Building AI Castles: 3 Takeaways from Customer Stories in AI Product Development

What does it mean to build AI products in 2025? What conversations are being had behind closed doors? Let's talk about it-

In this episode of Your AI Roadmap, Dr. Joan Palmiter Bajorek discusses the evolving landscape of AI product development, emphasizing the importance of integrating AI into workflows, the necessity of a solid data infrastructure, and the potential for aligned financial incentives between developers and customers. 

She shares insights from her experiences with customers, highlighting the need for collaboration and adaptability in the face of changing technology. Joan also touches on the future of AI development, advocating for a flexible approach to building products that can evolve with market demands. Modularity! Don't fall in love with only one model!

Want to talk about your project? Contact us at hello@hireclarity.ai

See customer testimonials and more: HireClarity.ai

Takeaways

• 🧠 Intelligence: AI integration is crucial in product development
• 📊 Data: Quality data can be more valuable than the product itself
• 💰 Cost: Using AI tools can significantly reduce project costs
• 🔄 Adaptation: Continuous adaptation is necessary in AI development
• 🤝 Collaboration: Working with customers enhances project outcomes

Chapters

• 0:00 Introduction and Overview of AI Projects
• 2:56 The Importance of AI in Product Development
• 5:52 Data Infrastructure: The Foundation of AI
• 9:12 Finances and Aligned Incentives in AI Development
• 12:08 The Future of AI Development: Flexibility and Adaptation

