EP 161 – Christina Cai – co-Founder and COO at LydiaAI – How Do We Link Health and Wealth?

Listen

Michael Waitze worked in Global Finance for more than 20 years, employed by firms like Citigroup, Morgan Stanley and Goldman Sachs, primarily in Tokyo.  Michael lived and worked in Tokyo from February 1990 until December 2011.  Michael always maintained a particular focus on how technology could be used to make businesses more efficient and to drive P/L growth. Michael is a leader in the digital media space, building one of the biggest and fastest-growing podcast listener bases in the region.  His AsiaTechPodcast.com show has listeners in more than 170 countries and his company, Michael Waitze Media produces some of Asia’s most popular podcasts.

Christina Cai is the co-founder and COO of Knowtions Research, an applied AI company on a mission to insure the next billion people. Having grown the company from the two table desk at the innovation centre at the University of Toronto after graduating, she believes the key to scaling is cultivating teams with strong roots in our mission and humble origin. Christina leads with empathy and has evolved to guide the team through different stages of growth as the company pushes towards using technology to advance insurability in every reach of the world. Insurance companies tap into the company’s Lydia AI risk prediction engine to make accurate dynamic health risk predictions based on alternative data. These actuarially validated health scores are used to make personalized product recommendations, accelerate underwriting and create new insurance products. Established in 2015, Knowtions Research is backed by Alibaba Entrepreneurs Fund, Information Venture Partners, 500 Startups and the Canadian Government.

This episode is brought to you by:

The Asia InsurTech Podcast spoke with Christina Cai, a co-founder and the COO of Lydia AI, about applying AI to healthcare data and the challenges of founding a startup.

Find the transcript of our conversation here:

Michael Waitze 0:00
Okay, the recorders on. Hi, this is Michael Waitze. And welcome back to the Asia InsurTech Podcast. This is the only podcast in Asia focused on insurance that gives entrepreneurs, we’ve got a great one today, thought leaders and investors a platform to discuss how technology is reshaping the insurance industry, frankly, globally. We are super happy to have Christina Cai, the co-founder and the COO of Lydia AI on the show today. Christina, thank you so much for doing this. I trust you are well.

Christina Cai 0:35
I am well, thank you so much for the opportunity to share.

Michael Waitze 0:38
It is my pleasure. And where are you based? Just for my edification.

Christina Cai 0:43
I’m actually based in Toronto with half our team based in Toronto and half our team based in Asia. Taiwan, actually.

Michael Waitze 0:50
What a second, if it’s Toronto? See, for me, it’s so hard to tell what month it is because I live in Bangkok. So every day is 31 or 32 degrees. Never changes. What is it like in Toronto?

Christina Cai 1:00
There is snow outside, there’s snow outside as of yesterday, I should say. And today we had like, double digit Celsius weather which is really nice. So we had like 10 degrees, 15 degrees celesius. And for American friends, that literally means that I did not have a very heavy down puffer jacket on, right. We have snow boots, and I could actually wear cute little ankle boots and not get snow stuck into my boots. That’s that’s what the double digit celsius means right? I think it was 15 degrees or something today. So it was nicer today. But in March, we’re seeing like spring slowly start to actually tease us a little bit after the cold snow. But even as early as like Monday, we were like negative five still. So it’s like a strange time right now. Not Bangkok weather.

Michael Waitze 1:52
No, for sure. And one of the reasons why I’m so happy to live in this part of the world is because I grew up in Massachusetts, Connecticut and New Jersey and then moved to Tokyo. So I was in the cold and the snow for most of my life. I’ve shoveled more sideways and driveways than I ever want to see you again. And being in the heat. I love it. Every second of it. Yeah.

Christina Cai 2:11
No more snow. And you know what the hack to the snow issue was to live in a condo building. I’ve lived in a condo building in Toronto for the last, you know, 10 years that I’ve been here and I have not shoveled snow. That’s the hack that you know, you should have used.

Michael Waitze 2:28
Sure. I’ll go back and tell my nine year old self that my dad should have moved into a condo, that would have been great. So I don’t have to look at my sister’s staring out the window of me while I was shoveling the driveway. So before we jump into the main part of this conversation, why don’t we get a little bit of your background for some context, right? Because you’ve been living in Toronto for 10 years. But I guess that means you haven’t always been living there.

