As AI makes building technology faster and cheaper than ever, having a better tech stack is no longer a lasting advantage in equipment finance. So where is the real moat now? We sat down with Patricio Pazmiño, Chief Product Officer at Kin Analytics, to talk about defensibility in the age of AI. Pazmiño has spent a decade in equipment finance building credit scoring models and automation tools alongside lenders, and he argues the durable advantages were never really about the technology — they’re about proprietary data, embedded domain expertise, and the trust built over years of industry relationships. Watch or read the full conversation below or listen on Spotify.
What AI Commoditizing Tech Stacks Means in Equipment Finance
Rita Garwood: As a chief product officer, you’re deciding every day what Kin builds next. When you think about AI commoditizing tech stacks, what does that threat actually look like from where you sit?
Patricio Pazmiño: For us, it’s less of a threat and more of an opportunity, because we see our tech stack as a tool among others we have, not as the end product we offer our customers. The end product we offer is business outcomes — that’s our North Star. So AI is actually allowing us to impact those outcomes in a better way and faster. In the era of AI we’re living in, building technology has become very cheap and easy, so deciding what to build is what becomes the relevant part. The way we decide what to build next comes from talking directly to our customers, being close to them, understanding their pains — not just observing them, but actually trying to solve them. That gives us the insights we need. And now AI becomes a tool for us to do that process even better and faster — putting together prototypes, validating hypotheses faster, and getting more valuable products into our customers’ hands faster too.
The Real Moats in Equipment Finance: Proprietary Data, Expertise, and Trust
Garwood: With technology evolving as rapidly as it is, let’s say a competitor could buy the exact same tech stack as their competitor. What’s left that cannot be copied?
Pazmiño: Let me start from the lender’s side. I believe the most important asset for lenders — something very defensible, that cannot be copied easily or at all — is their own data, the data they’ve been collecting and storing for years of being in business. That’s really what’s going to fuel any technology development in the future, whether it’s AI or anything else. I’d also mention embedded expertise. Our industry is quite complex, which creates a lot of very niche expertise, and typically that expertise only lives in the heads of managers or senior people. Being able to embed that expertise into the solutions you bring to the rest of your team is key — the more you can do that, the more defensible your operation becomes. Related to that is your industry presence and the relationships you’ve been cultivating for years. Our industry moves around relationships and human connection, so those relationships, and continuing to strengthen them, are very defensible.
On the vendor side — companies like Kin providing solutions to lenders — I’d say it’s similar, but I’d highlight three things. First, deep vertical integration tied to subject matter expertise: not spreading too thin, but really going deep into an industry — in our case, equipment finance — understanding its complexities, and embedding that knowledge into the products we offer. Second, trust, for the same reason I mentioned before — in this industry, decisions are made based on trust in your partners, so having a strong brand presence in the industry matters. Everything I’ve mentioned is defensible for the future because each of these things takes a lot of time to acquire, and once you have them, they compound over time. That’s what makes them so valuable.
Beyond Automation: Why Credit Decisioning Is the Next Frontier
Garwood: It seems like everybody in lending today is racing to automate the same things. Where do you think that race ends, and who’s left standing when it does?
Pazmiño: I agree — everyone is racing to automate very similar processes, ones that are sort of far away from the key decision points in the lending cycle. Things like getting data into systems, intake, document processing automation, or connecting to relevant data sources for things like KYB processes. Those were the obvious things to automate, and everyone is racing to do that. But where I see very few lenders and vendors focusing — and what I think is the next big thing — is how you streamline or improve the point in the lending cycle where you make the actual decisions, like the credit decision. That’s a lot harder to automate because it takes real subject matter expertise and it’s a high-stakes process. What I believe will happen is that a few years from now, having all these low-stakes processes automated — like document processing — will just be the norm, and the companies that stand out will be the ones who tied that automation into enhancing their decisioning process. They didn’t automate a process just for the sake of automating it — they used the freed-up time to make better decisions. At the end of the day, our industry is all about making better decisions and managing risk. We’re not in the business of automating things. So at Kin, we approach this with a risk-first mentality — whenever we work with a customer on automating part of a process, we always ask: how does this affect your risk decisioning? Is it improving it, or affecting it at all? Otherwise, you might just be automating parts of the process but making the wrong decisions faster, which is worse than making slow decisions.
