3429: How Credit Karma Scales GenAI to Power 60 Billion Predictions a Day

[00:00:03] What does it take to deliver more than 60 billion AI predictions a day, while keeping the trust of over 140 million members? Well, my guest today is the Head of Engineering at Intuit Credit Karma, and he leads the company's largest organisation to make that happen. So his name's Wan, and in our conversation today, he's going to share how he and his team are modernising their data platform,

[00:00:30] building AI-first experiences and merging the strengths of Intuit and Credit Karma's ecosystems. I will also explore how fintech leaders can balance speed, personalisation and compliance in a world where AI is becoming the default expectation. But how are they doing this? What are the challenges that they've come across along the way? These are a few of the things that we will learn from my guest today.

[00:00:57] Now, before I bring my guest onto the podcast, I want to thank our sponsor, Careerist. Because imagine being able to work remotely, earn a six-figure salary and finally have room to grow in a new career without having to learn how to code. That is what Careerist QA Bootcamp makes possible. And in just four months, you'll get live training, a 100% remote internship and personal coaching to guide you into a tech career.

[00:01:25] Now, right now, there are 37,000 QA roles open in the US, and with starting salaries as high as $105,000, it's an attractive option for so many reasons. And Careerist graduates are already working at companies across 42 states. And if you have a quick look on Trustpilot, users over there are rating the programme 4.7 stars. And here's the real clincher. It's risk-free. Land a job or get your tuition back.

[00:01:54] But seats for the next class are filling very quickly, but you'll find the link in the show notes. Please check it out. I'd love to hear how you get on. Maybe this could be your moment where you land that career in tech that you've been thinking of or convincing yourself you don't have the right skills. Because believe me, this is open to everyone. But now, back to today's show, and let's jump into the interview with my guest today. So, a massive warm welcome to the show.

[00:02:21] Can you tell everyone listening a little about who you are and what you do? First of all, thank you for having me. It's an honor. My name is Juan Agus, and I lead all the engineering functions here at Credit Karma. I've been here for five years, five and a half years. And a lot of folks, especially in the U.S., know Credit Karma as the free credit score company.

[00:02:48] But we go way beyond that, which I'll be excited to talk more about that later. But, yeah, the mission is to help all our members, which we have about 140 million of, to make progress in their financial life. Well, there's so much I'm looking forward to exploring with you around that and all things Credit Karma. But before we do, I always like to dig a little bit deeper on the origin story of my guest. So, let's start with your journey.

[00:03:17] I mean, from Walmart, Macy's, to leading the largest org at Credit Karma. I'm curious, what drew you into fintech, and what keeps you excited about this space? Absolutely. So, one way to draw the connection from e-commerce, which is I spent about 10 years on, to fintech. And I don't think this is 10 years at all.

[00:03:45] So, in e-commerce, oftentimes you have to think hard about how do you sell, right? Because it's all about selling. If you're going to sell AA batteries, okay, everybody needs one. You just make sure it's there. The price is reasonable. But when you're trying to sell something that's more up there, like sofa set, right?

[00:04:08] Something that's like a high-ticket item, then you really need to contextualize it. You need to really build the intent, all of those things. So, when you think about fintech, it's about that. So, when you think about it like credit cards, it's kind of like everybody needs it, just like you're getting your AA batteries. But then, when you start thinking about home loan, well, do you really need a home? Some people may not. Some people do. And you really have to contextualize. You really have to work for it.

[00:04:36] And you want to make sure it's the right thing for them as well. Auto loan, right? And in-person loan. So, that's one thing I would connect the dots there. And obviously, from my perspective, I mean, fintech brings a lot of complexity that I haven't had to deal a lot before in my previous e-commerce career. Things like privacy and security, which is very, very important. And I've always been drawn to like, how do you move fast?

[00:05:04] And still move fast in a world where you absolutely cannot make mistakes. Because in e-commerce, if you send more than one items that the customer order, okay, not the end of the world. But in a fintech, you have to be careful. Like, the room for mistakes is zero. And again, you have to protect the customers or, in the case, our members' privacy and data.

[00:05:29] So, the complexity there that's still moving fast and the potential of making an even bigger impact in their life, to me, it's the, you know, it's a draw there for fintech. And fast forward to present day, and it is all about credit karma. And you've been tasked with bringing together two engineering cultures and tech stacks all under the Intuit umbrella. So, what's been the most surprising challenge in that process? And what's worked well?

