it wasn't real yeah yeah that's what i was gonna ask about cause twenty twenty one is i mean that's like the that's right when the ai wave was like just starting to rise up out of the ocean i feel like twenty twenty one with lms and everything so you made this decision before it was really obvious that maybe you could automate relatively easily a lot of these things what was your plan initially for the automation side of things sure the vision has changed with the technology right because and and now it's even changing now every six months but that point yeah um and it was funny because there's this saying that as this is my my technical co founder he always says that well he used to say that point is that is that machine learning and this is twenty twenty one right machine learning is run on python and ai is run on powerpoint and that was that was a joke on twenty twenty one and and and that kind of painted the the things that we were doing at that point we weren't thinking about the chat gpt of the world we thought there was something that we could do on on data so executing on data at that point was something that we had out of clear we were pretty clear from the very beginning and having as much data as we could and having non traditional data to do all sorts of stuff um not not only on pricing but also throughout all the customer life cycle yeah that was one of the first things that that we decided but then of course then two years later we had to actually completely re redo our our vision because now you can do crazy things that you you couldn't even think about in two years ago right so so we we started crazy on one path we thought digital and machine machine learning and data basically it doesn't matter machine learning or any other any other way of doing it exploring data was gonna be what what would be a difference in the market but then just the genia explosion came in it's like oh this is i mean this is great for us we had to do a lot of work also because we started executing on on gna in twenty twenty three we're pretty early not as early as some other companies probably in the us but yeah we're pretty early here in spain but by the mid twenty twenty three and and we can't we can't forget now because all the models work perfectly but at twenty twenty three it was so dbt three point five and it was still hallucinating pretty badly and we had to do a lot of architecture around that because obviously that was a huge problem you didn't want i mean if you were you were gonna have like a chat bar or something customer facing you didn't want it hallucinating coverage that you don't have right no we all remember this this tweet that this um this chatbot from an airline i think it was that they they said that oh you all of you had three tickets and they had to actually go and give it to them because because i mean it is part of the company right so so you had to work around that and and now it's way easier and and we've seen some of the stuff that we've done we've now now we're about to release the the third version of of our original chatbot which was okay our first project and what we what it took us i think three months and a half to build for the first time obviously there's some some i mean you need to learn stuff like that and the second version took us i think it was a month and a half it's already kind of half of the of the initial version and now i think we're rebuilding it in like a month and it and it's way more powerful so it's just i mean technology just has the advance of technology just has these these i mean it propels you right it's just yeah have like having tailwinds is just for you and the thing that's amazing about the the models now um is how versatile and powerful they are that's obvious to say but i'm using that to point to something more specific which is that back in twenty twenty three twenty twenty two when they first released um when the apis first released uh they didn't have function calling or anything like that so if you wanted to say you know you wanted the the the el llm to be able to receive a command from a user and then do something you had to set up a whole infrastructure yourself to to yeah make that work um i set up a little chat app for myself that that sends uh ui components from the back end to the front end um and i had to rig up this whole redis stream where i'm you know i'm passing the data into an intermediary data source and then i'm reading that from the front end and then i'm passing another message from the sir anyways it's it's to your point one like it's crazy how much more powerful these models get every single month and and and that that power compounds in in in in the product yeah it's not really cool you didn't have blockchain or the whole ecosystem so yeah exactly so doing workflows was i mean it was kind of like trying to walk through this this desert because you didn't i mean you didn't have anything and context windows were were very small you didn't have any tooling there was no memory it was just like crazy i mean it's just like you had to throw in a bunch of retries because sometimes the streamer would just quit yeah and it's so funny talking about it cause it feels like it was so long ago and like so much has happened since then and it's like two years it's like a little over two years have passed yeah it's just crazy to me this this whole ai shift it's um it's been a wild ride so i guess one before we keep going i i meant to ask at the beginning and i'm curious for an end user like someone using twilio what does it look like what's the what's the back and forth how does what's the flow for a customer of twilio right we we we had this one thing and this inside that if you wanted to make this customer profitable then we couldn't let's say we had to put the opus on himself to just do everything on his own i mean we we we think