Ep 37 | 5 Product Traps Killing AI Projects (and What To Do Instead w/ Kurt Yang)

Build Real AI Products (Fast): Product Manager to Community Builder with Kurt Yang (Fintech & EdTech, RAG, Embedded Finance)

From banker to PM to community catalyst, Kurt Yang shares how non-engineers are shipping functional AI prototypes, validating with customers, and turning meetups into massive and strong ecosystems.

Chapters:
4:17 – Community lessons for PMs
8:34 – De-risking AI with stakeholders
12:50 – Tooling spotlight: Lovable in practice
17:07 – Prototyping workflows that scale
21:24 – Embedded finance & risk-based pricing
25:41 – RAG in fintech/edtech (what works)
29:57 – Stack, Supabase, and next steps

What you’ll learn

From frustration to community: how Kurt spun up GenAI for Fintech & EdTech and grew it to ~800 members.

Lean PM, real signals: why working, functional prototypes beat pretty mockups—and how to run smoke tests with real users.

Tooling that compounds: using Lovable + Supabase (+ ChatGPT) to ship usable prototypes you can measure.

RAG, demystified: practical walkthroughs and when retrieval-augmented generation actually reduces hallucinations.

Fintech shift: embedded finance + dynamic risk-based pricing and alternative credit scoring to hyper-personalised offers.

Enterprise adoption: winning over senior stakeholders in regulated industries (trust, governance, records).

Founder traps to avoid: solution-first bias, over-scoping; how to pick a lower-risk wedge to earn trust and revenue.

Mindset: “Don’t just think, start.” Momentum over perfection.

About Digital Nexus

A founder-led podcast where Australia’s AI builders go beyond the hype with real workflows, decisions, and lessons from shipping. Hosted by Mark Monfort & Chris Sinclair.