[00:05:59] I'm sure you've had more than a few stories during this process. But anything you can share? Absolutely. So, yes. Thanks for that question. Because I do want to, you know, mention that credit karma, Intuit acquired credit karma about the end of 2020. So, that's what is at about four and a half years ago now. Yeah. And these are two different tech ecosystem.

[00:06:28] So, credit karma uses GCP and, you know, exploit a lot of the cloud native features of it. You know, again, for making sure that, you know, we get the best, the best bang for the buck. And Intuit is on AWS and a lot, a lot of other kind of like Intuit built tech ecosystem as well. But really the value, right? That we've realized that through the journey.

[00:06:57] The value is the togetherness of it, the ecosystem for the member. And we started. And like you said, Neil, it's like integrating those is not easy at all. But the moment we realize how much value you can unlock, it's all worth it. And I think we've been well on the journey now for about two and a half years. I would start with this.

[00:07:25] For a lot of folks, for a lot of engineers, and, you know, I love board games. And, you know, for those people, when you play board games, when you play the Euro side of the board games, Neil, like that part of the continent we come from. Euro board games are usually much more complex. It's not your monopoly, right? There's a lot of rules. But, you know, so in this case, I look at it as playing two board games, right? So connected to each other. And it's even more fun when you're, like, really trying to figure it out.

[00:07:55] So that's the way I approach it. Now, getting into, like, a little bit more details here. So how to make this capability seem seamless? We first have to figure out how to integrate, federate the identity system, right? So because we don't want that. That's where the seams are usually shown.

[00:08:18] When you move from one capabilities to another, and then you have to, like, log in, you have to 2FA and stuff like that. So that is our first thing. And I think we've been largely very successful on that. I think the, you know, we, you know, in the recent taxes on April 16, ending there, we saw the stats about 70% of the credit comma members are able to use throughput tax without even logging in. Like, it's just like, it's just, boom, just get in there.

[00:08:47] And that number is going to get better. And it's going to unlock a lot more. So there's a lot up, up, down the line. Identity is one. And then I'll just close it with data. Having data, not only visibility on both sides, because Intuit has a ton of data. We have a ton of data. But really, how do you rationalize it?

[00:09:10] So we know that the NIL over there in Intuit is the NIL over there in CK, making sure the same NIL use. But then also, we have all these attributes about NIL. We have thousands of. And then we're trying to figure out which are the canonical. And then we're trying, in the mission of trying to really understand NIL to serve you better as well. That's a lot of work. But that's fundamental, right? Foundational.

[00:09:39] So if we don't get that right, all the features we built on top of it, not really going to resonate. So we really need to get, make sure, like, you know, combine knowledge about you so that, you know, we can be, like, hitting bullseye when we start building features on top of it. And you've got over 140 million members. So Credit Card has got this huge data footprint, which must be a huge challenge on its own.

[00:10:08] So how are you modernizing the data platform to deliver more personalized and more meaningful engagement at scale? Because for a lot of business leaders listening, they'll be interested in how you achieve this. It's a massive achievement, right? Yeah. The short answer is the work never stops. It's always there. It's a continuous journey. So I cannot take credit. I mean, I've been here for five and a half years.

[00:10:35] When I came in, one of the reasons why I think you asked earlier, why FinTech and specifically why Credit Karma, why Intuit? It's the amount, A, it's the amount of data. And B, what do you do on top of it? So I'll give you a bit more concrete things.

[00:10:54] So in Credit Karma, when I joined, I think the stats that blew me away was Credit Karma generates about 60 billion predictions a day, right? And what do we do with these predictions? I mean, this is what I came back to, you know, previously talked about trying to figure out what's best for you in your current financial state. Are you trying to buy a home?

[00:11:24] You know, is your credit card the best credit card for you in your wallet right now? What about loans? Like, can we lower your interest rate? What offer is best for you? Because we don't want you to be like, we are not here to make you click on offer on like, hey, here's a credit card. We don't make money out of that.