about how in a product in a digital product way of thinking we how how to make it everything very comprehensive very easy to do um very clear in a way it's very to make it that actually you make insurance very self service that's our angle when when we work a digital product so basically when when you stumble on on one of our ads and get into our website you just have you just do i wanna purchase this and then you go through an online onboarding and where everything is it's kind of clear we have reduced quest questions at a minimum just the the real things that we need and everything else we plug from different external databases and and we obviously we we plug on more data that traditional insurers don't have on the way your behaviors on the onboarding the things that you read the things that you click the things that you look at the things that you don't look at which are as important as the things that you look at and now through events and then with all that information we actually apply into our our pricing engine and almost a sentiment analysis as they're going through that initial process i wouldn't say i wouldn't say it's sentiment analysis but i would say we we do some cool things for example that that we can we can share and it's just a a way of improving decision making throughout the whole life cycle okay so for example if a customer when he gets to his quote he's reading he hasn't read anything but he's reading one particular coverage very carefully more often than not we there is something wrong with the customer we tend to plug it we don't do anything on pricing right we don't do it because i mean he hasn't done anything that merits different pricing than someone that hasn't that hasn't that hasn't shown that behavior but we we do flag it because uh we found that more often than not there's something wrong with the customer we those customers on average tend to have earlier claims during the uh the life of the policy so for example interesting customer would exhibit this behavior and then suddenly on his second week of of this policy i mean and second normally is just on the very first week or even very first day of their policy then they have a claim and one of the things that we're flagging is like oh this customer exhibited this behavior so you probably need very carefully into all the different different data points that exhibits that he's given us on our online on our online claim claims process right because for to actually file a claim we ask we ask a the customer to file to file a video and some pictures and all these different data points that we use then to evaluate the the claim and and when he exhibits this worries and behavior what we do is we actually look at those more more carefully and more often than not you find that either in the metadata or on the naming of the picture or the video you see they were actually taped or taken before the the policy was even purchased oh wow or sneaky or you see or you see all the or you see all the different behaviors and it's not very not very sophisticated it's one of those things that you actually laugh a little bit when you when you find it um you see that those pictures are from google images and not even on the very first on the like the tenth page of google images it's on the very first page very first result of google images and oh gosh so yeah those are those are those things that that we look for and and that behavior it's not sentimental analysis it's just what we call the customer dna we get all this information and and we see patterns on there that we actually flag so the adjusters in this case they have this information so so they knew they know what to look for in different cases and that's fascinating a week ago we got finally we got our first image just it was built by gen ai and it was one of these and the funny thing is that um probably uh well i would say probably it's not the first one because the really good ones you don't see i mean there's there's there's nothing about fraud that you don't you the real good one you the real sophisticated fraud you just don't really detect it um so probably all those using a nano banana pro they just they just kind of fool us but this one wasn't even using nano banana pro it was probably using just some cheap free uh online image generator and it was you could definitely see that it was ai because the the it was a window like a broken window the window was on top of a table and it was float this floating effect but it's not really like it doesn't look like it's set on the table it's like kind of like floating stuff and and the the the pieces of glass were from different sizes and she's like yeah this is too perfect i mean and and the broken glass was it was too perfect it's like one of those things that yeah this is definitely not it's just it doesn't work so it's so yeah we can have like yeah oh that's interesting our first our first dna i picture um on fraud yeah see and this is a perfect it's a downstream consequence of ai that i don't think anyone would have thought about like the impact on insurance and these hyper realistic images i yeah that's kind of terrifying honestly for insurance companies i'm sure i wonder if it'll become common to uh for for insurance insurance companies to hire you know more in person claims adjusters so if there's anything suspicious they're like you know what we're sending a guy out uh right and then and then the person gets back oh no no no no i'm fine and then you're like oh ok ha ha ha that's that's you're lying ha ha ha that's what they're doing anyway since they don't have these these digital process that we have they they always need to support more often than not and depends on the company and they they send these these people just just review the the claim because they just what they only have this call from a person that says i have