Middesk CEO Kyle Mack on AI, Good Data, and Business Onboarding in 2026

What happens when you try to build AI-driven fintech products on top of messy, incomplete, or unreliable data?

In this episode of Fintech Layer Cake, host Reggie Young sits down with Kyle Mack, CEO of Middesk, to unpack why data—not AI—is the real bottleneck holding fintech innovation back. Kyle explains how KYB has evolved far beyond a compliance checkbox, why poor data foundations can amplify risk instead of reducing it, and what it actually takes to support real-time, automated decisioning for business onboarding. The conversation dives into agentic workflows, performance constraints, first-party data strategies, and why trust and explainability still matter in a regulated world. Kyle also reframes business verification as revenue infrastructure, not just risk management, and shares why identity should be thought of as a “rail” connecting businesses, banks, and governments. 

Transcript

Reggie Young: Welcome back to Fintech Layer Cake, where we uncover secret recipes and practical insights from fintech leaders and experts. I'm your host, Reggie Young, Chief of Staff at Lithic. On today's episode, I chat with Kyle Mack, CEO of Middesk. Middesk is a KYB platform that helps banks and fintechs verify business customers. But as we'll get into in the episode, KYB is about a lot more than just verification.

AI is obviously a huge topic in 2026, but in fintech, there's a critical, sometimes often overlooked, foundation you need if you want to leverage AI, and that's data. If you're not careful with your data framework and strategy, your AI tools will just cause and exacerbate problems. I invited Kyle on since I know this problem is something he's been thinking about a lot and working on solutions for.

Fintech Layer Cake is powered by the card-issuing platform Lithic. We provide financial infrastructure that enables teams to build better payments products for consumers and businesses. Nothing in this podcast should be construed as legal or financial advice.

Kyle, welcome to the podcast. Excited for a conversation today. I've been following Middesk for a while. I know y' all are doing a lot of fun things with data and AI, a very important topic in 2026. Maybe for our listeners, the best place to start is maybe just with a quick spiel on what even is Middesk? Why am I here talking to you?

Kyle Mack: Totally. Yeah. And again, I appreciate having me on, Reggie. Middesk is really the leading business identity company. We provide a whole set of infrastructure products that largely banks and fintech companies use to help onboard, verify their customers, usually where those customers are small businesses or it's their commercial accounts. Some real-life examples would be companies like Mercury or Ramp or Cash App use our data and intelligence products to be able to automate KYB.

A very different example is companies like Gusto. We have a set of business registration APIs, and they use those products to offer embedded account setup where businesses can actually go out and register themselves directly with the government, and Middesk is on the back end, kind of the API layer for doing that. So I really think about the business today as our strategy is we're trying to spot all the friction points that exist for two businesses to be able to work together, and then we want to be the API layer that removes as much friction as possible.

Reggie Young: Yeah, I love it. I worked at Bluevine before Lithic. Sweet spot for the SMB market, and there are a lot of quirks which I want to get to in this episode. There are some funny quirks, like KYC is hard when you're dealing with consumers and all the state licenses, but there's a whole different set of problems for KYB, authenticating and setting up accounts for businesses. We'll circle back to that.

Main reason I wanted to get you on the podcast initially is to talk about AI and data, because I know Middesk is thinking a lot about building with AI. I know from our conversations, you've hit on data being a really important thing that a lot of companies don't think about when they're leveraging AI. Why is data such an important thing that builders working with AI need to think about from day one?

Kyle Mack: I think it's really three things. We'll go through them, and then I can do a quick dive. I think about access, solution performance, and then reliability and trust. To just double-click on those things, we'll talk about access. We started Middesk about seven years ago. When we started, the problem we focused on was there was this huge delta between the data that existed out in the world about a business and what you could get through a third-party provider. For us, that was a LexisNexis or a Dun & Bradstreet or something like that. And then the impact to the customer, to the bank, to the fintech was that it became really hard to automate decisions with confidence because if you have a bunch of gaps in the underlying data, it's very difficult to know whether the company that's opening an account with you is actually a fraudulent fake business or whether they just formed their company last week. Those things effectively look the same.

We thought a lot about this access problem in the beginning of Middesk, and today, we manage relationships with about 400 government agencies across the US. We've built this massive data strategy to be able to go out, acquire that information, index it, keep everything up to date. More recently, with agents and LLMs, access is becoming a bit easier. And so you now have tools that can go out and do some of that work. Access is a big challenge. It's becoming easier, which then leads to this next point, which for me is around performance and what you're trying to deliver with the solutions.