[00:11:40] So, and we know that if you just click on and apply for a product, for a financial product, and you don't get it because we poorly done our job because, you know, your credit score is going to come down, right? And it's loose-loose for everyone. Then you have to wait for your credit score to come back up again to figure out, you know, which financial product. So it's really important for us to get it right. Right. So to your question, it's kind of, you know, what do we do with this data?

[00:12:09] Well, it's the data itself. We continually, you know, when you have a massive amount of data, you're going to have to, it's very important. It's like you have a big house. You have a lot of furniture. It's like, well, first of all, like you need to, you know, you're not going to use all the space and all the furniture all the time. So you need to make sure like you really rationalize like why using the data and then what are you going to do with it? And last thing I will close is that the infrastructure underneath it continually, you know, evolving. Right.

[00:12:39] So like at this point right now, we are working on migration to, yeah, migration to a lot of the NVIDIA Triton kind of server, infraserver that we've been looking at. And so we are well on our journey. And that's something that because the massive amount of data, that's something that we started like a couple of years ago. So when it was like brand new. Right. But so we have to be on the cutting edge.

[00:13:08] But knowing that the cutting, whatever is on the cutting edge is not going to probably scale to what we need. But we have to start because the massive amount of data usually take us quite a while as well. Right. To be able to to really productionalizing it. But at that point, that what used to be cutting edge is now ready for scale. Right. So so we just have to like keep keep at it, which makes the job really fun. I'll be honest. And we're going to have a lot of AI involved, making things easier.

[00:13:36] AI first experience, another powerful phrase. So can you walk me through what that looks like in real terms, especially when you're trying to balance innovation with regulatory complexities? A very fine line to tread a difficult balance to achieve, I would imagine. But tell me more about that. Yeah. Yeah. That's oh, my God. Yeah. That's giving me a lot of white hair.

[00:13:57] You know, how to make sure our AI is safe, you know, because we talked about earlier about, you know, you know, if you're an e-commerce and I'm not going to be an e-commerce here. It's just like, but this is the feeling that if you're in commerce, your AI made a mistake. All right. I mean, don't get me wrong. Repercussions are there. You know, wrong things at ship and charts. You can always refund it. But if you're in space like ours, right?

[00:14:22] So I feel a lot of like, you know, empathy with Tesla, of course, you know, because if they mess up, I mean, it's life's loss. I think in our case, if you mess up, well, it's people's financial life, which is absolutely you can't. And so here's the way I look at it.

[00:14:48] Andre Carpati, right, who coined the term, like he used to work for, he's a pretty famous speaker. I think all the audience probably know who he is already. But I would say he had this famous talks about what software 2.0 is a few years ago. And then recently he talked about software 3.0. So let me just go through that very quickly. So software 1.0 is like you writing lines of code, right? So that's what everything, what writing software is. And software 2.0 is what he means.

[00:15:18] It's like how do you use machine learning model, neural nets, all of those things, right? So your traditional machine, traditional AI, so to speak. All right. So and then software 3.0 is using Gen AI, using LLM, you know. So and he had this talks, people can look it up on YouTube and podcast. It's very famous. You know, to create, to deliver value, typically it's not only just one or the other.

[00:15:46] I think it's a mix, right? Everybody started, you know, in the old days writing it with 1.0. And then, so in this case, he made an example of FSD, right? So everything is just like, you know, C code, a bunch of like C++, C code. And then you start replacing certain functions with neural nets. And then over time, those functions grow and grow, grow. And then you ended up having to, you know, you can just start deleting your 1.0 and making it a footprint smaller.

[00:16:15] And then now you have 3.0 coming in, right? So, but it's not going to, because of the safety, you want to make sure, you know, that is like really, you know, got rails around it, observability and all those kind of things. So, so the job is making sure like, you know, you, you know where to use the right tools. And then for CK, we started out with, with a good mix of 1.0 and 2.0.

[00:16:40] Like when I started, it's all, we already like rocking and rolling with like this 50 billion, which is now 60 billion dictions early, which is basically your software 2.0, right? So you come in, you see a credit score, you see all this experience that that's 1.0. And then, but when, when we are trying to like behind the screen, predict like what are you trying to do with your goals, which offer is best for you. And you're going to get the proof, all of those kinds of things. That's 2.0.

[00:17:05] So when, about three years ago, when LL, you know, when open, open AI, chat GPT 3.5, you know, really hit, hit, hit, hit, you know, hit the masses. We, you know, I think in three months we released kind of ask me anything like a chat bot that can talk about your financial life. And the aim there is to learn, right?