Transcript

0:01

Hey folks, welcome back to another episode of Your AI Roadmap. I am so excited to talk to you today and it's been a minute. I uh know we haven't dropped a podcast episode recently and the reason is because we've been so busy with customer work. I have been delighted to be working on some really cool projects and uh my team is small and we just don't have as much bandwidth as we scale related to the podcast. which currently doesn't bring us any revenue. So I'm glad you're hopefully enjoying this podcast as we scale it across the years. Today, I want to talk to you about the projects I'm building as much as I can share, but also like what I'm seeing from customers and kind of how people on my team are thinking about AI builds, how our customers are thinking about AI today, and kind of AI product development cycles. in the field, in the wild, let's say. So there are three things. I was just talking to my customer today. uh He shared a project bid that he'd gotten and said, hey, this is of interest to me in parts. he, this is very common. So most of my customers come from a certain domain. uh Agriculture, I'm working on a manufacturing project. This new one would be a marketplace opportunity to. to build with the customer, but the customer themselves, most of my stakeholders are CEOs, and most of the CEOs have domain expertise. My agriculture customer, he is a luminary in the agricultural space. Does he know a lot about data and AI infrastructures? No, and that's okay because we work arm in arm together, respecting and building really, really cool things together, like really cool. Anyway, so not everyone has to have AI acumen. But I will say what's really cool about my manufacturing customer, for example, is when he came to me, he's like, Joan, I want to make sure we're using AI as we build AI. And I was like, woohoo, this customer uh sets the stakes high, knows a lot what he wants, but even that question and that mindset is a relatively new thing for me to experience in customers. And so with this new customer that I'm speaking to right now, the bid I looked at, for the development had no AI usage in the build. Now that might've been okay in 2015 for it all to be done manually, but here, as I record this in May 2025, using no AI to build products, I think is bloat. I think it's a bit of a waste. I love when we can be efficient with our tooling, AI and otherwise. and use human time strategically. Yeah, does that make sense? So I'm gonna go through three things that I talk to customers about right now that as if you are on the business development side, I would love for you to understand more about how we think about on the development side. And maybe that you could share this with your stakeholders as they up skill to ask great questions of development groups. And if you are on the build side, I'd love you to think about how we talk across functions. and kind of build your acumen. And if you're just here to learn more about AI, I love that too. We're all upscaling. We're all learning together um in the trenches. So let's get into these three things that I have been learning about and would love to share with you. Number one, as just mentioned, are you using AI to build AI? uh I think this was really obvious, but some people aren't doing it. There's some really cool products and hopefully I'll do another podcast episode about cursor, and some other tools, but just basics basics, you can use Chatch GPT to get code, right? Is it the best code? Okay, remains to be seen. But my team leverages tools that have AI in them or are AI tools for development as we build quality products. Now, what we need to do, so let's say I get this code from whatever tool, I still need customizations. I need some human. to review the code. My CTO makes sure that everything is organized. Our documentation can be read by a human. One day maybe we'll just have AIs, building AIs, but I still think this human oversight to make sure that it matches the goals. Checking it before it goes out the door. It's like checking your homework before handing it in, right? You're like, is this the right assignment? Is this essay for English or is this actually my math homework, right? You wanna make sure that there's that check before it goes out. So if you are a customer or you're planning to build a project, if you're not using AI, I think there's a huge component missing here in 2025. So for example, I recently was conducting a competitive analysis for a customer as we build the stack of what tools we're gonna use as we move forward. It's really, really exciting. I wanna share more, I can't right now. But in the competitive analysis, I did some overall cursory look. And then I also asked three different AI products in this space, what are strong products I would be wise to benchmark against? And actually we found this really cool augmented reality product I'd never heard of before. Maybe they have bad SEO on their website. I don't know why, but I had not heard of them in our overall competitive analysis to look at different tools. And in fact, they hit what my customer is looking for to a T. He reached out to that company's CEO. for a meeting because it could be an excellent fit. So just being a product that's findable, like AI helped me find the right product for our building blocks of this next product phase. Do you see where I'm going here? Like AI is integrated into workflows. It's making me do better project work for my customer. He's getting better results, right? and we're gonna be building, hopefully, I really wanna get that next project scope signed off and could be working with him for years, but this idea that it's integrated, that we use it to augment, to help, to pilot with us for something I may never have found otherwise. And I'm pretty good at finding things. I'll just let you know. I don't think I'm starting from zero over here, but whatever. So that's number one, AI used in building. The second one is that when we're building things, I still want to be thinking about our data infrastructure. So you may have heard me say this on this podcast. I've said on another podcast when I'm interviewed, your data is your foundation for whatever you want to build with AI. You have bad data, you have a bad foundation. Maybe it's not organized. Maybe it's in different data silos. I don't know. But if your data is a cracked foundation, if it's poorly set up, whatever you build will be on shaky ground. I don't like shaky ground. I like a great foundation, solid, fantastic, and we're gonna build castles, yeah? So for example, a recent project I saw, it was really beautiful. Visually, this product, stunning. But behind the scenes on the backend, this was not getting a data lake organized. The data that was being brought into the platform. was not going to be successfully repurposed to build personalization tools, understand more about user behavior. It's like a pretty, object, but like there isn't depth. ah I have a lot of metaphors, but point being just because an AI can build something pretty on the front, if your backend doesn't support a long-term progress, there's also something missing. Does that make sense? Very shallow. It reminds me of people who are gorgeous, but then you talk to them and there's no depth. I love a deep conversation myself. So maybe there are people who just like to be pretty and that's their thing. But as far as AI development, data infrastructure, data is so rich. I've mentioned this on this podcast. I'll mention it here again. I work with a customer that the data sets they're building could be more valuable. than their regular product in a three to five year time span. Let me say that again. Their data could be more valuable than the product they built today that they sell. So data is so valuable, that foundation, that solidified thing. Now, if you're building something dinky, you just need something fast, you need something quick, you don't care about long-term, you don't wanna scale to be a unicorn, you don't wanna be a multimillion dollar company scaling to billion dollars of revenue. Okay, whatever. Build something dinky, something cute. something attractive. I care about huge growth potential, and that means that your backend is organized, it's clean, the data is robust. You are building, like I said, castles. You're building