Most of our customers today, they're trying to drive real-time automated decisioning at the point of onboarding, which means they're looking for responses to come back from Middesk in less than three seconds usually. And so that becomes really hard if you're going to build a set of agents that at the time of request are going to go out, they're going to go hit the Delaware Secretary of State website, they're going to crawl through the page, because the sites are unreliable and they're slow. And so the performance today, at least, that you're able to get, it's measured in tens of seconds or minutes. And that has an impact on the customer experience. And so we think a lot about the work that we've done has now enabled us to start building agentic products that have access to the tools that we have internally, a lot of our underlying data platform.

The last thing I was just going to mention quickly is this idea of reliability and trust. We work in a world of regulated decisions, and it's important for people to be able to understand sort of data provenance, how information's flowing through systems, explainability to understand where decisions are made. Today, if you make a call to OpenAI, just a quick LLM call, you can get back interesting information, but it's a bit of a black box for where those things are coming from. And for somebody thinking about building with AI, the data that it accesses, there's how are you going to access it? How quickly do you need information back? What types of information and sources are you going to be willing to leverage to make decisions? I think our approach has allowed us to thread the needle a little bit on being able to offer agentic solutions and AI-generated products. They can leverage a lot of the data that we have readily available and solve some of those performance problems.

Reggie Young: I love it. The first one resonates a bunch, the access and inclusion problem of the easier access to financial products, you want to increase access, you want to increase inclusion. But also you're going to let more fraud in if you relax things too far. It's really hard. Yeah, access is a really important thing to get right, and then the reliability thing. It's funny, I think maybe two, three weeks ago, the hallucination risks have gotten, I think, a lot lower and better with time. But even two, three weeks ago, I was chatting with an LLM platform that will not be named and got a verbatim quote from Regulation B, the federal regulation on discrimination and credit that was just completely fabricated. It was an exact citation and it says this with a quote that doesn't exist.

Kyle Mack: I'm sure the lawyer in you is, oh boy. 

Reggie Young: Yeah, you got to still be careful out there, folks. Hallucination risk is still definitely a problem you encounter.

What can go wrong for a company if they're not thinking about this data problem? If they build a great AI tool and they haven't invested in setting their data strategy up correctly, what are the problems they're going to encounter?

Kyle Mack: I think about a few things. One is taking on unnecessary risk, if we're thinking about making decisions, even to your point, around hallucinations. But if you're actually trying to drive to just automated decisioning both to open accounts or to reject customers and those tools are making mistakes or they're using returning data that actually is for a different business than is actually the customer you're trying to onboard, companies can be called the same thing. That's definitely one challenge, and that can lead to fraud risk, credit risks, just a bad customer experience as well.

And then the second thing was, again, the industry we're in and the customers, I do feel we're still at the very onset of the adoption curve for really integrating these products into workflows in a more meaningful way. As I think about on one hand, there's a top-down push to be able to drive efficiency, drive costs down, and explore these tools, but given the nature of the way these products are being used, I think there's a risk that if the tools are making mistakes from day one, it gives naysayers maybe the tools that they need to find ways to slow the adoption of some of the stuff out, slow it down, or limit use cases.

It's getting to the point where the utility of these products can be very significant, but the organization needs to have trust that they're going to work, they're going to be explainable, they're going to do the things that they say they're going to do. I know we keep coming back to it, but a lot of that for me still gets back to, do we know where information's coming from, when it was updated, how that data is being used, which ultimately is at its core a data problem.

Reggie Young: Yeah. I took a good writing seminar in law school, incredible professor who's very well regarded. It was an award-winning constitutional blog. At one point, we did this exercise where he was like, okay, here's the thing that you're going to read, and then you have to go verify all the sources. And the exercise was realizing, even when you have human editors in these well-reviewed law journals, you look at the primary sources, and the humans get stuff massively wrong. And so I love the LLM problem now of like, what are the actual sources, and is the LLM reflecting them? Are the actual sources reliable? You got to look at the whole chain.

What are some of your top learnings from building products with AI so far at Middesk?

Kyle Mack: A couple things. One that I've been thinking a lot about right now is what the experience for leveraging these products should look like and how to really integrate these products into our existing solution. Every week, right now, our customer base, it's about 600 customers, we're helping hundreds of thousands of businesses get access to different types of financial products. I'm thinking a lot about how can I make these more agentic solutions as integrated as possible and almost feel invisible in our product.