[00:17:29] So I think that was what we thought about as like the best way to, to figure out what our members actually need, need this new technology for. Because it's, you know, it's very nascent. I think at that point, I think everybody knows. So, you know, so, but that's the point of it. Like we quickly launched it and we're trying to learn what people are using it for. You know, we observe what's going on. We look at the efficacy, how safe is it, it is all this kind of thing. So we were able to do that.

[00:17:57] So again, like I thought of that as like a 3.0. And then from there, we've been evolving those features, looking again at the safety part of it and what the members actually find valuable. So we launched features such as tax advisor, wallet analyzer, looking at the credit cards in your wallet. And, you know, hey, maybe there is a gap, like looking at your spending habit. And maybe there's a gap there where we can, we can help point out. It's like, hey, you are eligible for this 2% cash back.

[00:18:26] So, but you really keep using this credit card that you had for 15 years. Maybe you should consider that. We've looked into debt consolidation. So we look at a lot of different use cases, which in itself is now evolving into agents. But, you know, there's a lot of these things, you know, still, I want to like put a dose of realism here as well. We want to move fast, right? But we also want to make sure that we've done it, we are doing it in a safe way. And, you know, we can observe.

[00:18:55] And from there, that's how you improve these things as well. So, but yeah, it's just basically a mix. Like, it's like you're a chef in the kitchen and you need to make sure you have the right ingredients at the right ratio as well. And looking at everything that Credit Karma offers your users, your team has to support everything behind the scenes from credit cards and loans to auto and mortgages, etc.

[00:19:21] So how do you manage engineering focus across so many different verticals without or while maintaining a seamless experience for the user? Because as a user, you kind of take it for granted that you've got access to all these different things. But it must be quite challenging behind the scenes. Yes.

[00:19:40] We have engineers who are, first of all, we want, when we look at our engagement survey, typically, not typically, it always is. The number one, our strength has always been mission, the belief in a mission. So whether you're engineers or marketers, it doesn't matter. The mission is what's the most important. We really feel it, right? I mean, like, I'm an immigrant.

[00:20:09] And the first time I learned about, you know, a credit score was through Credit Karma. And I learned how to make financial progress very quickly from the app as well. And we know that's not a one-off experience. Like, millions and millions of people have told us that's the case. So I think that's the first thing, first and foremost. So everybody understand what the mission is, which means that, you know, the importance of getting it right,

[00:20:36] the importance of protecting the data and the privacy, it's like number one. So that's non-negotiable. Okay. From then on, if you're a product engineer, so we, you typically, you know, we want to make sure that you're an engineer, first and foremost, but you also have the product head on and really have that empathy of, you know, so if you're an auto learner, I think you mentioned a bunch of articles there.

[00:21:06] So we also are in Canada. We also in UK. We also have Neobank. Like, so we do have like this broad, you know, very, very large surface areas. And now with Intuit, we have PAX, which is our biggest vertical, actually. Right? So, and PAX is, you can, it's like very, very deep, very complicated. You know, it's a lot of this kind of thing. So, so we do want to make sure where we have our engineers who spend the time on the engineering side, but also think through what, what, what, what depends on what vertical they're on,

[00:21:35] what exactly it is that, you know, our members need to do. So those are on a, on a, on a, on a product engineers, um, working with the product, having that product mindset. And so we do have data science team and the machine learning again, like we have to make sure that, you know, they understand like, um, to deliver the best value of the members. We always want to make sure they understand like how to translate that into what they do. Um, so horizontal functions, security and platform, uh, a lot of those kinds of things.