proverbial, what are they called, castles, motes, realms. Okay, I don't need to go all sci-fi, but point being we're building some wonderful things on wonderful foundations. So that's the backend and the data. The third part, that I think about more and more as I bid customers, as customers pay me on time, yes, I love that, uh is the finances. So let's imagine that I'm working with a customer who is not aware of the AI tools that my team is using. I'm just building something, the customer says, great Joan, love this cost, wahoo, we move on. Now, if I'm using AI efficiently in our workflows, my customer knows that I'm using AI broadly, but doesn't necessarily know how Like what if I'm building at a fifth of the cost, I could just keep the rest of profit for myself, right? We build lean, we build efficiently, our profit grows and grows, right? A lot of software companies see revenue and profit, you know, like 60 to 70, 80 % profit. I know companies like this, right? They are profiting off of their customers, which is great. I understand it. That's how we stay afloat. That's how we're in business, right? But I believe in the future, when we're building smart, when we're building together these multi-year relationships of my team being arm in arm with customers, I believe that a smarter model might be around aligned incentives. So something I talked about recently with a customer is, you know, my team could bleed you dry. My team could, you know, make quality products and you don't even know how much money we're making off of you. And frankly, I might not care because if it's just about the dollar signs and you'll sign off on it, like I could just make very, very expensive products. And in fact, one of the investors for my customer said like, one day it'll be dangerous to have our AI team outside of house. We might want to acquire you. We might want to bring that AI acumen, that AI skill set in house. But for the stage of company, that's too much for the stage of company it's at. And I totally get that. And they need to stay in their zone of expertise, right? Of what they're currently doing today as they scale the company. What I believe is part of the future, and one of my customers already talked to me about this, is we do a revenue share. We do equity. Like they want to prepare the company to sell in a three to five year time span instead of bleeding them dry on just the project cost. And I don't mean bleeding them dry. It sounds very nasty. Maximizing my team's revenue. maximizing profit just for me as the person developing the team and the AI products, there is a huge opportunity to say, hey, I'm going to help pass on that efficiency to my customer. Hey, customer, I can build it for way, way cheaper because of the AI tools I'm using that we are literally optimizing your workflows. We are making this excellent product work. You're going to see that in your revenue. We know this company has places it's going to go. We would like a cut of the revenue, we would like equity. So when you sell, we also benefit. So that's what I think about as aligned incentives. Our values are shared because instead of just taking from that profit, that profit is a cut of mine that I'm interested in or my team is interested in. And if they were to have quality foundation, let's say I'm building that castle with my customer, the customer selling the castle, it's both of our shared goals for it to be a really quality project that the data is set up so that when someone is snooping around and saying, do I want to buy this company? Like, you know, do we want to sell to private equity? They know it has a quality back end. The documentation can be readable by a human. Right. Like we have amazing products built that shared alignment of goals is built from day one. The foundation, the way my customer respects me. It's an amazing feeling that I can do amazing work. Really cool projects, impactful for the world, get paid, and I like my customer a lot. Do know what mean? Like, it's an amazing world. So that third piece being the finances, aligned incentives. You know, I see a lot of gig work where you just build a project and you let it go and they pay you for it and that's great. And I see, I don't know, you know, you can also, you know, I'm talking about like robust partnerships. in customer projects. And I've definitely seen people who are like, let's get the cheapest, cheapest version. You know, I don't care the quality. I just want it to be really, really cheap. And the thing is, one of the companies I worked at, I'd say a third to a half of our business was rescue projects. They built it really cheaply. They maybe got it done offshore and they didn't speak the same language. Offshore can be built really well. I'm not saying otherwise, but I'm saying like, if you were looking for the cheapest option, You may get poor quality outputs. You may get bad customer service. um It might bite you in the butt in the long run. And I've had to fix some of those projects and it is a pain to try to fix it if you have a bad foundation. You may have to start from scratch. Anyway, I don't mean to be the negative. I think the positive and the opportunities to work together are far more powerful if you're talking to the right customer. So as I'm giving you these three things, these three topics about AI product development. Let's go back and review them. Using AI in your workflows, in your builds. Number two, not just a pretty front end, not just a pretty product on the outside, but on the inside, your data, your backend, the value that's being brought is there, is solidified, is being built out also. And the third thing being those finances. The aligned incentives that I see is part of the future of amazing AI product builds. That's a lot of the things. The last thing I'd like to say, guess bonus fourth one, would be to think about kind of how modular we build. I was talking to a data company this week and they were talking about they are not married to Azure or Google or like they want to always have flexible components. And similarly for my team, I do some work specifically just benchmarking models. So OpenAI has lots of different models that you can use. for example, and some are the right fit, some are more expensive, some are customized for different purposes. So for that customer, I'm constantly benchmarking, this is what OpenAI says it can do, this is how fast it actually runs. Here's the quality of the outputs, good, better, best. Is it giving us good product? Is it just cheap and not helpful? Like we constantly need to be benchmarking, checking the pieces of the puzzle as we build it. You know, is brick the best thing for today? Do we wanna build with cement? Do we wanna build with algae stuff? I saw a really cool project this week in algae. Anyway, point being, we need to be constantly thinking about adapting, being curious, being open to growth in different ways. So I hope this has helped you think about development in the future and the role of developers, which definitely are still needed, just probably a lot fewer in these type of builds I'm talking about, the way we talk to customers, how we can upskill together. and having excellent questions and growing in our knowledge overall, I think it's just super cool and really fun, which is why I have been doing so much customer work and can't wait to close more so that I can hire more folks on my team and bring you all more podcast episodes. So thank you for your patience with us in this brief pause, but it's been so much fun to do customer work, get paid for really, really cool, fun. projects that I get to talk more about hopefully in the next few months publicly. If you liked this episode, please make sure to download it. My team seems to downloads and it gives us a rejuvenation of all the hard work is paying off. If you'd like, subscribe, leave a review about the type of projects you want to hear more about. I know we have a backlog of guest episodes. They take a long time to edit, so please be patient. I'm so excited that you're here with me learning more about AI and building your own AI roadmap, leading with curiosity and excitement and just learning together about the future of products and projects. Okay, all the best from Seattle and look forward to seeing you on another episode of Your AI Roadmap. Cheers.