Embedding Domain Expertise Into the Technology
Garwood: You talk to lenders and credit teams constantly. What’s a capability you’ve built that looks simple from the outside but is genuinely hard to replicate, and why?
Pazmiño: The first thing that comes to mind is the capability of embedding all that expertise I mentioned into the technology you implement in your operation — to the point where you can actually see your expertise and your new technology working hand in hand, complementing each other rather than being two separate, competing approaches. We’ve heard dozens of stories in the industry of failed technology implementations because clients obsessed over the technology and not the problem they were solving. There was no training, no change management, and the companies bringing that technology never took into account the expertise of the users and how they’d been running the business for decades. So the two approaches end up in friction, and it fails, and you don’t get the benefit of the technology you invested in.
For example, part of our offer at Kin is building custom credit scoring models for our customers. When we build these models, we bring a lot of expertise on the data and statistics side to extract value from our customers’ data — but we also put a lot of effort into understanding their current process, their expert-judgment-driven decisioning. We can’t just throw that away — they’ve been doing it for decades, and it works, they’re in business and doing great. So it’s about taking the best of both worlds and combining them into a model that understands the business rather than working against it. That leads to much better and faster adoption once you deploy the solution. The other key thing is building the capability within these teams to adapt to new technologies, which comes with a lot of hand-holding through implementation, training, and change management, so teams actually understand and can leverage the new technology.
Data, Domain Expertise, or Technology: Where the Advantage Really Lives
Garwood: For a lender, is the moat or advantage in the technology itself, or is it in something else — data, domain expertise, relationships, implementation know-how? Where would you put your chips?
Pazmiño: Definitely not in the technology. The tech stack you have as a company can give you an edge, but I don’t think that advantage holds in the long term, especially now with AI, since access to technology is becoming easier and every company will tend toward the same level of sophistication. It might give you a short-term edge, but not a long-term one. In my opinion, the true moats moving forward are, again, proprietary data. Imagine you’re a lender — your competitor could acquire the same tech stack tomorrow and level with you on technology, but they can’t acquire your proprietary data. If you’ve been disciplined and consistent enough to capture and store that data, and made it accessible to your team, and your competitors haven’t done that, they can’t go back in time and do it — that’s just lost to them. Proprietary data is key, and now AI is also allowing us to capture and store unstructured data, like data from documents, that wasn’t available before. I’d recommend both lenders and vendors — especially lenders — really obsess over their proprietary data: capture everything you can, every touch point with a customer, every call, every email, and invest in the infrastructure so that data is stored correctly and can be leveraged in the future.
I’d also mention domain expertise and vertical integration again — this applies more to vendors like Kin. If you’re a vendor in our space, really make the effort to understand how the industry works. That’s been very key for us — we’ve been in the market for ten years working in equipment finance, sitting down with customers, understanding the ins and outs of the industry, and that’s helped us advance our product. If you’re a vendor new to the industry, I’d recommend making that same effort. And then, applying to everyone, whether vendor or lender: trust and distribution. Our industry moves a lot around human connections and relationships. I’ve talked to a lot of customers, and they don’t make buying decisions based on the shiniest product — they make it based on things they can validate, references from people in the industry, how well they know your brand and your work. So bet on that too — on your brand and on generating and protecting that trust. That’s what’s really going to make the difference and be defensible moving forward.
Same Tools, Different Outcomes: Two Real Credit and Automation Examples
Garwood: Can you walk us through a real example — without naming names, if you can’t — where two companies had access to similar tools, but the outcome was completely different because of something underneath the technology?
Pazmiño: A couple of examples come to mind. First, we engaged with a customer who already had a credit scoring model up and running, built on genuinely impressive technology — you’d feed it all the data you had and in a matter of weeks you had a validated, very predictive model. We reviewed the metrics and they were amazing. But in practice, it wasn’t being used — the credit team didn’t trust the decisions the model was making. We revamped the model through our approach, which puts a lot of effort into understanding the customer’s expertise, and found that a lot of the variables in the model didn’t make sense from a business perspective — some weren’t actually available at the time the model would run, they only became available afterward, and some were unclean or biased. That’s a good example of where the outcome was different because one approach really took into account the nuances of the business.