Christina Cai 2:55
No, actually. So I’m Chinese. I was born in Beijing, China. Well, my family actually emigrated to Japan when I was three years old. And we lived in Japan. So you know, I lived in Yokohama, Japan, right? For for three, three and a half, four years. Then when I was six years old, we moved to Canada. And unfortunately, I do not speak Japanese anymore. And I became English and Chinese kind of by two languages. And I have a very Canadian upbringing. So I’m from Vancouver. So from like, all the all the West Coast kind of thing going on. And then, you know, when I chose University, I ended up in Toronto at the University of Toronto. And from then on, I think in the last year, like in the last year of school had this idea of like, going to start a company, right. So this is actually like kind of a dorm room from the school kind of concept that came to us. And in the last year of school, we didn’t really know what it even meant, right? We just went Canadian government website and registered a company. It sounds like third year. It sounds ridiculous. But it’s true. He went and you downloaded something like articles of incorporation for free coffee on Facebook, then you have a company now, right? And so that’s the beginning of us, right? And back then, we actually had a little piece of technology because back then, you know, you a lot of a lot of people like your first exposure to startups or technology or whatnot, is, you know, the newsfeed it’s like TechCrunch is all of this and every day, the way you’re reading headlines, you’re like, you know, reading this 700 words, summary of years of work, where you know, people have raised 10 mil overnight, or like couple of what got acquired for so and so. So you’re in school and you’re like, Okay, I have this really exciting piece of machine learning technology. It’s in the space of natural language processing. And now so I’m going to do a startup I’m gonna do a tech startup. So you’re full of this are going through school you haven’t graduated yet and it starts off really simple. So when we first started in the University of Toronto, we had a little innovation space. And I like to tell the story a lot, because it’s literally the size of our current server room, and we shoved three desks in there, right beside the men’s bathroom. So we used to get water in our little kettle, from the little tab. And the building was so old, that you had to physically open the elevator doors, right? Like it’s a physical pull open doors, like there’s not much of those left anymore. The building is you have to now like actually demolishing this building is how old it is. But that’s, you know, where we started. And when we first started, it was just a simple piece of natural language processing tech, what we were doing is we were taking key concepts from Tech, we were parsing it out, so unstructured text, parsing it out and being able to map that across multiple languages. And that that’s all our tech did. So we did not start off as an InsurTech actually. We started off with this, like, full of passion, kind of all we have a technology, technology is going to save the world. This is wonderful.

Michael Waitze 6:09
Why didn’t you and you said we so there must be other people involved? Right? And we can get to them too. Why didn’t the three of you the four of you, however many there were, decide to take this stuff that you were working on and like go get a job somewhere, you know, go to Microsoft or go to Google or something like that, which is would be normal for university students? Like what was it like is your family entrepreneurs? Was the other founders family entrepreneurs?

Christina Cai 6:33
My co-founders family’s entrepreneurial? And I feel like I just read too much TechCrunch. Right. We, we have a phrase in Chinese is called, which literally means dumb courage. And I think that’s what it is, we get asked that question quite a bit. And I think that’s what it is dumb courage, when you don’t know anybody, and you’re full of passion, and you’re just courageous. Because you don’t know. Right? So you know, kids are sometimes like that. And they think they can climb things they can’t because they don’t know better, right? So so that’s that’s how we first started actually piece of technology. Myself, Anthony, my co-founder and CEO, Alex, our CTO, Fedora, our UI UX person. He like chief UI UX, he’s actually the he was just us in the very beginning, actually trying to figure out how is this thing gonna come together? And the first problems that went after was actually Hey, like, we’re going to use this in language translation actually wasn’t insurance related. And we were in healthcare, because we were parsing technical text. Right. And then, you know, because of the ability to do it in multiple languages.

Michael Waitze 7:46
But why healthcare stuff?

Christina Cai 7:47
Technical, because there was there was like, medical ontologies. Right. So essentially, there were webs of knowledge. And you can do that it was really cool. bluntly, put, right? Like health care test was a demand there is really cool to work with healthcare text. So we picked healthcare tech, like there wasn’t, the tech was cool, right? So we ended up there. And then from there on, we realized that we weren’t like, we couldn’t make it work solving this problem, right. But what we did stumble upon was actually people found our ability to work with unstructured Chinese text, Asian text, Japanese texts. Really, really amazing because all of a sudden, there’s all this unstructured medical data, all of that you could actually use, right? And following our progress. This is also how machine learning was also evolving in the background as well, like we were evolving the company, but the greater machine learning, technology and community was also evolving. We were at the earliest and it was also evolving. So essentially,

Michael Waitze 8:51
what year was this?

Christina Cai 8:52
This was like 2016, 2017.

Michael Waitze 8:57
You know, one of the things that you and I talked about offline before we started recording was this idea that you can’t separate who you are from what you’re doing. Do you not find it even slightly interesting. And I don’t know the found the other co founders, we can talk about them too. But born in Beijing, moved to Japan. Living in Toronto raised as Canadian, so it’s English. Japanese enough. Yeah. Used to speak Japanese. I’m guessing if you lived there for three years when you were from three to six, right? It would be weird if you didn’t speak Japanese. Yeah. But now you have all these language skills, and you’re sitting there studying NLP. So you’re thinking about this, and then you move into all this other stuff. It can’t be separated from you. Is that fair?

Christina Cai 9:42
I think it is fair. And I think like, actually, this came up in conversation with my team once. I feel like having to learn to being thrown in and having to learn to brand new languages that are not even remotely close to each other as such as small age had this like, it forced me into this, like, you jump in with both feet and you jump in quickly.