this claim and i need to and i need to file it whereas we yeah and that's a big different for us we have the the video the pictures yeah right we may we may even have if it's a some something to do with the stolen goods we have the police report which is which is so we just have so much information at every step of the process that it's just like it's an unfair advantage in a if i've ever seen one yeah yeah yeah yeah interesting okay so they so the customer submits the claim images video they or i guess they they sign up they answer this entire questionnaire with information and then do they just have an online account and yeah yeah everything is like everything is just like like the standard online business that we've all you got used to this point you can get into your into your private space and you have all your policies you can browse your contracts we you have it all signed over there so basically it's online everything is self service you can call us or write to us and we have this way of thinking around it's like oh if someone has has had to call us it's because we haven't we have done something bad in product so we try to learn about all that and we try to plug all that into the like a like a closed loop so we can improve the product and stuff mm hmm so yeah basically you treat insurance as you would treat your account on amazon or netflix or anybody anything else interesting okay so maybe a little more technical i know it's from what i've seen it's kind of an ecosystem of agents with with trio can you walk us through what that looks like i think one of the biggest insights that we got is probably taking a bit of a step back but it kind of preaches what i what what i explained in terms of technically and what we think we still think we'll still see a lot of companies do is everybody is chasing pure efficiency and in insurance efficiency is cost to serve and cost to serve is about ten percent of your cost base only it's like the the largest chunk of your cost base is marketing and claims right so if you even if you try to reduce and if you managed to reduce by fifty percent your cost to serve what you're gonna you're gonna earn what five percentage percentage points on your profit margins that's not a lot and even yeah that's considering you you just got a fifty percent reduction which yeah more often than not you're not gonna get but if you look at the eighty you look at marketing and claims that's eighty five percent of your cost base so if you if you only take one like ten percent you're you're having eight eight percentage points added to your profit margin and let me tell you there's a lot of room for improvement in in marketing and claims in in insurance industry and just to close the loop when we did this switch we are able to run our oldest product just homeowners insurance at fifteen percent profit margin which is three times the average in spain the average is just five percent and um and we're looking at with some some stuff that we're introducing now we're looking at probably running at eighteen percent by the end of year so you're looking at four times the average profit margin that we're running and just and we're not refusing more than the average we're not doing we're not doing some shady things we're just doing things better and that's because we had this this uh insight and around half around mid twenty twenty four that this is not to just reduce the size of your call center and and and do and making sure that you're able to scale efficiently is one part of it but it's a small part so the the the thing is that how we can do things better and how we can be more granular in our decisions how we yeah how how we can think around insurance better and then for us that was we organize around three levers and that's how we've been deploying ai since we we we organize around growing more efficiently which is obviously marketing um and the writing smarter which is not there's ai but it's more on the machine learning side it's just not as much gen ai as as as you would expect and then managing claims more effectively and and what you find is that um when you use for both marketing and ai and for making decisions it's not about like the traditional automation process like you first have a then you have b and then you have c because the thing with marketing and claims is that's not it's not a very linear process either of them if you look at claims for example it doesn't follow the i mean claims don't follow a straight line they they branch out depending on the coverage that there are the coverage that actually is the severity if there's a a a fraud suspicion if you need to schedule a repair if there's a third party involved so so there's it's a very non linear process so there's a lot of divergence in in among claims so yeah what we what we start when we started thinking about what what can we do to automate this is that that we figured out that you cannot automate it as with the traditional technologies so what we started doing is we use g a i to build a a next best action machine so at every stage of a given claim you get all these different data points all the different information you plug from different places and you run a set of analysis and then you get a set of recommendations and a confidence interval on on every recommendation and some of them you automate and some of them you have a person review them right so so that's and and we did that both for claims and for marketing and we just got this this amazing results just by in claims you're able to do more and do better and marketing you're just you're just you're just able to be very granular on your campaigns so before and and you guys i'm sure you know you couldn't branch out a lot on your campaigns and and you use broad matching because if not it's completely unmanageable if you have this right crazy amount of