[00:22:04] I think those are, are, are all drawn from that, from that value. Uh, but then, you know, they, but they themselves then translate into what's really important for them. Um, and then we want to make sure those are measurable as well. Um, and then, you know, lastly, it's, um, it's a culture, right? Um, I think when you talk, you, you know, uh, you, you talk to maybe the most amount of tech leaders, maybe anyone in the world, um, and, and only they can tell you that, I mean, at the end,

[00:22:32] at the end of the day, um, you know, it's, it's the, you know, it's like, uh, culture that's, well, you know, the culture here is that, uh, you gotta know, you know, um, you gotta know your stuff, right? The craft skill is really important. And, but number one, number two, uh, instilling that sense of, um, uh, you know, uh, I don't know what to call this under the, you know, kind of like this, uh, uh, the culture of debrief. So we've talked, you know, we had this blue angels pilot come to us and, uh, you know,

[00:22:59] and, and they, they, they all talked about like how every performance they went into a room and being very self-critical of themselves, uh, what they can do better. And it's not a finger pointing, like they think point their finger to themselves and go really deep into like what they can do better. And I think that's what I love about our culture. That's kind of what we want to instill. It's like, there's always ways to do better. Um, it's easier to, it's always easier to think about how other teams can do better. Uh, you mean, you can give that feedback, but then, you know, but the first exercise should

[00:23:29] be like what you can do better. Right. And then collectively we think about like a system together, like how we can like, uh, do better as well. Um, so I think that, that culture of like making sure that you know, your stuff, you know, what you can do better, um, and, and then, and then work together. So that, that build that trust among each other, because I do the days like, yeah, you're right. There's data science. There's, um, uh, the ML ops, there is a security, there's, you know, core services. There are so many of them.

[00:23:58] So, but it starts with like the belief of the why and then that culture of like, could I have done better, uh, you know, and, and let's what, you know, let's have like a very tactical plan to, to keep, to keep getting better at it. And listening to you there, I think one of the repeated themes in our conversation today is how behind the scenes at Credit Karma, it's balance that plays such a massive part in your success and delivering those experiences that as users, we're probably guilty of taking for granted.

[00:24:27] And also I would say culture is often the invisible engine behind, behind high performing teams. We always talk about the technology, but it's the culture that needs to bring that to life. So how do you foster things like autonomy, trust, and velocity all while keeping technical and ethical standards right at high? Because once again, it's balance, right? Yes. No, you're absolutely right.

[00:24:51] Um, I feel like, um, you know, on autonomy, um, I do feel a bit more fortunate at that part of it, partly because I, I'm, I, I value autonomy really highly. So to me, it feels like I want to make sure I treat folks the way I want to be treated. Um, so, you know, it starts with like, um, what, you know, uh, if I don't have to be involved,

[00:25:18] if I don't have to make that decision, you know, it's like, why? I mean, so let them like, so starting about like, what are the, the parameters there that I really do not need to be involved in most decisions you don't really, like I don't, right. Uh, like I think it's pretty famous now, the whole one way door and two way doors. Um, so, so those things are, you know, I always index on that and making sure that it's not just about autonomy, you know, brings more engagement, but it's also learning, right?

[00:25:48] I mean, uh, and, and, and I like surprises too. Good surprises. It's like you let people, you know, and they, they do it differently than you. And then the outcome, a lot of time is even better. And that's nice. I mean, so I think it starts with like, just, you know, uh, understanding the value of autonomy is not autonomy for autonomy sakes. Um, what is the value and, you know, do you really need to like believe in it? And then, and then after that, it becomes, you know, part of your day to day.

[00:26:14] Now that said, you also need to make sure you have a very clear, um, understanding and guidelines when, you know, you need to, you need to step in. Um, you know, just, you know, I don't know if you have kids, Neil. I mean, you know, it's like when I have two now, 21, 15 years old, I don't have to do this, but when they were, you know, uh, younger, I think I have to like step in. I was like, Hey, it's about when it's about the health and safety, maybe there's one more I forgot, but like, I'm sorry guys.

[00:26:41] Like I have, I have to be like, you know, your, your, you know, mean patterns here. Like, it's kind of the same thing. Like, I mean, like I thought again and again about making sure that, you know, the privacy and the safety, uh, a lot of the time is about, um, contextualizing as well. Right. So I forgot to mention that actually to give people the best autonomy, you have to contextualize is very important. Um, if they don't know why certain things happen. And so you have to be, you know, they're not going to be able to do a good job.

[00:27:08] And then, and then a lot of people ended up blaming autonomy or the people executing. No, maybe it's your fault for not contextualizing it enough. So I think contextualizing is very important. Right. And part of the contextualization is to also let everybody know like what parts of things that are non-negotiable, uh, if, you know, if you're negotiating too much with your, um, you know, with the, with another engineering team or whatever other teams, what is the rule there? It's like, you know, how long is too long?