Michael Waitze 10:08
And you’re talking about Chinese and Japanese? Yeah.

Christina Cai 10:10
Chinese, Japanese, English. It’s not so much Chinese that’s like my mother tongue or so you have Japanese and English. Right. So imagine, I think most immigrant kids can empathize with actually showing up at the first day of school. Sure, not speaking a single word of like, the local language. I remember this really, clearly. My mom sent me to school with like, essentially two phrases. My name is Christina, and I don’t know. That’s my first day. Thanks, Mom. Great three. My name is Christina. And I don’t know, right? There’s this she sent me to school like that. And I think that you have to just jump in fully and adapt. And I feel like that adaptability, like is literally in my blood. So in many senses, when I was sharing with my team, it felt like, you know, out of school, why didn’t I go get a job? Why didn’t I do this? Because I was used to just jumping in full hearted, I think, right? And you just didn’t think that much. Right? You just You just kind of did it? And then yeah, much like learning a new language, you kind of learn the ethos and the language and the environment of startup land, right. And you kind of just roll with the punches,

Michael Waitze 11:24
But I think you you have to be a person like you are is uniquely qualified to deal with NLP. Because you’ve been doing natural language processing yourself for years. People don’t understand, I think that aren’t bilingual or multilingual. Just how different the language structures are. Right? So in English, you would say, I went to the movies. Really simple phrase in Japanese. It’s completely backwards to the movies. I went. But you have to understand both of them. So when you go to NLP, you just your word searching, you’re doing all this other stuff. Sorry. It’s just language has been fascinating to me forever. Please go ahead.

Christina Cai 12:03
I remember when when we were working on this, you bring up a really fascinating point, our algos and this is this is mostly like actually Alex’s work here. And my role was actually to make sure his work was understandable by investors most of the times, and he would call what I do suit talk, which is pretty funny. Right? And he would, and I remember this one of the challenges we had was, you know, in English language, we have spaces between words, right? Chinese character, we don’t have spaces. So how do you know? Yeah, when? When the characters make together is a word. Right. So part of the algo was actually cutting words. And sometimes two words, plus, together make a phrase, all of that, like not having spaces was a very unique challenge for itself. Right?

Michael Waitze 12:52
Yeah. So how did you get from the all of this NLP, all the language processing all of this movement into what you’re doing today? Right? Because you said you would be talking to people, and they’d be very impressed with this language processing stuff that you were doing. But not just in the translation part of it right, in the kind of larger healthcare space. Is that fair?

Christina Cai 13:14
Yeah. Exactly. So that’s, that’s when we started getting into the health healthcare space to structure unstructured data to prepare it for training machine learning models, right? And that’s what machine learning was starting to evolve in a way where they were doing like, you know, representational learning all that cool stuff was actually happening, right. So essentially, what we learned is we were able to collapse all the medical records of a person into the way to understand this is people have different data has dimensionality, right? There’s when you think about hospitality data is it’s temporal. There’s like, days, there’s hospitalization events, there’s this, there’s that there’s that so there’s actually multiple layers of data. So essentially, what we’re figuring out is how to collapse all of that into one point in space to actually capture all that data. And now I am radically and grossly simplifying all the complex work does, but that’s essentially what they were doing. And that actually became the input data to make predictions on top of, right.

Michael Waitze 14:20
Once you get that data and figure that out, what do you do with it that’s relevant to healthcare,

Christina Cai 14:26
To healthcare, right? So that’s what that’s, that’s, that’s where we actually walked a few more paths around the way to actually figure this out, right? So in the beginning, we were like, now we can make healthcare predictions on that. We can do 30 Day readmissions, we can do this. We can do lots of cool stuff. Because now that you’ve reduced the dimensionality of the data while still capturing that you can make lots of predictions. Now you get to the next problem, which is like, okay, so how does this prediction actually drive value and where does it drive value? How do you take a piece of technology so in retrospect, we went through this whole thing in reverse, right? Like we had the tech first instead of you have the start with the problem. And then in the end, you figure out the tech to solve the problem. In the end, we had, what we had discovered was our ability to actually make health care predictions was actually extremely useful in the life and health space. Right. And essentially, that’s how we act health Life and Health Insurance space. That’s how we actually came to life and health insurance. And we realize the value and impact that we could actually drive in this space, because a couple of things were actually happening. The first is essentially, like the proliferation of data, there are now new alternative sources of data, the source of data that we’re particularly excited about is a source we’ve called open health data. And what that essentially is, is that it’s the result of governments investing in these eHealth initiatives. And a part of that eHealth Initiative actually consists of a patient data kind of repository. And sometimes it actually, they’re trying to support digital health initiatives. So what they do is they actually have an API that allows consumers to consent their own data to trusted third parties. Now, that’s super exciting, right in the world of data, because that represents a really predictive, highly predictive, and highly descriptive source of data that can actually insurance could actually use. Now those things could actually use to create new products, they could actually be used to underwrite, they could be used to accelerate underwriting, they can be used to personalized products better, like, do all of those things, right. So that’s how we actually came from a simple language, natural language processing into healthy AI and making healthcare predictions. And finally applying that to for impact and actually realizing where the value of our technology and the market it drives in. So it sounds really cool when I say it like this, but this incorporates about five years.