campaigns and all and all exact matching it's just like it's just not manageable for a person but now you have a machine that can go through all your campaigns through all through all your keywords with exact matching and just give you recommendation and a and a confidence level yeah so so now what used to have what used to take a couple of people and a couple of days or or a week then we can just do in twenty minutes wow it's it's yeah it's really a new world um and i'm curious why why don't other insurance companies in spain do this what what's holding them back are they are they just trapped in the past um um no actually actually this this is a great question and and one of the things that we just realized that when we had this switch on on our mindset and probably explains why not only the insurance in spain but you also see i'm sure you're aware of this this um papers that say like oh ninety five percent of pilots just don't show any any improvement for companies and the problem is that and this is something that we found ourselves and and and now what it seemed very easy and very again very matter of fact and very logical it took us quite a while to just figure out how how to build it and and what what made sense to build is that you really need to rethink your company from the ground up right so we what we did is we we didn't but i mean and we were pretty lucky because we only started in twenty twenty one so we only had like two years of legacy at this point but one of the well three years of legacy at this point but one of the things that we just realized that if you bolt ai into an existing consumer into an existing insurer it just won't work or or an existing company yeah from from that fact so so and and this jose maria is able to show it very well in one slide so we started thinking about our our company as a substrate of data at the very bottom then all your your business processes on top of that and then all your interfaces and interfaces it used to be uh screen interfaces your app for your customers your your your app for your adjusters where where they do all their things and but now really it's it is a subset of screens or or human ai or ui but it's also apis and it's also mcps and now on top of that on top of that on top of that let's say organization you run your human agents and your and your your ai agents and and they do and they do work for you so that's that's the way we started thinking about it and when you think when you think around around it in that way then then you just realize your processes are not geared towards humans only and are not geared towards screens but also from apis and mcps and ai agents right so so when you think around that way then you just realize you have to rebuild and rethink your company it's fascinating yeah it makes it makes sense i think a lot of companies just everyone got excited about ai and wanted to use it and saw all these shiny new toys that had the potential to be cool had the potential to have an impact in the company but no they just tried to buy it's like they bought a set of legos without the manual and try to just like assemble it like it has the potential to turn into something really great but you're not gonna get there unless you actually think through okay what do we how where do we start how do we integrate this into our company which we were just talking to another guest actually the new episode that we just launched today with jake ambrose i was gonna bring this up as well yeah we were just talking about how like the best practices have not changed whatsoever i think i maybe i'm just beating this dead horse at this point but like the the best practices are still the same with and without ai exactly what you're saying like you have i just love having seen this art for twio where you identified the problem i i i feel like that's marketing that's my marketing one o one class from college is ok you wanna start a company what's the problem what are you solving what's your ideal customer and you literally i think this could be a case study for a marketing class like okay here's the problem now here's the here's the market that you're going for and then you build around that whereas these existing companies aren't either they're not willing to make that investment or some are like this guest we had on with this company called brex um they have made that investment and they've made many pivots and kind of rebuilt processes to integrate ai and it's made a huge impact but it is a large investment like it's not it's it's almost harder for an existing company to rebuild than for a new company i agree with that and i think a lot of these these big larger companies one of the things that um so ai to me there's there's two things about ai which is gonna make it very difficult for existing companies to integrate one of them is is it fundamentally change changes your business your business processes i should and your and your technology stack and the way you organize your company so basically that that means if you actually make a bold investment into ai you need to change your organization you need to change your processes you need to change your technology stack so there's gonna be a lot of people that's gonna be upset because either their job is changing or they have to they have to do something else or you are imposing new stuff onto them so to begin with it needs a lot of courage from from directors and and and leaders because i mean you're you're getting you're putting yourself out there just by saying if you do the real investment you're you're definitely putting yourself out there because there's a lot of risk a personal risk involved the there's a lot of and there's this this argument that if you don't do it then there's company wide risk because it's gonna fail but we still have to see that and and we will have more call out or