Transcript

Speaker: 0:00

Within two days. He actually built an entire functional sort of mobile banking prototype within two days. We can build proper functional And like you said earlier, prototypes that we can actually tests with real customers, so they can actually based on Biggest advice to everybody is just to start not talk about it, not start. Think about it. Start. Welcome to the Digital Nexus In this series, we're going to be getting down and dirty with leaders and founders in the AI space. We want to get an understanding they're building businesses strategies and future impacts of Welcome back, folks, to another This time we've got a special This is Kurt Yang. Kurt, thanks for joining us. Thanks, Mark. Kurt and I have met and also Chris who's behind the camera right now. You would have seen the last front of the screen. We do the Swapsies every now and But Kurt, you're doing some terms of like bringing people and myself that have run know what that's like. So, um, Kurt, if you could just give us a bit of a background for the audience that may not know you? I'm sure a lot of people do, but if they haven't heard of you yet, what is your background like? What's the stuff that's led to For sure. Uh, yeah. It's a great pleasure to be invited to this, um, fantastic venue. Um, yeah, it's a bit about me. Uh, my name is Kurt Young. Uh, I'm a digital product doing product management for the various industries across like enterprise, a little bit like And so I do have a lot of, uh, product management experience, and I'm really passionate about using that to help startups, entrepreneurs to succeed, especially help them to, you know, like sort of build their next innovation cheaper, quicker, get to success much faster. Um, I think my how I found my passion to this new adventure, this made up thing I've been doing is, was actually through my frustration of I just couldn't find a meetup or a community to, for me as a fintech, PM, whatever, to learn about Jai or even trying to trying to transition into AI product management. So hence why my friend was telling me as a joke, hey, why don't you just create one yourself? Yeah. And initially I took it as a And that evening, I think, I don't know what got through to me. I just said, you know what? Hey, what the heck? I'm just going to pay a meetup You know, there was a special and I paid the fee. And then that's the moment. That's the moment I realized, I better just to find a way to So it took me a little while to Yeah. But I think it's like a I'm sure you've been running get in, get into that bit of a kept going and going. So I've been running this for Yep. So starting from from scratch now we got about nearly eight hundred members. I held about fifteen events both I mean, some online, some virtual, some online, some in person. Um, it's just been fantastic, I couldn't tell you the amount personally, I have gained and actually even trying to do some well, just out of like Yeah, I think that's great and But you said VMware, when were Uh, that was twenty twenty two, About two years I was there for. Yeah. Oh, okay. More recently. Yeah, yeah. Okay. I won a competition through They had something called the IT So I'm dating myself here. Um, but in any case, a lot of, And then you mentioned, out of frustration, like setting up the, you know, the meet up group, the gen AI and fintech and edtech. And we'll get to that, setting you weren't seeing something in what you wanted. And we did the same thing. Whether it was, um, we joined the Data Science and AI Association and became president there. Um, but we also created the Oz DeFi Association because out of that frustration of not having a place to chat in between other meetups. And I think it's interesting the frustration of, hey, why know, at night? And then Edison invents the So I am comparing you to Edison here and stuff, but could you tell us a bit more about what Gen AI and fintech and edtech is all about? Some folks, they may have stuff, and that's great, and But what what to you what what about for sure. Um, to put it very simply, my a it's a community. You know, right now it's online But the point is that we have this community, all the community members, we can come together. We can learn, share all the you can call the latest or the most useful AI or AI in general as well. Use cases, um, even like or even like we actually had one perspective, you know, like how AI career perspective, right? From all different aspects. So that's kind of the but like I fintech financial services. That's my background. But later on we're actually know, broaden it up to including because I personally have a lot background covered that as well. And also we have a lot of members saying to me, hey, in the education space, there's a lot of applications or applications happening. Why don't we just try to see if we can get some really amazing guest speakers to cover those topics? So I'm also going to have some sessions coming up as well. Great. And speaking of, um, you know, there's so many different kind of pathways we could go into, but I'd like to set the scene a little bit. Like your background we were discussing before is in project management, actually. Product management, there's a I can go deeper. Yes. No. That's true. But I said the wrong p word Apologies to my project and I've probably equally pissed another p word. But um, in in terms of the that the background? You said you were at the banks and you were at some other like, you know, fintech kind of places before. Has it always been? No it hasn't. Okay. That's a really good question. I mean, that probably can go on for like another ten minutes or whatever, but I'll try to keep it short. Sure. Um, yeah. I actually started my career in banking as a commercial guy or relationship slash commercial guy. So I was a banker. I was a stockbroker, sort of online stockbroker for a little while. Then I transitioned. I transitioned into more like a commercial contract manager relation, institutional