Michael Waitze 17:09
Here’s the thing, right? So you’ve never had a job at a big company. You all graduated from school and just kept working on this business, right? But how do you sustain for so long? While you’re figuring this out? Again, I had a conversation with another entrepreneur a couple of days ago. And to me, entrepreneurship is really about discovery. Right? You have this idea. And you said we did it in reverse, we had this technology. And then we found a problem to associate it with and I guess there’s so many things I want to ask you about this the first being, do you read the tech news differently today now, after being through this for so long? Do you know what I mean? Because everything does happen in reverse? I think, right? You do all this stuff and you’re like, How can I now apply it to a place where it has value as opposed to saying, here’s a problem to solve? And I’m going to go back test or backfit technology to solve that problem? I don’t think it works that way.

Christina Cai 18:01
Yeah, like now, when I read tech news, like, I think there’s a lot more I want to use the word camaraderie, and there’s a lot more celebration than it was in the beginning. And what I mean by that is before you read it, and you’re like so and so did X, Y, Zed, you’re like, whatever, like, and there’s almost like a sense of jealousy, right? Like giving every entrepreneur can do it like that. We even have a say it’s called valuation is not validation. Right? I genuinely think every one of this being honest, like all read valuation, like, and fundraising reports and felt that twinge of like, maybe I’m not good enough for like jealousy or whatever, right? Come on, right. But I think now like, now I read it, you’re like, Okay, if they manage to do that they’ve moved mountains, and I applaud you. So there’s a much stronger sense of camaraderie, because you understand how much is below the surface? 700 words? Right?

Michael Waitze 18:58
Well, yeah, I mean, you said earlier, and I love this. I wrote it down. You’re like, yeah, they raised their $10 million overnight, and I have this philosophy of everyone’s an overnight success 10 years later. Because it just takes that much time.

Christina Cai 19:13
Yeah, I totally agree with that. Everybody’s an overnight success 10 years later.

Michael Waitze 19:19
But what? So now that you figured this out, now that you’ve been through this whole process, I’m presuming that all the founders are still together. Is that Is that true?

Christina Cai 19:27
Correct. Yeah.

Michael Waitze 19:29
Do you feel again, like the camaraderie between the four of you now is stronger than it was even when you were students and stuff looked like a lot of fun? Do you know what I mean, hey, let’s do this thing together. Now. You’re just like, Oh my god. After six years of this, or seven years, whatever it’s been, you’re thinking like, I know you so well, we’re so deep in this. And we’re actually moving now kind of thing.

Christina Cai 19:53
Yeah, I absolutely think so. And I think one element here, is that there there’s A lot of things that it becomes a cadence of working together. Yeah, right. And I think that’s actually really important. You kind of learn how to communicate with each other in a way that like, pushes what needs to get decided faster, right. And I think one thing that’s like, I think that, like, that’s really good. I think that’s like a little bit difficult when you work together for so long is you forget to give them the chance to grow and to update your frameworks by which they operate, because you keep expecting them to work a certain way. But you know, in their own special contexts in the areas that they’re doing, they’ve grown as well. And you’re, if you’re still making, if your mind is still making that assumption, then you’re, you’re you’re like, you’re kind of creating almost reinforcing loops where there doesn’t need to be because you need to allow growth. And I love Yeah, it’s tricky. Like, and analogy, actually, our CTO gave once is really funny. He was like, hey, you know what, like, I have my own machine learning algo model of how Christina behaves, I do this, and Christina will probably x y Zed, right? One day, when Christina does not behave in x, y Zed anymore, I have to go back and be like, okay, something changed. You know, that’s exactly like and you don’t sometimes, especially with people, you’ve I don’t like co founder ship is a very special relationship. I also want to extend it to, like people you’ve worked a long time with, you don’t necessarily give them the room to do that. Right. So I always thought that was really funny, because he was like, you know, we’re just a giant ml algorithm. Okay, I guess that makes a lot of sense.

Michael Waitze 21:41
But this is one of the biggest difficulties of all interpersonal relationships. I mean, marriage is the one which most people identify as they get older, right? It’s like, I met somebody when I was x age. And now we’re 20 years later, we’ve both completely changed and yet, I still deal with that person like they were when they were 25, that’s just not true. Right. So it’s the same thing with co founders. You’re like, I’ve known you since you were 19 or 20. But now you’re 30. You must be a different person. And I love the fact that you’re just one big ml algorithm with Christina doesn’t do this today. I have to go back and check my data.