or or um more kodak or more blackberry moments or nokia was the one that i was looking for yeah nokia you're definitely you're definitely gonna have it but it's on the it's on the more longer term whereas if you do this bold investment and you do this a whole organizational transformation and you and you are the the face of it and the the face pushing for it and it fails it's gonna fail in one to two years and then you're gonna have you're gonna lose your job right so there's a lot of personal risk into these transformations which which i think to ai by its nature require and and we are the perfect example of that we started doing one thing and then we had to completely reengineer the way we run claims with with the this next generation machine in the center we needed to we needed to completely reengineer marketing with this this different agent in the center and and so basically but we are three co founders and it's a small company still so we there's a lot of our imprint into it so that's what we need to do and still but if you're looking at a big public company there's a lot of personal risk involved a lot of agency cost on that one and that's one thing and the other thing is that the usual strategy just to outsource it to the ibm's or the accenture of the world if you do the traditional projects where technology and the the outsourcer i mean the the consultant company it's pretty removed from your business that won't work yeah i mean there to me and and this is something that i tend to believe that ai implementation will only work if you are done if you're if they're done within the company with within company talent i can see it's somehow working if you have your consulting company very close to your to you and your processes and and you give them you have them onboarded as as someone from your company yeah but the traditional engagement where they are really far from you and they just they're sitting on their on their offices and they kinda know your business but not really that definitely won't work right because it's it's so to do it well i believe it's so core to your company that that that a lot of companies if they think they can engage the traditional consulting firms they're gonna struggle that's that's my prediction and maybe i'm wrong who knows but i would agree that's that's what i believe it aligns really well with with what we've been hearing from from a lot of our guests and and we have guests from everyone from you know big large companies to startups like yourself and yeah like jake was talking about our last guest ironically also named jacob um uh pointed out a lot of the similar uh a lot of very similar things that that you're pointing out that if you are wanting to lead out on like an ai initiative at at your job or whatever like you gotta have courage to look dumb you gotta be willing to acknowledge that you don't know everything yet um especially in a field that's changing as quickly as ai is and if you and if you and if you keep yourself into in like the safer projects just just having chat gpt or copy of given to all your employees that's not gonna have a lot of impact really so right exactly that's where you get into that ninety five percent that don't see a lot of value yeah in ai yeah totally and i i've seen that so much and i the company i currently work for is called nasuni and they it feels like they have done a good job where they did they rolled out chat gpt enterprise everybody has access to chat gpt and the same day actually even before everyone got access to their license they rolled out an education plan and said okay everyone you are required each quarter to go through these classes which will teach you for your function how to use chat gpt enterprise like use cases and how to integrate it and i really admire the way they did that because i think that step's often forgotten they companies want to say that they have achieved their ai initiative for the quarter or the year and that goal was to have every single employee have access to chat gpt yeah like if i were to say check if i was working as an hourly customer service agent at one of these companies they're like you gotta use chat gpt i'd be like alright sweet i'm gonna use it to generate some pictures while i'm at work and i'm gonna mark off on my sheet that i used it and we're good hahaha yeah if you're not gonna tell me how to use it i'm not gonna i'm not being paid to figure out how to use it unless you tell me to right right right um we're pretty clear in the way and and and yeah if you see our use cases they are they're a bit different than that and one of the things that that we we found that even even in a company like ours is this cultural resistance to change and yeah and yeah and the way i am this is one of the one of the things that we're probably seeing in society as general just not in a in a companies like absolutely yeah the the kinda like resistants to now you're looking at now how people are resisting more and more of ai and and how contrarian views on on the success of we are more and more successful and one of the things that we do is that um we will give you access to any within reason obviously to any tool that you want even if it's for your uh side side hustles or side side jobs or side projects or whatever we will just give you access because we want you to be to have this ai literacy and then oh i mean we don't know we don't we don't know where the next great idea on how to use it within tuya is gonna come from maybe it's from an adjuster maybe it's from from a customer support agent you just we just don't know so that's why we did that and we've seen some great results one of the first things that we got was um so we have pet health insurance and we have the guy running that it's a it's an actual veterinarian he's um oh interesting he used to be a veterinarian he's by trade he's