[00:27:36] Um, again, safety, I would talk about that privacy and then all those things. And the rules like, uh, when you're scaling, I mean, this is very important too. We do experiment a lot, right? We experiment a lot and we go fast, but not all experiments are, are, are, um, successful. Actually, a minority of them are successful, which means like when you're, we have to move fast and when, how do you do that safely? But then when you're ready to scale, it's another thing entirely.

[00:28:02] Like we, now we have to bring in all this like processes, you know, which is the reason there is, is for, for good reasons, because scaling in, in our world is something else, you know, with the amount of data, you know, and a lot of those kinds of things that we have to deal with. So I think, so I think that's as long as people understand, right. Here's the paved path. Here's the rules of the road. Um, and then you contextualize to them, then you, you, and then you, you have to reinforce that by letting them explore.

[00:28:29] Um, and then when they come back, when, you know, make mistakes or whatever, you have to celebrate it. I think very quickly people figure it out like, oh, okay. You know, this is, this is the way it works. And most of the people that I know, you, you know, will, will, you know, they wouldn't want anything else than that. So. Love it. Something else I wanted to highlight here is I think traditionally speed is always been seen as a competitive advantage across the tech industry, but on the flip side of that in fintech trust is absolutely everything.

[00:28:59] So again, back to that B word again, balance, how do you strike that balance when developing new products or platform changes? Yeah. Um, uh, we have, uh, uh, I'll, I'll, I'll go over the, uh, the, the, the, the, the, the boarding part first, which is like necessary, right? So a, uh, you gotta have the process. Um, you're going to have ways to make sure the process is followed. Uh, you know, there is no magic behind it. You know, this is the whole checklist, you know, mentality and manifesto.

[00:29:27] Like we, you know, we've, we've developed that over the years and it has worked for us, you know, knock on words. Um, you need to have a team. So this is a process of the tools in our culture, in a team. So we have, um, uh, we have security team, uh, we have Intuit, uh, security team, which is, uh, a security team is part of as well. So we've been able to like, uh, cross, uh, pollinate, you know, and, and, and from both

[00:29:54] sides and, and, you know, trying to like, make sure we have the best of both worlds over, over there. Um, and, and people who are, uh, whose job is just that, right. So, uh, making sure there is no conflict of interest. Um, and, um, and then we always have to go back and look at the data, right. And then to see, and then, uh, what, what works, what didn't. So I'll give a brief example here, you know, just to like, make it come to life.

[00:30:20] Um, you know, everybody now is trying to figure out how, how to like bring Gen AI tools to, uh, to their engineering teams. Um, so if we had used, uh, our previous process to vet, um, uh, third party tools, um, we would not have, uh, onboarded Cursor or Windsurf or Copilot because we, we are, you know, we have to be conservative. We have to make sure, uh, that those things actually add value. And then there's, you know, it's like a thousand other things we have to look at.

[00:30:50] But I think, you know, but we know that I think, you know, for, uh, for where it is right now, uh, we have to adjust. So I think we've been able to work, uh, with the legal team and security team and, you know, making sure that we, we balance that. Like we, we've been able to like quickly onboard these tools. And a lot of the time, Neil, I think it's a lot of it is like, yeah, there is, there is no, again, there's no magic pill here other than you're going to have to understand it's a risk reward race, you know, uh, equation.

[00:31:17] Um, and then, um, uh, if you're taking on more risk, you also need to make sure the observability is there. Right. So, and, and so you can react very, very, very quickly. Uh, there is no such thing as a hundred percent secure system, but, uh, what is there is that, you know, the mentality of like, um, uh, bad things can happen, but you want to be there. You want to be the first one to catch and then, you know, uh, nip it in the bud as quickly as possible. So, so I think that's, you know, uh, we it's a, it's a, it's an equation now that, that we

[00:31:46] are very comfortable moving, you know, from one spectrum to the other, uh, because we have, you know, we, we've, we've developed, uh, the muscle, what it means to be on one extreme end to the other. And we are comfortable kind of operating in between as well. And if we were to look a little further ahead, I'm curious, what are emerging technologies or indeed platform evolutions excite you the most?