Christina Cai 22:17
Yeah, I have to check something changed, maybe have to update my models. I thought that was really funny. I forgot what the context was when he was explaining this to me. I thought that was pretty funny.

Michael Waitze 22:26
That is awesome. Can I ask you this about open data? Right, because this is true across the spectrum. It’s open banking, open insurance, open health data and stuff like that. It’s it is really good, right? It’s really powerful. Because you get all this sort of anonymized data, you get these massive data sets would have been difficult to accumulate over time anyway, right. And then you get to run your algos through them or develop the algorithms around them. What do you do as a team to clean that data and make sure that that data, which has been what’s the right word, it’s had been decided to be shared by the individuals that own that data? Right? How do you clean that data and make sure that the ML algos that you’re building or that you’re working with are actually valid based on clean data, if that makes sense.

Christina Cai 23:10
So I feel like our own team would be able to do an entire podcast just speaking about this. And just how messy real world data is. Right. And so I won’t go into details on what they do, quite frankly, because I can’t do it justice, right, or no, but we actually have built this in into an interview question for machine learning people. And the way that we actually ask is, because a lot of times when you’re coming out of school with the machine learning background, you’ve actually worked with datasets that are that are like academic data sets, right? sanitized, right, very clean, very whatnot. And the real world, you’re like, hey, is supposed to be like this. Why is it like that? Right? And essentially, we’ll be like, Okay, how would you if you were to be handed a data set that looks like this, like you expect this? But you got that instead? Right? How would you think through dealing with it? And we’re not really looking for right answer, per se. We’re just looking for a creative way of dealing with it right, or the way that they’re actually thinking, but that’s actually what we’re testing for. So it’s such a big problem, that it’s something that we actively test for. Yeah, right on during our interview process. And I feel like it’s interesting, you bring this up, because I trust for a lot of machine learning startups in insurance probably in in other sectors too. The challenge here becomes, okay, in order for an enterprise like like an insurance company to adopt us, their data is extremely messy, it looks a certain way. We would need to invest like a lot professional services in order to even get their stuff into a place where you can actually do something with it. Then you run into and especially if these are ventures startups, it becomes difficult because you run into the problem of professional services. So it becomes this, like, really interesting way of thinking where you’re like, how do we actually think through as an InsurTech? How do we balance the need to do professional services in order to get them on to a new system, and actually help them digitally transform with our own own need from an investment and from a venture perspective to prove out recurring revenue and exponential revenue growth? Right. So I love that you asked this question, because I feel like it’s essentially a it’s like a it’s debated amongst every startup. And I think everybody has a different answer on this issue.

Michael Waitze 25:57
Right. And that gets back to what you were talking about earlier, the interview question that you ask, doesn’t have a right answer, per se. Yeah, it really just has a mindset around it. Because, you know, it’s funny, you said, your mother gave you you two phrases when you went to school? My name is Christina. And I don’t know. And I think the answer to that question really is, you can still use one of those things that is, I don’t know. But here’s what I would do. Right? Because this is this iterative testing that you have to do constantly. It’s kind of a metaphor for building a company from scratch. The real answer is, what are you building? I don’t know. But we’ll get to that at some point in time. And that’s why I love the fact that you asked this question as an interview question, because it’s really a mindset test for the person who’s trying to answer Yeah,

Christina Cai 26:42
Yeah, it is you’re trying to test for if I threw them something slightly different from what they’re used to, or they’re going to be able to deal with it? And also, like, the reality is, is that the data we work with is extremely messy. Right? We need to do lots of magic tricks on it to get it to something we can actually work on. Can you actually be creative enough, flexible enough? And at the same time, make it work.

Michael Waitze 27:05
Yeah, I mean, I did a whole podcast series on data and data analysis that I called Data Driven. So this is, again, something that’s very interesting to me. And I would love it, if we could get one of your cohorts onto one of your colleagues onto the show to talk more about data. It’s something I think that fascinates people, and is really relevant to startups, because it’s so expensive to do properly. Right? You don’t just get a bunch of data, analyze the data and go home and you know, have dinner, there’s so much work in the machine ops, like ML Ops is almost as hard as the data analytics, right. And most people don’t talk about that. So we did a bunch of work on learning about ml ops as well. And it’s just really interesting to me, anyway, could spend a ton of time on this. Can we just talk a little bit more about data? And then I want to talk about the overall business, if you don’t mind? One of the other things I think about a lot, particularly in the context of distributed ledger, technology is self sovereign data, right? There’s all this open data. But at the end of the day, that data is about me. Or you? And do you think at all as a team about that data has value, and if you attach that data to some sort of blockchain or distributed ledger technology mechanism that people will actually at some point, start getting paid to share that data and extract some of the value themselves?