a veterinarian this is his first office job it just happened to ok to know him and he wanted to do something differently and he he came with us and a lot of people and yeah so it's so it's first first office job and he he's running well with with our help obviously the bed health peace and then one of the first things that we we we started looking at chat gpt and then he built a custom gpt to just help him uh do the claims because claims pet health claims are just kind of easier because they just give you a bill from the veterinarian and you just see if it's uh just i'm just simplifying but you just see if it's covered you just see if if the uh if the health history matches and that's it you just have you just look you just have all the information that you need so you build this uh custom gpt this is not even very very sophisticated it's just a custom gpt from chat gpt um he just he will feed in all that information and the custom gpt has that policy he will just look through it and and we look at the different coverage and we look at the pictures and and the history and the and the bill we looked at everything matches and then it's like yeah yeah it should be covered so so he just got himself a nice little automation that he he uses on his own on his own on his own job just to do it better or just have more free time or whatever and we just got a nice cool idea that we can implement on on our systems and it's implemented now on the on the on the process for the justice for for uh for pet health and that's how it came to be because because we open up all these crazy tools and we will pay for the from the company so they can just play with them oh see that is i love that cause that just shows trio's willingness one to invest in their employee base but two the recognition that like we're not the end all like our what we our processes are not final and i think a lot of companies especially mature large companies have reached that point that our our our processes are tested we have so much data behind our processes and they are set and they are the gold standard and i think a lot of companies need to realize that that's been obliterated like you gotta be everything lean and mean but this i just love the idea of the employee benefit of that would attract me to a company like hey we'll give you access to ai tools i'd be like that's huge i would love that we don't even not only we don't think our processes are final we don't even know what's the next big idea is gonna be we don't even know what we're building in a month in a year from now because we just we just don't know what the capability will be and and what we'll be able to do and that's that puts pressure on your product organization because we with ai we we cannot work the same way we work with traditional tech we have the product organization that figures out that figures out what we want to build and and why because of we have this customer need or we have the business need or whatever and then and then we build it but there's always yeah a customer need or a business need behind everything that we build we just because if not it's very easy to just fall into this building for the sake of building because it's fun because it is fun of course that's why it's fun that's why we got this this side project kind of idea just to make people just more aware of what what can be done and and and being able i mean just get more ideas because if you're aware of what can be done then you just sometimes some people get ideas on how can their process their their process that they're working with can be improved and then it feeds back to us and it feeds back to improve improve to you so for us is definitely a win yeah yeah yeah i love that i'm curious with these large insurance providers now i mean you've in the words of miley cyrus you've kind of come in like a wrecking ball to this this insurance industry where things are very traditional it's in the past it's kind of been like an old boys club like the men and it's just such a old industry and a lot of the players are old have you gotten much push back from the traditional insurance providers yeah definitely another way of and and we used to do describe to you as as this way not not so long ago so a way of describing to you and what we do is that it's been a a contrarian story so basically we've been told at every stage of the company that what we wanted to do couldn't possibly be done so when we started we just like oh we want to build this online self service digital based on boarding process like no insurance it will never it will never be self service it's always sold it's never bought that's something that we used to be and it doesn't make probably doesn't make as much sense the the idiom in english as it is in spanish but we we got at every stage we got told that like insurance is not it's not bought it's sold and and we were betting on no actually there's people that want to buy it on their own if you if you put it on if you build if you build a good enough process and a good enough product and a good enough price and that was the first the first thing that we actually proved that was against a little bit against the grain that's what we used to say and then the next one was well you won't be able to have people file their own claims on online and then again we build it we we had to iterate on it and then we had this now it's eighty five percent of the people file the claims online through our online process obviously wow there's fifteen percent that's our urgencies and they just i mean your natural reaction is calling and that's okay that's that's part of the process so eighty five percent that's a that's really healthy that's incredible yeah that's basically everyone file their claim online and and we have to at the very beginning we have to push a little bit and they would call us just file