[00:32:11] As you begin to build the next phase of credit comm as intelligent, uh, financial guidance, what, anything that excites you out there, any technologies or anything out there that got you right? Yeah, absolutely. I think, uh, again, like I'm, I'm going to join the herd here. You know, I, I, I tried to be, you know, snowflake special sometimes, but Hey, there's some, you know, sometimes the, uh, the crowd is right. I mean, agents, you know, um, uh, so, you know, into it is, um, part of the contributing,

[00:32:40] uh, group, uh, on the A2A agent to agent, uh, specification. Uh, I think there was an announcement a few months ago during, um, uh, during Google's, um, um, you know, I forgot the name of it, but anyway, uh, yeah, the, the main Google conference. Um, so that is for real, right. I mean, so in the FinTech world, um, in London, like in UK, you guys have open banking. Um, so we, we have a UK, uh, credit comma product as well.

[00:33:09] So, uh, so that is a big advantage. Uh, so we, so because we have a team there, I mean, like, you know, uh, I'm always jealous. It's like, Hey, in UK, uh, some of these things that in the U S is so hard, like over there, it's just, you have like the way you guys do auto insurance as well. I mean, anyway, there's a lot of things to like about the, the forwardness of the, of the, of the UK, uh, uh, uh, law there in terms of the financial data in the U S I feel like we're still lagging. So why am I mentioning this?

[00:33:39] It's because I think when we talk about agent to agent, we didn't talk about, um, you know, having, uh, MCP, uh, with, uh, so a lot of these things is not less, it's about the technology, right? I mean, if you look at the spec, it's kind of like, Oh, okay. Like what, what's new there? Well, what's new. I mean, it's simple. It's pretty easy to understand, but what's really exciting though, is that it's the consensus, right? So, uh, suddenly, um, companies like into it and I don't know, I look at the list

[00:34:09] of, it's like, is everybody there? So we, so my hope is that without federal regulation, like heavy handed all in and taking years and years, we just take a long time. Anyway, this is more about like, Hey, uh, if the critical mass of the, of the, of the, of the, of the fintech players, whether you're a big bank or a small startup agree on adopting this, um, then you can basically, um, that's the fast, faster way to, to create an open

[00:34:36] banking, like, uh, environment, environment that's enabled by open banking, like in UK. Okay. So, so it's, yeah, this is the technology is there, but you know, the consensus needs to be there, which I think is creating, then you being able, having like agents to agents talking to each other and then using tools to MCP protocol, all those kinds of things. Um, it basically only spells win for the, for the users, you know, for all of us, for our customers. Uh, but again, like the safety part of it needs to be there.

[00:35:04] I'll just going back there because it's easy to, it's easier to build fast. Uh, but it's a lot more interesting and harder to make sure that, you know, it's the safety is there. So I'm really excited to see where it's coming, uh, where this is going. Um, there is so much, you know, potential there, but yeah, let me stop that. Oh, one thing is that, uh, I'll just add that I think into it from, um, you know, uh, when we talk about agent, I think there's a lot of like, uh, uh, most people, most companies

[00:35:31] that are building agents, as far as I know, it's, you know, there is no such thing as a, uh, fully autonomous agent. Right. So, but, but a lot of agents are like impressively autonomous. Um, so, but a lot of people are also building agents that are, um, you know, it's a glorified version of a, of a workflow engine. Um, and the truth is it's, you know, it's probably something in between. Um, you know, it's like, if you really build an agent that is actually useful and doing

[00:35:58] something for you, uh, not just like words, you know, just like advice and, you know, chat and stuff like that. Uh, the truth is you're going to have to, to mix those. And I think into it has done a really awesome job. If you saw our latest, uh, press release there, how many agents are right now available on QuickBooks. Um, that is impressive. I think those things, uh, like the bookkeeping agent, you know, uh, uh, uh, payment agent,

[00:36:24] an accountant agent, um, those things are, um, uh, I think is, is really like you have to start somewhere, but where it is right now, I thought, you know, the team that has done an awesome job. Absolutely. Love it. As you said, we are living in an incredibly exciting time and all this talk of AI and agents, agentic AI is incredibly exciting, but it can be challenging just trying to keep up with the pace of technological change right now. So before I let you go, I'm going to ask you to leave a few tips here.