Christina Cai 28:31
So that’s really interesting. The way that I think that what you said in the the end is people getting value from their data. Right. So whether the technology be blockchain be whatnot, I think the central part we’re coming back to is people getting value from their data. Fair enough, we’re not in a place where people are just willing to consent. Yeah, you have to give them a reason to consent, you must give them some sort of value, they must see the value in consenting. Right. And whether that be, you know, in the traditional wellness applications that insurance companies have pushed out, whether that’s discounts in the premiums for exercising or like 15 $20 off here, there. And there, whether it’s engagement, whether it’s a sense of community, whether it’s understanding their health a little bit, whether it’s whatever. Our hypothesis here is that people are willing to share their data, if you can prove to them that you are trustworthy and using it for their benefit and that they find it valuable. And whether you use blockchain tech or whether you use other tech or whatever the tech it is that you want to put it on. I think the central piece of that is that people find sharing data as a value service, right. Like it’s valuable to them.

Michael Waitze 29:47
Yeah, fair enough. And I think that people will share their data, if they do find value, and particularly if they feel like they’re in control of us in control that their own data as well. And I guess the flip side of As you mentioned before that gathering all this data doing all this ml work on it, applying artificial intelligence and algorithms to it allows you to create new products, new digital products. I’m also very curious about because health insurance is this thing that’s like I subsidize your health insurance. Sometimes you subsidize mine, right? Because if I don’t make any claims at the end of the day, and you do, part of the way, underwriting and actuarial math work is at scale, we’re helping each other out. Is that fair? The risk mitigation, yeah. And the risk pooling? Yeah. But how do we then take all this data and then create personalized products for people within that concept of risk pooling? Which if you can do great AI and ML on it should allow you to do personalization? Like how hyper personal can you get do you think?

Christina Cai 30:57
I think that’s a great question, right? The way that I want to actually answer that in two parts. The first is okay, how can you actually how does machine learning really impact risk pooling? I think the promise of it is essentially more granular ways of looking at health risk, right, more granular ways of looking at health in general. For example, traditionally, you look at exclusion criteria, we used to just exclude everybody who had diabetes, right. But now, we know that it is not the case that everybody who has diabetes is going to be uninsurable right there to high risk, especially because of lifestyle changes, and all of that. Now, the question becomes, how can you actually figure out how to, how do you tell? How can you tell you can’t tell the checklist because the person checking that box looks the same on a checklist basis, right, you need some better way to actually pick it out. So that’s the promise of machine learning where you’re able to get more granular risk groups. Right? For example, though, that diabetes is actually a pretty good example. So that’s like a, that’s an example of risk pooling. Now you’re like, Okay, so if you have insurance product recommendations, really, really personalized? Does that then mean that if someone is extremely high risk, who’s going to send the dice, they don’t just get it? Is that what it means? Right? And I think the way that we think about personalization is giving them like, when we predict the risk, it’s always a percentage of their thresholds, right? It’s how do we actually balance the threshold between what is a tolerable risk versus what that person actually needs? And I think the beauty of machine learning is its ability to get a closer and much better and much better, like almost guess, assessment on that front. Right? So the question becomes, like, if we know this person is below health risk for their for age and gender group, what can we do about it? What can we do about it? If it’s there above their age, gender group, right? Can we actually motivate them to get healthier to come down and all of these things, right? So it’s actually the the personalization becomes the the balancing point between the threshold of tolerance for risk, versus what can actually we offer them?

Michael Waitze 33:22
This data and data analysis is great at the insurance level for them to understand the like you said, the granularity of individual’s health and hyper personalization, then allow them to underwrite and write policies for people that are maybe better or different than they would have been originally. Because like you said, if I just check a box that I have diabetes, and you check the same box to we look the same, right, particularly if our age is the same and other sort of qualifications are the same. But is there a way again, to flip that on its head and say, Wait a second, here is somebody who’s part of that risk pool and I can now give that data to them and the predictive analysis back to that individual all the way down to that granular level and say, you know, if you just stopped smoking, or if you exercise more, or whatever the result of that prediction is give it to the individual and say, do these things. And then your premiums can go down? That’s simplifying stuff, for sure. But you know, what I mean, help the individuals as well, as opposed to just the whole pool.

Christina Cai 34:18
Yeah, for sure. So that’s actually how we do it, right. So whether we do it is when the consumer the goes back to an earlier point, I make value to the consumer, what is the value to the consumer? So when they actually consent their data, what they get in return is a personalized health AI report. Right? It’ll actually show how their health risks compared to people in their age and gender group. Are they a little bit higher, are that a little bit lower? Are they a little bit higher, but still within the relevant shareability? Are they a bit too high and they should get that little bit down? And I think that they’re able to get a healthy eye report. I think the next step there is to actually be like, okay, so are the levers they can actually do in order to actually improve what it is that they’re they’re doing. Right. So I think that’s the I would say that’s the next step that we want to look at. And we want to be careful, we’re talking about health improvement and giving any sort of like, medical advice, because that’s not what we want to do. There’s obviously like, behavior level stuff that are proven that I think definitely can be integrated into this.