The other example: another company we worked with already had a document processing automation tool, but it was breaking on 60% of the applications processed through it. The reason was that this company received applications in hundreds of different formats — the tool worked well on standard formats, but broke on edge cases. What we learned, not just in this engagement but in general, is that our industry moves a lot around trust — a tool can work nine out of ten times, but if it breaks even once, you lose trust, and the team goes back to doing things manually. So what we helped them implement was a tool that took the business into account instead of forcing 100% automation. Applications eligible for full automation were fully automated, but whenever the tool encountered a new case or an outlier, instead of a wrong automated decision, we built in partial automation — the AI processes the document but alerts a human, who reviews it and makes any corrections. That worked a lot better. Both examples speak to the same insight: go deep into the business, really understand it, and make sure the technology adapts to the business, not the other way around.
Does Cheaper AI Software Make Defensibility Harder?
Garwood: AI lowers the cost of building good-enough software. Does that make defensibility harder for everyone, or does it help companies like Kin that have deeper domain expertise succeed more?
Pazmiño: For companies that placed their bets only on their software and technology, this definitely makes defensibility harder. But for a vendor like Kin, who’s put in the effort to understand the industry and work alongside customers, embedding that expertise into our products, this becomes an enhancer to our offer. Right now, the cost of building technology is cheap, but deciding what to build is the key — companies that have placed their bet on domain expertise are able to make better decisions faster, which lets them move faster than competitors betting only on the technology side. There’s also the idea of replicability — getting a platform or tool up and running is a lot easier now with AI, so what used to be a competitive advantage for some companies is becoming a commodity.
How Lenders Should Build Their Moat Over the Next 3–5 Years
Garwood: Let’s say you’re advising a lender’s leadership team on how to think about their own moat over the next three to five years. What should they be protecting or investing in?
Pazmiño: Number one, I’d say data — really invest and protect it, make every effort to ensure it’s captured correctly and available for the future. Any technology you want to deploy in the future is only going to be as good as the data you have. Second, when you’re looking for partners to bring AI and technology into your operations, look for partners who already understand your industry, your niche, your business — or at least have the genuine intention to. I’d move away from vendors that are spread too thin across many verticals and aren’t going to make the commitment to advance their offer to what you’re actually going to need. The other thing I’d recommend is making sure that whatever technology implementation project you’re considering starts with understanding what business outcome you’re expecting. Don’t let decisions be driven just by wanting AI or a specific tool — first understand the problem you’re solving and the outcome you expect, and hold your partners, or even your internal team, accountable for that outcome.
The Defensibility Belief About Equipment Finance That AI Has Already Proven Wrong
Garwood: What’s a belief you had about defensibility in this industry two years ago that AI has already proven wrong?
Pazmiño: It’ll definitely have to do with the technological aspect of things — the tech stack, how advanced and sophisticated your technology is. How advanced is your proprietary technology, is it ten times better than your competitors’ — that was probably the number one moat, at least within the tech industry among vendors: owning your own tech stack and having a technologically superior product. That’s what everyone was trying to do, and that no longer holds, by any means, right now.
One Question to Bring Back to Your Leadership Team
Garwood: Last question. If a listener takes away one question to bring back to their own leadership team after this conversation, what should it be?
Pazmiño: I’d recommend asking this: if a competitor — whether you’re a lender or a vendor — bought our exact tech stack tomorrow, what would actually be left that they couldn’t copy from us? And if the honest answer is “not much,” that’s where you have to start working. As an extra one, if you’re a company that’s already invested in AI or technology, or in the process of implementing something: ask whether you know exactly what business outcome you’re trying to move with that implementation. If you can’t answer that with exact numbers — “I’m trying to reduce this metric by X percent” — that’s something you need to go back and think about. Go back to the problem before you keep investing in the technology.
Garwood: Well, Patricio, thank you so much for being on the podcast today. I enjoyed our conversation — I learned a lot. So thank you.
Pazmiño: Thank you, Rita. See you soon.