B2B, kind of like a customer success manager. You know, a lot of the today's those, um, experiences, I got I got exposure to some you can call project management or product sort of related work as well. That's how I got like really But I didn't even know that was Yeah. Yeah. So I landed an opportunity by accident within the bank, as I say. And that's that's when I think my skill sets really can And, uh, and then from that point on, I've just been product management ever since, you know, at ups and scale ups and enterprise. Yeah. And it's interesting. And I'm asking all this because AI is not something that a we haven't we have seen that it doesn't work. Well if you don't bring in experience the context per se for sure. And everyone has the context of the expertise that they bring in. Sure. AI can help level up things. If you don't know much about you information that you can ask far faster, but it works so much background of some sort of edtech, from, you know, product And I say all of this because I interesting perspectives their experience with what's Now, in terms of that, like what mentioned before that there's having these events. Could you could you mention and, you know, framing for the audience. You're this product manager You're running these meetups and been some of the top of mind just running these events? For sure, for sure. Um, I think the biggest learning or, you know, you can call like a pivotal moment or something is just realizing that anyone can do it. Anyone can create their own. Might not be like AI startup or really get into this space. And the reason I'm saying this, especially speaking from someone, not from a technical AI or even tech or data, sort of a background. I was a commercial guy before I I got a bit of a technical like sort of stuff throughout the years of doing this stuff, right? But even that's the case. Um, what I realized is that, like some of my amazing past guest speakers, they are not engineers or data or whatever people, right? So they come from, say, CFO or of the person. So he was a former owner of a I think it was a tire reseller or tire expert, sort of a company. So people come from all, all what they like. What I saw in common among all passion, their passion to solve Right. Because they were frustrated with whatever the way they were solving before. And all of a sudden they solving those problems. Number two is what I felt is So they they all found their own Maybe that niche is because of their experience or their background or, you know, maybe their unique sort of market understanding. And they bring that to the I think you probably brought up like initially a little bit as well. Right. And then they really go hard at And then that's how they I can succeeding in what they're doing I think it's really interesting. It reminds me of, um, I can't the scenario is like Steve Jobs, He he learnt all about calligraphy and style and stuff like that. And that experience coupled with wasn't he was the tech guy. But coupled with his experience thinking that's helped shape And for many people, it's the Like before, it was like, okay, I only knew education and maybe But now everyone has this opportunity because the barriers are lowered, like you said, like if people are passionate about their space and they can bring in a second skill set, which is like the AI side of things, it's a dangerous like they become a dangerous weapon and in a good way, as in, like they can tackle opportunities that they wouldn't have. And passion is a big thing as I mean, Mia, I don't know if you I'm a very passionate person Yeah. And actually talking about what Right. There's one thing I've been because I'm also coaching some startup founders on the side as well. One thing I do kind of basically the pattern, I've been their past expertise, their their heavy focus around the So there's a lot more, um, focus around the solution, not the problem. And even with the problem, because they have their preconceived notions, but their previous understanding of the pain points. So sometimes there's a lack of or maybe a bit less focused around really understanding the problem and also validating a lot of the assumptions within the problem they're trying to solve. And that can be dangerous in a lot of those sort of AI startups or teams. That's something as a PM, I know, just do that validation space a bit better. Because in that way they can potential, you know, failures or Okay. So that's really interesting because I think a lot of founders normally just without AI, even with AI, maybe it's accelerated that problem of just going out and trying to solve what they think is the problem without actually validating the problem. And it's kind of backwards. It's like, well, I have to create a solution to take it out there. And we hear this a lot, but it sounds like it still happens today. Maybe there's like a this is the people like yourself can help insights sooner, because they months, years, even, without kind of problem. Can you give us an example, scenario where, you know you've describe that to us, because I this sort of stuff, like how change for sure. Are you talking about the gone down a particular rabbit Right? Yeah, exactly. Let me think of one. I think there's a couple of I really good example. I can give it to you. Sure. Um, there's one example. Um, this is a lady from the. I believe she, um, like, she had an idea around an area she want to focus on, um, which is a this is more in the proptech sort of area. Okay. Yep. And, uh, and she had a very strong belief because she's been in that field for a very long time, um, you know, very experienced. And she had very solid Um, but then when she actually problem, what she soon realized Hey, that problem. It is worth the while solving. It's very big as well. But because of her, you know, early stage startup, hasn't got capital solving that. So she used to get very, you going deep on that one problem. And, um, I think where I came in Um, you know, maybe we've validation