Michael Waitze 35:28
Yeah. I mean, I wouldn’t suggest telling an individual person do this or do that. But yeah, again, just saying, people in your cohort have done this to change that kind of thing, right? Yeah. Yeah. Yes. super interesting. Do you feel like with all the work that you’ve done, and your team has done, and I’ll tell you the context of this, I used to work on a trading desk. Right. So every day, we would process like tons and tons of micro little bits of information and make micro decisions on them in real time. And what ended up happening for most traders, and you can ask them about this is that everything, then ended up looking like a trade? Does that make sense? So that if you walked into a restaurant, you’re like, Okay, what’s the best way to get the best table, it all felt like a trade transaction to you? Because you could malt you could process all this data coming into you look around the room, see who was there, see who was waiting and try to figure out your best way in real time to get the best table? Do you feel like you’re so deep in this data analysis in the healthcare and all this other stuff that you’re doing? That when you leave, work, whatever that means these days? Everything looks to you like this big solvable data problem? If you know what I mean, outside of what you do,

Christina Cai 36:39
yeah. Like, what care what do you bring home from work base? Yeah, right. What what’s thought patterns or habits even bring home from work? I think for me, there’s a, I don’t work with the data. Personally, I understand. I think what I do bring home is this concept of Hold on, where are we disagreeing here? Let’s isolate it out. And just decide in one way or another, because it’s probably like, not deciding on this is probably more consequential than just move on. Right? Right. Like that. That’s probably more than mentality that I am bringing home. It’s that we decided on this, like, why are we going mulling this right? Or like, the, like, another thing would be like, there’s definitely thought patterns that that come come back, right, like your the situational analysis, which is essentially a lot of what I do, I assess the situation, try to put a new lens to try to look at the parameters and help drive a decision, right? That’s like a lot of actually what I do. And I find myself like, I’d be talking to like, a friend, and they’d be talking about, you know, something that they’re doing at work. And all of a sudden, I’m like, Oh, this is really cool. Like, I would think about it like this, and like, oh, that’s actually really cool. And that would come back. So that’s kind of like the the lens by which I’m looking at the world constantly. That’s coming home.

Michael Waitze 38:03
Yeah. And that’s kind of it’s interesting, right? Because it takes a while for you to notice that in yourself. But your friends are probably saying like, Christina, you know, I mean, you’re doing this and you’re doing that together. It’s weird when you kind of realize it’s happening now.

Christina Cai 38:17
Yeah, yeah. And it’s so funny, because I’ve never put this this this goes into this, like, I put together like a giant charcuterie, like crazy table for my friends for Christmas, right? And they saw my prep work for it. And they’re like, what’s wrong with you? Because I had like photos from Pinterest that I isolated out what the patterns of like the cheeses and meats were, in order to actually figure out how I should play it on all this. And I’m like, there are patterns, and they looked great. And they nobody could believe is my first time doing it. And I told them, I was like, this is like my startup life. Like, you know, I try to look at proxies of how it is I assess it. I tried to put together something that works for this situation. Right?

Michael Waitze 39:02
Right. And again, this is part of the startup world that nobody talks about is that it becomes such a big part of what you do and who you are, that it bleeds over and spills over into the rest of your life. And that’s kind of why I asked that question because it was the same. You know, for me, everything looks like some kind of media event, right? How can I take this and talk to people about this so that then other people can learn more about what I just heard. Right? I even said this You before we started recording. Can we save that so people can hear it? Yeah, yeah. Yeah. So it’s weird the way just imbues these other parts of your lives. Anyway. We joked earlier about being an overnight success. 10 years on 10 years later, I want I just want to end with this right because I feel like we could go on and on and I just want to end with a little bit of a data point for people and just you raised money last year at the end of last year, so congratulations for that. But what does growth look like to you? Like what is effective growth look like to you and the team Where are you operating? Now? Where do you want to be operating as well.

Christina Cai 40:05
So for us, we’re really glad actually that our first insurance product actually with a health score actually launched in Taiwan. So growth for us looks like number of insurance products that are using our health score that’s sold on the market growth looks like number of new alternative data sources that our health score can ingest. And lastly, growth looks like number of different country markets that we are able to penetrate and enter. So those are the three key metrics of growth for us. Because really, I think we’re using Taiwan, which people like it’s a little fact tidbit of knowledge here, they’re actually like the 11th, world’s 11th largest insurance market, right? They’re actually pretty big. They’re right after Canada, very high insurance penetration. So they’re really great market model a lot of this and prove out the prove out the ability to use alternative data, use it and underwriting and actually carry that same model into other countries across Asia.

Michael Waitze 41:08
What’s the population in Taiwan?

Unknown Speaker 41:10
It is 23 million people. And it’s got to be kind of close to the population in Canada now. And it was 35 million.

Michael Waitze 41:17
So not that far away. I mean, okay, that’s our 50%. But still 25 to 35. Not too far away.

Christina Cai 41:23
There’s like parts of Canada, or it’s just land and no people. Right? We have a lot of land. We have a lot of land. Yeah.