your claims like no there's this process and you can do this and you can do that and this is the way it's gonna work and it's gonna be better for you and it's better for the company and um but yeah now we're eighty five percent the the self service online purchasing is at ninety eight percent so every basically everybody we we don't we don't we don't sell policies just dispatch them in a way and that's amazing and at every step of the company it's a story of of uh of going against common sense and and the industry common knowledge and and yeah proving them wrong so so that's that's yeah that's we have definitely had experience a lot of pushback and now now people are just mesmerized by us not really but a little bit hey we are just a little bit i'm mesmerized no i'm i'm a big fan if i if i lived in spain i'd be i'd be logging on right now to buy an insurance policy i mean is there any plans to expand yeah we definitely see a gap in southern europe we that's that's the our claim to fame with with fans and and potential investors that we want to build the southern europe champion in insurance because we what we see is like more advanced well more penetrated come uh countries where insurance has more penetration per per gdp kinda like the uk the nordics germany switzerland they do have companies similar like ours there's there's space for everybody but we we see way less competition in the south because insurance yeah probably less attractive uh in in terms of pricing and margins probably yeah so we see we see a space in portugal we see space in france italy greece awesome probably some others more countries that will that they will go after so so yeah probably in twenty twenty six more h two twenty twenty six we'll be looking at expanding internationally yeah okay awesome that's big it's a big project it is indeed yeah wow so i'm just amazed at the you've given several examples one building the i think you you said it was like the chatbot from first iteration three and a half months then was it two month or a month and a half and then it a month with the third iteration yeah and and at every iteration way more powerful with way more functionality right so at the very beginning yeah it's had this little chatbot that only replied to questions about coverages it couldn't i mean and then we evolved that and then it became a bit more things and now we're building a tablet that basically is a text based interface where you can do everything that you can do on our app but through text but also it's has complete observability and evaluation so every conversation we have is its own evaluation and we've built this closed loop that it will it will see what go went well or what went wrong and it will we will have a more llms or more agents sorry running on top of that and they will it will advise on improve improvement on the on the actual agent but it will also advise on improvement on the product like the insurance product yeah and we will also advise us on improvements on the digital product so so we so through so cool every customer conversation you will get advice on how to improve their business which is which is basically what you would have always wanted to do but because we didn't have this technology you just need to reduce yourself to sampling sizes and and customer and like non tradition like traditional customer research but then now we can actually do it at scale yeah it seems to me that toyo has uh that you guys have kind of optimized for an extensible platform more or less as a company um and i really i find that really interesting and really admirable in a lot of ways i'm fond of saying that like that that the correct amount of friction in a product is not zero um and so i love i love hearing stories like to yo that you know i then get pat myself on the back and be like oh yeah i'm totally right ha ha ha um and and um it's just it's really interesting to hear stories of of you know how fast you guys can iterate now um both because it it's pointing to something new which is this this speed of iteration but going back to what we were talking about earlier like it's really it's just the same core business processes that have existed for a long time hyper optimized right like yeah um you know iterative development has been a thing for software for forever right and now it's just it's you can you have a fast forward button you can press and it's incredible scale yeah yeah so i love that the kind of the organizing philosophy of of of taking full advantage of that yeah i like that too i'm curious um wanna if we get mixed replies to this question and i'm very curious what your answer would be because you have almost achieved human a lack of a human in the loop on a lot of this insurance these flows that you've built do you think there will be a time where there is no human in the loop for your processes or do you kind of always see that a need for human oversight well just two things we we don't have and maybe i explained it quite wrong we don't have we all well we always we have still a lot of human in the loop but that's by design so the confidence levels and the outcomes and the suggestions that we have in our bots on our on our nba machines some of those for example if you look at claims which is i mean it's it's always better to just figure it out with an example like for example a refusal a claim refusal we will never out fully automate that because we think that's a moment when a when a customer is which is a paying customer it's pretty vulnerable because i mean imagine if you're outside of your house yeah um and you just lost your keys and you just need to do call us or or text us or or just get in contact with us and um chat gpt says no exactly that makes i mean it's just like having this empathy i think it's part of insurance and totally that's one that's something we designed that there's