The way we're approaching this, we'll see if it's the right strategy or not, is when Middesk flags anything, any sort of risk factor, we want to be able to go out automatically, investigate that. We want to present information that will allow you to have the complete picture. And I want that experience to almost feel like magic. If we can think about the UI and how the flow between triggers and research and the workflows on the back end work, then I think it allows companies to leverage the existing distribution that they have and drive usage of these products super quickly. That's the big thing for me. I go back to the ultimate goal for Middesk. We want to make it easier for businesses to access the products that they need to start and run their company.

Today, about 10% of the companies that are formed in the United States, Middesk will help to verify in about three months after the companies are formed. When I think about how much inefficiency there still is for a business to go from formation to payments and banking, the ability to have our customers adopt these more agentic tools is really important because the reach is starting to become more significant. That's definitely one, what's the experience look like? How do we integrate it with products that people are already using for us to be able to drive adoption, and then that allows us to scale impacts more quickly?

The second learning is just how important the context is. It's easy to make an LLM call and just say, return me information about Middesk, and you're going to get this kind of profile based on our website and LinkedIn maybe. It's a lot more difficult, though, to be able to coach and lay out the chain of thought for hyperspecialized workflows around risk. Maybe an example would be within lending. I have a borrower who's coming, and I need to trace back the ownership structure of this company through nth degree, nth hops of ownership, because I'm trying to identify any additional guarantors that need to be listed on a loan agreement. And then for those guarantors, I want to be able to understand who has the authority to sign on behalf of the business.

Doing like that is not a quick search. It's many different tools based on the things that we find. We need to be able to invoke other sorts of actions. I think the context then is super important for how do people think through and reason through a problem like that, and then how do we think about all the branches, the edge cases where things are going to get stuck. Those have been my two big things, the experience itself and how to make it integrated and then just how important context is and how much knowledge you have by just doing a bunch of stuff manually really.

Reggie Young: Yeah. I love that. The 10% reach is awesome. 10% of SMBs is a massive chunk of the US economy. So that's awesome that you all are touching that. And your point about agentic workflows, there's a lot of headlines six months ago, like agentic payments, they're increasingly becoming a thing. Still super, super early, but really it's about the agentic workflows. It's sufficiently advanced. Technology is indistinguishable from magic line. I hadn't thought about it until your comment just now about agentic workflows are that sort of, oh, we can make magic happen. In a perfect product experience, it all happens behind the scenes. You have no idea it's happening. Agentic workflows are a great opportunity for that.

Kyle Mack: Totally. Yeah. 

Reggie Young: I love the context point, too. It's funny we talk about, depending on your pronunciation preference, either riches in the niche or riches in the niches. The very specific unique cases, if you can solve those, people care about those, they're hard to solve, and so if you can figure out the hard things and they become incredibly lucrative as a business to solve, but also hard because there's a lot of unique context. That was fun to get the escalations. Oh, this is a trust arrangement. Who do we verify? Okay, we got to figure out state law.  It's fun. They're fun problems as a lawyer, but also it's very hard, you do the context for those arrangements.

Kyle Mack: It's interesting, when you think about the strategy of a company like Middesk where it's a usage-based business, we generally want to skew toward use cases that have scale. Generally, where there's scale is where you have more simple things. If I'm a Cash App, I'm onboarding primarily sole props and single-member LLCs, and we can talk about the complexity of some of those things in a second, but that's where you see high volume. But by nature, single-member LLCs are fairly straightforward. But then you swing to the far other end of the spectrum and you use trusts as an example. The use cases of onboarding trusts are generally lower volume, but the amount of overhead going into them is significantly different. So the economics of supporting those use cases are very different.

Historically, the upside of opening an account for a trust, or maybe- I don't know, take another example, originating on commercial mortgage, $100 million commercial mortgage, the upside for the bank of closing that transaction meant that it was okay to have a lot of cost associated with it. And now there is ability to start to think about how to make those transactions even more lucrative because the depth that we can go to in the research can start to replicate what that person is doing, which I think then the benefit to the bank or to the lender, those transactions can actually start to become even more profitable and more efficient for the business who is waiting on the money or the account or name your financial product.

Reggie Young: Yeah. What about the concept of first-party data? I know that comes up a bit nowadays for Middesk. What is that concept, and how does it fit into this AI and data discussion?

Kyle Mack: We thought from day one about the business. If our strategy is going to be go out and acquire a bunch of publicly available data, there's some delta that can be gained in the beginning. That's not a forever thing. We've tried to place several bets. I talked earlier about the use case of today Middesk actually going out and registering companies with the government, and that's an example of that.