of the problem, but really about understanding, hey, like, why why is it capability? I mean, from a resourcing from a, you know, like, sort of solution solving perspective, what are the, you know, all the top problems within your space and which one is actually more suitable for you and your organization at this stage to solve? I think, I think for a lot of that, you can call a bit of a, um, you know, discovery analysis. And we actually found out a much organization, her organization And I said to her, hey, it's not like our eyes not on that ball, you know? I think that eventual problem Yeah. The startups, the journey as you Right? Because sometimes you might start with something really small. But the thing is you're getting Momentum is in terms of getting customers, getting traction, getting revenue. Right. And it's a snowball effect. Once you gain, you know, those people's trust you most likely they're going to allow you to solve their bigger problem as well. So it's going to be a iterative You don't want to go hard on the If you don't have enough like ability or resources to solve it. Yeah. So that's a good example. I think that's interesting because it's like they they need to have those runs on the board before it because it's a trust game. Like there's still this Like we need to see that yes, someone can actually create the things that we um, and we can trust them to solve these like bigger problems. So so that's certainly true. Um, I think that's great because AI is great, but it does make it into areas where potentially you So it really does depend. But do you think that there's on the one side you can call it side, um, what are your thoughts with things and tinker and test? And some of those tests may lead might actually lead to new thought about before. So do you think that there is a hole one hundred percent? Another great example I can give So, um, this is actually a fun, uh, it's not a founder, but he's, um, he's actually was my first guest speaker in my very first, um, first show at the end. I think that's his pronunciation So he's a technologist. So, you know, technologists, people like to tinker around things, but he's always been a software solution architect sort of space. He's he's also fairly new to AI. Um, well, I think what he was days using ChatGPT. That was about a year and a half Right. That's when you know now they lovable and the sort of interest So he was actually using ChatGPT He actually built an entire functional, um, sort of a mobile banking prototype within two days. Wow. So, so the point I was trying to say is that the capability is there, right? I'm not talking about building a full on, you know, sort of a production ready, commercially viable financial products from, from digital banking perspective. But if you have something like to people, you can show to the alignment, getting customer much more, um, what do you say? Uh, much more effective than if you just talk about an idea, because when you talk about an idea to, to to people saying, I want to do this new thing, um, new things or something, sometimes people may not really able to visualize that well or on the same page misunderstand, right? But if you within a very short amount of time, you can get to that. That's really helpful. But more importantly, and because it's actually it's a functional prototype. He's able to get real data from people sort of using feedback as well. Although this is not a, you banking app that doesn't have a know, you can get live data, um, I think that there's, um, vibe sometimes hate in terms of vibe people think that, oh, you're do stuff and you don't have the And I'm getting to, you know, to a, as a product manager in your a front end developer. I've done a lot of stuff with of SQL, a lot of data science I know how HTML works. I know you know, that I've seen doing, you know, some of the CSS kind of stuff and doing front end stuff. But the interesting thing has have that bit of knowledge, even knows how to actually do all the to work with databases, you can is more than just a prototype. It can actually become a product because you knew how to set it up well. And I think that that's the A lot of people think that AI is I think it's great. I think it will get us like to, you know, functional kind of prototypes. But I think we can even take it somewhat what you're doing. What do you think about all Oh, one hundred percent. This is where, you know, I was telling you how passionate I am as a lean product manager, right? So I really want to focus on So that's lean. Lean. The reason being is that in lean product management, we very much emphasis on the importance of actually building something that could be a, you know, maybe a Figma Figma interactive prototype. It could be just a mock up or I mean, that's before the AI like AI age. Now we got tools like Replit know, ChatGPT as well. We can build proper functional And like you said earlier, they we can actually do smoke tests real customers, right? Yeah. And as a PM, that's actually even more powerful than you having a just interactive prototype. Yeah. Getting feedback from customers. What if you're going to use it? What do you think real customers using it because you're not actually selling that to customers. What you want to do is you want to test out there is that product somebody's actually going to be using on a regular basis. Once you get the real usage proof comparing to just getting whether they're going to buy You actually get to see the Like you've got real data points And the key difference here is People are saying, hey, it's to be able to sell. That might be true. But but here's the kicker, Once you've done the validation using that, getting that true confidence, knowing this is actually the one this is going to work, then you can you can in today's world, you can build it properly. I guess you can actually hire a small team, really build it out front end, back end, but only once you actually done the validation. So you do that stuff