several there's several actions that we will never fully automate and only and there's always a confidence level where you automate things even if it's a a decision that we we can we can see ourselves automating so i think it depends on the product and the customer for example there are some these these telemetric insure the telemetric insurance pieces that i i fully that we don't do obviously it's more b to b but they they can be fully automated because i mean if if whatever if you have these winds on top of whatever knots then maybe you just have a payout and that's something that's easily automated because you have the the data source you can have the machine looking at it and then you just you just pay the customer that's probably b to b b to c i definitely think that there will always be a human rule and because there are decisions that are so critical for your customer relationship that they cannot be otherwise uh or they shouldn't be in our point of view they shouldn't be automated even if you can i like that yeah and then when to say no is important it's not only on claims as well for example some of the the different topics that our chatbot can can deal with for example problems with payments payments are so delicate and even if it's a small payment it's just the customer's money right so that's something that we don't have to worry about we just have we just we just go on the conversation and we just switch it and whenever we just realize that it's about that we have we pull the information from from our systems we have this this copilot uh copilot screen that that uh the customer service agent can look at but he will always be a person that will answer your doubts on payments because it's very it's very critical for customer even it's ten euros a month it doesn't matter right it's just his money and and it's not up to us to decide what it's worth for him or for us so right for example payments in in in the traditional cost to serve ai agents we we don't know we don't automate either yeah i love that and it i think the key is what you said at the very beginning is that you have a human in the loop but it's by design it's very purposeful it's at the key you know crux of each kind of decision conversation which i i admire i think that's where most if not all companies need to get to where ai that what ai is great at is all everything in between it's great at the in between stuff it's great at automating filing a claim checking if a claim is valid like you mentioned with pet insurance like that is it's bread and butter that's that's what it's great at but at least currently and frankly it's because of human bias maybe it could be better if we were more open to it but currently ai whether it's the ai functionality and capability or if it's just human trust is not great at conversation at having those difficult conversations at reasoning through an angry with an angry customer it's not great i mean and you can do it and we just feel it's not great it's not good for our business right so so different businesses different founders different directors are probably have different i'm gonna probably have different views on this but our view is that there are several things that we don't think we should automate even if we can and maybe it gets to a point that our view changes but i think this is probably a matter of principle as opposed to a technology capability so that i mean definitely our future is something that i don't see ever being automated yeah um yeah there's some other things that maybe we we now don't feel comfortable with but maybe maybe in a couple of years i mean at the at the pace that this this accelerates yeah maybe there there i mean maybe there's even agents buying uh insurance for our for our customers at some point just you just plug into chat yeah he's like i want home home owners insurance in spain and they just they don't even buy themselves right so yeah we don't know where we don't know where the world is gonna go in a couple years but for sure there's several there's several decisions and several actions that if there is a person on the other side we will probably want to do it ever yep that makes sense that makes total sense as we're uh wrapping up here one i'm curious we always ask this at the end if you were to give advice to somebody who is um skeptical about ai but interested in learning more what advice would you give them for becoming more comfortable with the technology at least having a greater level of knowledge um even if they don't necessarily like it there's a i mean there's and a i don't know a thousand way of answering this question probably just because of what we're living right now is i would hand them a cursor um a cursor uh monthly fee and just let them build something and then just yeah i mean once you've seen it you just cannot see unsee no it's true yeah or or or whatever or whatever coding agent you want you want to have it's just like you just you just i mean just build something and whenever you just build it then you just i mean you cannot even see it yeah it's addictive it's life changing it really is i love that um one if people want to follow you follow Tuio what are the best ways for for listeners to do that sure it's just tokyo dot com we were able we were lucky to to buy a four letter dot com url that's already taken but this one this one was okay wasn't too bad um so that's probably the best one and then linkedin is j u a n g a two that's my linkedin and i just just look to your on your socials and we'll be there perfect awesome yeah we'll definitely we'll drop links and and everything it's been a pleasure thank you thank you for coming on and we'll definitely stay in contact you're part of the ai rebels family now so we'll stay in contact it's been a pleasure as well it was it was fun thank you sure we'll talk soon thanks fone bye