[00:36:53] Where or how do you self-educate? How do you keep up to speed with everything? Um, this, uh, we live, I mean, uh, when, when, um, I remember when I was, uh, I mean, right now, first of the short answer is just like, um, how do you not self-educate? Right. I mean, like when I grew up, uh, I was born in Indonesia and, uh, when I was like, I got my first computer when I was eight years old. Uh, and I remember like, um, you know, there's magazines back then. And this is like mid eighties.

[00:37:23] I'm dating myself here, but you know, going to the library and then going to magazines and then like PBS. I mean, like it's, you have to work a lot. You have to learn, you have to figure out how to learn. And it takes not just that. And then you have to spend a lot of time these days. I mean, you know, you throw a stone. I mean, you open up your, you know, you go to YouTube, you go to conferences, you talk to people, you talk to your team. I think it starts with like, make sure you're very clear, like what you're passionate about.

[00:37:53] Um, you know, um, the, well, first and foremost, uh, be good at learning, you know? So if there is a new thing, um, and you need to, to get to, to a decent level, 60, 70%, uh, make sure you're like, figure out how, you know, most people can do it. Like, you know, I don't have to tell you how, but you know, to get to a decent level, you know, uh, very, very quickly. There's so much resources. Number one, it's all just coming like from whether you want it or not. Right.

[00:38:21] Um, like I said, YouTube books, audio book, whatever it is podcast. Like there's a lot of them. The, what I, what I found, you know, I think it's trickier. I think it's like when you're trying to pick and choose now, like what areas, because from good to great, it takes a lot more effort. Right. So you really have to be very intentional. Uh, what areas do you think, you know, because you want them all, but then you can't have them all. So what areas do you think like you really want to invest time? Um, so this whole explore versus exploit, you gotta, you gotta balance it.

[00:38:51] You gotta keep exploring. You gotta be good a lot in a lot of things, right? Because otherwise, how do you know which ones do you want to be a great at? Um, and then you exploit, like you figured out like which one, like pick two, uh, pick one, whatever. One, two, three, three is the most and start getting deeper and deeper into it to a point where it felt like, you know, there is a reason why you want to master it and you actually get, uh, get what you need out of it. Um, but again, like I think that all the mode of learning are all there in front of us.

[00:39:20] Um, so we are living in a, you know, embarrassment of wealth right now in terms of like, you know, information out there. Yeah. And I think that is a powerful moment to end on. Thank you so much for, uh, joining me on the podcast today and sharing your insights for people listening, wanting to find out more information about all things, credit karma and the latest developments there. Where would you like to point everyone? Um, well, uh, credit karma is an app, uh, a download it.

[00:39:47] And then we also have, uh, uh, our, our website, uh, you know, where we have a ton, I think of information about your financial progress, uh, into it. Of course, uh, we have a very active, uh, I think engineering, uh, blog. Uh, you can, uh, take a look at that as well. Uh, but yeah, um, uh, really, uh, an honor, Neil, uh, it's been a pleasure, uh, talking to you, uh, this morning or this afternoon for you.

[00:40:15] Uh, the honor is completely mine. I will add links to everything there, including the engineering blog too. I think that'll be interesting to keep up to speed with. And I, for one, have loved learning more about how Intuit and credit karma are modernizing your data platform to unlock new levels of personalization and engagement. And with credit karma's 140 million members, it's an absolute staggering number there. And the technology and all the things that go on behind the scenes is great hearing about

[00:40:45] it today. So thank you for sharing your story. Thank you, Neil. My pleasure. So here we are at credit karma. We run about 60, more than 60 billion predictions a day. I think hearing my guest story today shows just how much thought, culture and engineering muscle all goes into making financial guidance feel effortless for over 140 million members.

[00:41:15] And whether it be unifying tech stacks or running more than 60 billion AI predictions a day, I think it's a masterclass for me in balancing innovation with trust. So please, if you've enjoyed today's conversation, share it with someone who's curious about where fintech is heading next. Maybe someone who would benefit from some of the lessons that we've learned today. And ultimately, talk to me.

[00:41:41] Tell me if you have that data platform and AI power like they have at credit karma, what's the first financial challenge that you would solve? As always, tech blog writer outlook.com, LinkedIn X, Instagram, just at Neil C. Remember to check out Tech Talks Network. We've got eight shows over there. We've got some pretty big names and so many fantastic insights. So check that out and I'll be back again tomorrow with another guest. Bye for now.