Michael Waitze 41:35
And is the business that you do in Taiwan, Olden Mandarin. Is that and is that purposeful as well.

Christina Cai 41:40
They speak Mandarin. A lot of the especially the folks running digital transformation do speak English. So we have hired for for folks that speak both on our team, right? So when there is a need to actually translate there is someone to actually guide that process here. And it’s funny because our team jokes that they’re all going to learn Chinese by the end of this right. But I think one of the things that we do want to do is okay, like conduct, like are we we need local market presence. At the same time, how do we actually operate as a truly global company? Right? We’re like, does, like how, what part is global? And what part is regional? Like what part? Is there even such a thing as global anymore? Or is this is a purely divided by different functions? Right? Like this? Is the AI machine learning function? This is the data function, this is the data for this country data for that function. This is the mobile application with this country, this the sales blah, blah, is it going to be country specific? Or is it whatnot? So I think a unique experience here as because we went international on day one, right? We’re faced with a lot of these very interesting, and I think, a lot of times, like advantageous as well, because we have 24 hour working and we don’t even force people to work 24 hours, right by as difficulties, but it’s not all bad. Right?

Michael Waitze 43:03
Yes. So HSBC used to have an advertising campaign that said like, think, act globally, think locally, or vice versa? I can’t really

Christina Cai 43:12
Yeah, yeah, it’s on every single Yeah, when you fly, you see it inside of the planes, whatever those things are, what are they called? This is how you know, you didn’t travel. Another thing that attaches to the plane that you walk off, I’m just trying to

Michael Waitze 43:27
think of the word to and I can’t remember, this is gonna, we’re both gonna look it up as soon as we get off this call? Oh, gosh, yes. But it’s really tricky. And I actually think it’s really important, right? I, you know, I read a lot of things, obviously, because I’ve lived outside my home country for 30 years. And because I ran international businesses, when I was in Tokyo, I dealt with Europe all the time, the United States, and also the rest of Asia, I lived in Hong Kong for a little bit. And I also lived in Korea for about a month not long enough to understand Korea, per se. But again, you live in all these different places. And you have to ask yourself this question, right is how do we create a mission for the company that resonates with all of these different stakeholders. And the reason why I think this is so important is because let’s go back to the beginning of your company, it was a natural language processing company that was meant to analyze multiple languages. That was the idea in the tech that was getting built. But over time, you changed now into InsurTech. But you had to take the team along with that. And if there’s not an overall mission, it’s hard to explain to people why there’s a change, particularly as the company grows, right? When it’s just the four of you. It’s easy to just like, Okay, let’s go do this. But as you get global, but have local people on the ground in different places, like explaining that mission just gets harder? No.

Christina Cai 44:45
I think in the beginning, yes, we had a piece of tech but the mission statement itself was actually make the world’s health care data useful. And that was actually where we started. That was the mission like make health care data useful. That’s what it was. So when you put it like that, and now it all makes sense, because we figured out how to make it actually useful. And the more important thing is we actually went deeper and figured out, why do we want to make it useful? First of all, why we want to make it useful? And like, Okay, how is it useful? Right? So our current mission is actually to use bleeding edge technology applying leading bleeding edge technology to improve the health and prosperity of the next billion people. Right. And those words were actually selected really deliberately, right, we select the word apply, because, well, we’re not research for research sake, we’re applying things. It’s all about the application. And it signifies our understanding, and actually our journey of getting a piece of tech into its application, right? Like that part is hard, like you think you’ve now found a problem solved. You’ve got to apply it. Now. I think our front, our growth team can really talk about how hard it is to get adoption training all those sales agents. Oh, my goodness, right. So apply as a key word here. And we use leading technology, not AI actually, because, you know, why not? Blockchain, you mentioned blockchain earlier, like we used leading tech leading edge technology, not AI. Because whatever technology allows us to get there we’ll use and we chose health and prosperity, health because of our roots in health and prosperity. It’s a very special word we actually chose. Because, you know, it’s like insurance is a financial instrument. Right? Right. That’s what is I think, is miss a lot of times like prosperity, right? It builds prosperity. And in the future, how do we actually link health and wealth which is the theme here likely prosperity is what we’re going after? And for the next billion people? Right? And today, on our website, you behind you really see ensure the next billion people because insurance is our first stop, right? So when we look at the company mission, like where we really evolved it from was make the world health care data useful, right, which is like, okay, that’s cute. But what does useful mean? I think as we evolved in that process, we got really deliberate about picking words, applying bleeding edge technology, to improve the health and prosperity of the next billion people is where we came to.

Michael Waitze 47:21
I think that is the perfect way to end I want people to leave with that thought. I want to thank you, Christina Cai, co-founder and COO of Lydia AI, for doing this today. I really appreciate your time.

Christina Cai 47:33
Thank you so much. I’m honored to be here.

Related Podcasts

Episode 224
X