beforehand just by yourself using those tools. Launch the market. It's all about what I call get where you can have a very high is something that customers And then you can build it up So there's no there's no I think I think that's great. And you know, the things that Like you might initially build something with a bolt, a lovable one of those kind of tools and where you've got some, you know, for example, a chat bot or something else that's got AI in it and put in your OpenAI API key, for example, just in the front end code, and you don't realize you think it's like safe and stuff. And it's not until you get an your key is being compromised, going to work and stuff. So then you think about it and then you can even if you if you are passionate and you know what to look for and stuff, and you can ask the questions of AI, you can see that, oh, well, the only way solution is that I have to have people log in to super base, to Azure, whatever it is, and I can put the OpenAI key in a function that's sitting behind the scenes. So only logged in users can that is more protected than So the thing is, the problems whatever, that's one type of problem. There's multiple other problems that even if you don't know exactly, you can either talk to your friends and they can tell you. Then you can ask AI, how do I Or you can talk to the AI. But if you're passionate, going back to what you said at the start, you can solve these things. Um, what do you think about like the how do you think AI adoption in fintech and edtech is going to go, given what you're seeing and given the things that you're passionate about, where do you see this space kind of like going over the next couple of months or, you know, it's a couple of months is already a long period. But yeah, what are your thoughts For sure. I'll probably start with fintech this stuff in this space. Yeah. Um, from what I've seen so far, bit of with the context first, on top of the existing sort of infrastructure and everything. So within fintech or financial services in general, there's the rising trends is what we call embedded finance. So there's more. So I'll just explain very briefly what embedded finance is. So think of this way. I mean, I don't know if you like, when was the last time you opened your NRMA app or Qantas app and those apps, and you will see more and more of those companies. They are not financial They are like, you know, they of area context in the airline. A lot of those non-financial companies, if you go into their applications on the web, on mobile, they offer home loans, they offer car loans, they offer, like all these different lending. Or you can call banking or And in most cases, those financial products, obviously they are not financial institutions. So they would have a white label I used to work in before. They are white labeled, sort of the financial institutions or technology partners. They will be the one providing those both from a technology standpoint and also from a financial product perspective to those non-financial companies like Apple. If Apple tomorrow, you know, like sort of yeah, they sell iPhones, right. But they also do like some sort of a, you know, personal loan to you while you're buying an iPhone. They must done that through Right. So that's what we call embedded finance because of the rise of embedded finance. So there's a very increasing, very much increasing trend of providing customer with a what we call hyper personalized sort of embedded finance lending experience. Mhm. For example. So just imagine if you go on to a, maybe a carsales.com like one of those car sales, you want to buy a beautiful sports car right. And you don't want to pay too You want to get along or So right now I mean if you do that, you have to go to the dealership. They they would only have maybe two or three of their preferred lenders. And the whole process is very very old fashioned. It can go very long as well. But in today's world, we're actually more and more of those sort of companies that are using AI to really providing what you call a almost like a dynamically risk based pricing, so they can actually based on your own risk profile. And that actually also tied to scoring capabilities, which is just using your your past credit Right. There's a lot more social data. There's different type of data There's already engine can help them to do that, but then that information can fit to the front end from a risk based pricing perspective. So for example, your offer pricing might be quite different from mine because of our, you know, our risk profile because of all different things, how it should be based on the product as well. Right. Which is how it should be. So hence why I think that's And there's traditionally that to do that. But nowadays there's more and more applications can actually even further feel that you can call this improvement of the front end experience, which is back then backed by this huge back end of capabilities driving. I think that, you know, what you personalization that was harder machine learning and you know, people talk about AI. A lot of laypersons, they think, But like it's a wider spectrum, as we know, and we're going down the line of the neural networks, which is like what is driving GPT. But in terms of like an overall um, this space kind of like usage in finance. And we maybe even touch on edtech there because like, you guys are doing more edtech related stuff, that how do you see these things being, um, adopted? Uh, further, do you think that more people need to be coming to do you convince people that are Because like robodebt was such a Australia and people still feel wasn't generative AI. That was, that was a very But how do you convince the Yeah, that's a that's a really I don't have audiences, but all see there is a need in getting stakeholders in both. This is actually not just for edtech as well, because in education, those industries, um, I mean, generally speaking, a lot of senior leadership, um, they probably