We're coming up also on about 10%. We started on this use case of, if somebody hires an employee in a new state, they need to go out and set up withholdings and unemployment insurance accounts. Some payroll companies actually don't allow you to process payroll to the accounts or set up. The process of going through and dealing with the government to be able to set these accounts up, it's brutal. And so we built the first APIs for doing the creation of state-level tax accounts. We sell that. Like I mentioned, Gusto, it's through Rippling, Deel, and Paychex, and others.

The benefit on the back end is that as we move into this position to almost manage the read and write of government data, the states require UBO information to be provided, entity structure data. And so it's allowed us to build a perspective on these companies that we wouldn't be able to get by just going and setting a relationship with the state and pulling their data down. That's an example.

We're also coming up on roughly that number. I think it's a little closer to 5% of all the companies registering with state-level tax agencies are being done by Middesk today, which still means there's 95% to go, but we're getting there. It's at least whole numbers now. That's an example, and that's super important for our customers because some of the hardest questions to answer around onboarding businesses is data that is not publicly available. So it requires an additional strategy.

We also do other things. Our data strategy has been to go out and canvas every company today in the country, sometime in the world, and build this massive entity graph of all the businesses that are registered in the country. We try to keep it up to date effectively every day. We're talking about agents and AI. Interesting things that we're starting to look at now is we have these internal, I'll call them tools, internal tools that we build our products on top of, but we may not make directly available to our customers, that we can give access to some of these agents that can start to do some of these more nuanced tasks, like we talked about tracing chain of ownership back and finding subsidiary structures. Our entity graph is great for that.

We're starting to just think about, across the network we have, given the volume and scale that we start to see, then how do we use that to derive some signal that ultimately will help our clients remove friction, make a decision faster, get their customer up and running. Hopefully, for Middesk, that means more defensibility of our business over time, more moats, more compounding value than just go get the Secretary of State data.

Reggie Young: Yep. In fintech, folks hear phrases like business verification or KYB, and they think, oh, this is a boring compliance box I need to check. It just means I need to do a simple verification or whatever. We were chatting before recording. You think about it in a little bit of different light. I think it's an important reframe. Would love to hear, how do you view that concept of like, is this just a compliance box that you need to check? What does business verification mean to you?

Kyle Mack: It is a compliance box to check. My pitch is, make revenue and growth and risk and compliance just two sides of the same coin. It is especially true if your end customer is a business, because onboarding businesses is extremely inefficient. Even with the work that we've done and the new tools, it is not uncommon to be having half of the businesses that apply for accounts with you getting some form of document review or deeper dive investigation. And it is not uncommon that when you send things to doc review, as many as half of the applications that go to doc review never reply. So a checkbox, how inefficient it is, actually is a huge customer conversion problem.

Today, we have two of the three largest banks in the US are leveraging Middesk for opening up small business bank accounts. When we launched with one of them, their own auto approval rate was 10%. The information requests and doc request rate was 90%. And of that, up to half of the businesses never replied. So the funnel conversion was crazy, which messes up all of your CAC to LTV. That is the reality of what the problem space is. And our goal, I keep coming back to it, we want to remove as much friction as possible because it has a real impact to the bottom line. That's at the point of onboarding. And then we can also talk about making better decisions hopefully should lead to less losses, whether that's credit loss, fraud loss. Having more reliable data will help you inform those models. I think it's both driving acquisition as well as the bottom line through losses.

Also, the space is boring, but it's crazy. I'm like a nerd for these just super-boring, unsexy businesses. I think it's super interesting. Before Middesk, I was at Checkr. We were doing pre-employment background checks. talk about something boring. But it's just this super-messy data problem. It's like picks and shovels style business, and it's always changing. There's so much depth to the problem space. That's what keeps me excited about it seven years in and for the next decade.

Reggie Young: That's funny. I find Lithic and our issuer processing stuff fascinating, but no reasonable person knows what an issuer processor is. Basically, as far as my parents, or where I work at a company called Privacy.com, because that's actually a card product they can understand, they don't really know what Lithic is. But it's not necessarily a- we are doing fun stuff with stablecoins and agentic stuff. But yeah, there's a lot to be said for businesses that don't seem like seductive businesses necessarily, a lot of value to be had.

I love your point, too, on the 90% of the bank's apps being sent to doc review. That trickles through the entire business and the entire strategy. A bank is probably looking at that and being like, this just isn't an effective product. In reality, it's just their tooling needs to be fixed, and then it changes how they want to invest, how they want to attack the market. Yeah, it's interesting, like compliance and risk. To your point, business verification is a compliance checkbox that you do need to check, but it also informs your product strategy and the broader company strategy if you don't do it right.