have been very successful with the way historically they've been running the organisations and, you know, like whether it's traditional software capabilities and structured and everything. So I think that adoption, both industries are fairly whether it's from a regulatory internal digital ability to So there has been some struggles, I have to say, in both industries. Yeah. Coming back to your point earlier, whether it's from a consulting consultancy perspective, from a workshop perspective, industry sort of awareness perspective. So there's definitely that need of bringing more, both AI experts or even people like ourselves. Actually, you can call community Yeah, basically because like one with you, one of my vision for edtech, it is actually not just If I can build enough momentum actually partner up with some sort of organizations where you really bring them to together. I think once they actually have appreciation of, um, you know, can just trust us or trust this Then I think you will see a lot more adoption or a lot more collaboration among different sort of industry bodies and everything. That's kind of the goal. I actually that's very Yeah. Um, I hope it gets there because been part of that. It's great to run them, but there has to be some strategy to it. Otherwise you can easily burn people don't realize is behind. And you know this well behind the scenes, how much work goes into even running just one event. So I'd love to hear that there's And I think, you know, if more of these groups get connected, we can all be learning from each other. And speaking of the meetup, um, now, obviously this is going out in a few weeks time and this meetup will already have happened, but you just mentioned that the next one coming up for us, we're filming this on the fifth of September, so I'm dating it. But, um, you've got one coming up next week, which is a practical one. Can you tell us about that for Uh, the one next Tuesday night. Uh, I mean, this is Yes. This is about, um, sort of demonstrating, actually a live demonstration of building a rack model. So for those that doesn't know what Rag is retrieval retrieval, retrieval uh augmented um generation. So this is basically ability that stuff as well. It's basically if you only fit that you want from an output combined or confining, then you actually doesn't allow the AI to So that's kind of a it's been it's been around for a little while over the last year or a bit. I think there's definitely increasing adoption for the different organizations, but I just really want to practically telling everyone, and this is not me demonstrating I have an organization. They're very technically They're going to help us to do that demonstration of a particular fintech use case, just showing to people how accessible that we can actually all, you know, sort of do it not just for the really, really technical guys. And there's a lot of you can sort of providers out there. Yeah. This is further lowering the bar organization to apply those set it up yourself. You can use a service now for And, you know, for rag for for folks out there, if you if you just ask a question of the model and it doesn't have the information in there, it's got a higher chance of, say, hallucinating answers. Whereas with rag, it's only going to answer from what's in those documents. But there's still nuance, which the, the meetup. So definitely like I know you will have missed it people on the show, but, um, you can watch the recording later on and watch the recording. Yeah. I was going to say, is it in Uh, this one's online. Okay. So normally we would have the recording after for members only. So make sure you sign up to our Yeah. But in that way you will get the I think that's great. I think there's a lot of from that, and it's great to a people can go through the back what it's looking like there. But, um, do you have any like this, do you have any kind of want to leave them with if in their AI journey? Like, what do you want to have when it comes to AI? For sure. I think my biggest advice to everybody is just to start not talk about it, not start think about it, start action on it today. So why don't they start acting For me personally, I started community and it's just been I learned a lot and everything, but at the same time I am doing something on my own as well on the side. Yeah, so I think really get your You don't have to be a technical do this, this stuff and do a meetup or join my community, feel that you can do, but do not of watch some YouTube videos or really kind of just get into it. That's probably the biggest one. And just just on that note and like final, final question, but like, what kind of let's call them tools. What are what's inspiring you moment in terms of tools to help seeing really cool that you want Yeah, for sure. So the tool I used most frequently, obviously everyone used ChatGPT. Yeah. Um, but I think me as a PM, as a using a lot is lovable. I'm lovable. I use that to for a couple of Um, one is actually just to kind of help me to really build out some initial designs and just sort of from a conceptual, conceptual stuff. But more importantly, I am it a bit of functional because connect with Supabase and I'm actually using that to build we can actually get live data sort of related stuff that I'm Um, but yeah, I think there's a I think the most important thing actually not most important. Just find the one that actually Because for me, I'm not as So I haven't been be using, you know, basil or a few other tools as much. So just really come down to what really go deep and hard with make the most out of it. Yeah. Brilliant. Well, look, we'll we'll put all There's going to be some great, we'll have from going to your back on the show. But, uh, Kurt Young, thank you Thanks, Mark. Really appreciate it. Cheers. Thank you for being part of If you could really help us out would be incredible. It'll help us to get in front of in the AI space, to share their back to you in the future. Thanks again.