Kyle Mack: No doubt. And doing it well, it can create really elegant, unique customer experience. It is also true that when you have to go back to a customer who wants to work with your product and you place the obligation on them to prove that they're legitimate, worthwhile to spend the time to onboard, that is just bad customer experience. So yeah, it's customer experience, growth, conversion and also a checkbox that you have to do for compliance.

Reggie Young: I know we chatted about a time to loan. You mentioned this at the beginning. Sometimes the verifications can take a few minutes in businesses if you're doing it the old school way. I think about working at Bluevine, it's like those SMBs that applied for loans, they would go out and they'd try multiple offers, and whichever one comes in first, they're probably going to take. And so that time, that minute difference of the confirmation email can be whether or not you start building a relationship with that customer. So yeah, it's a compliance checkbox that generates revenue and converts SMB customers.

Kyle Mack: Totally. No doubt.

Reggie Young: From what I've heard, SMBs and sole props are being formed generally at higher rates nowadays, but their verification coverage isn't that good. Tell me about that. This is what I was alluding to at the intro. I think this is a super interesting problem. What kind of coverage do sole props and SMBs have that makes it so difficult for them?

Kyle Mack: Okay, here's one of my other fun headline stats. Did you know that there are more companies formed in the US every month right now as there are people born?

Reggie Young: Huh. Interesting.

Kyle Mack: That's your learning.

Reggie Young: I like this. Most guests don't come on with the data at their fingertips like you have. So this is great. I love it.

Kyle Mack: That's our business, right? And we've been really trying to dig into this stuff. Every month, right now, there's about half a million companies that are formed. There's about 300,000 people that are born. Not all of those companies go on to hire people. If you have an Airbnb, you could set up an LLC to own the Airbnb, and that is counted. But that LLC still has a bank account that they need, and Airbnb does payouts to the bank account. So yeah, to your point, there's an increasing number. I think it's just going to keep going with Claude Code and all these other products that allow people to vibe code their way to something that works for the most part. Will work in the future for sure.

And then another interesting one, of SMBs out there, 80% of them are owner-operated with no employees, and 30% don't even have a website. That doesn't mean they're bad companies. So this kind of takes you to your question about, why is this hard? When you take the leap to go out and form an LLC, or you're going to do a start-up and you need to get funding, you go to Delaware or Texas, you set up a Delaware C-corp, at that point, there is a footprint of data available about you to pull from. For sole props, though, there's really no obligation to have to go out. Some will set up a DBA maybe. You might have a website, but if it makes sense for your business. And so there's just a data availability problem, and then evaluating risk is just very contextual because- I don't know, make up some examples.

So you have Kyle who starts a greeting card company, and I'm going to sell greeting cards on Faire. Okay, so I might have a website, but I'm not going to have a Google Places page for my greeting card company because I just do it from my house. But that's not a bad thing. And then I compare that maybe to Reggie, you start a coffee company that sells coffee at a farmer's market. You might have a Google Places page, but maybe you have a website, and then you have a restaurant who should have all these things. And they employ people, so they have an LLC, and then they also have a liquor license. The context and how nuanced the shape of businesses are and can be is so dependent on the industry they operate in, like their size and maturity. And so that's what makes this hard.

As another great application for LLMs, for AI tooling to be able to take in that context and be able to synthesize all these different data points down and be able to provide this kind of overlay, that would be really difficult to do in more of a rules-based world because there's so many versions and combinations of those things. And for us, the way we think about it is if you're just talking core regulatory compliance, what is nice is if somebody's operating as a sole prop, the general guidance is you can KYC the individual of the sole proprietorship, and you can check the box we were talking about earlier. At the point somebody's gone through and set up their LLC, now you need to do a more proper KYB. You verify who's the entity itself.

But that is just checking the compliance box. To actually evaluate, is this sole proprietor a real business that actually operates in the world and what are they doing for their business, that is not KYB from a compliance standpoint. It touches more on risk management and fraud. We try to help our customers. If they're solving a compliance problem, let us help you get great at cohorting and segmentation of your customer base into registered and unregistered businesses so you can apply the right policies and procedures on top of them. From a risk management standpoint, then let's look at the other data points and intelligence that we can layer on top of that to then help you evaluate risk appropriately.

Reggie Young: I love it. Yeah, there are a lot of SMB providers out there that just won't touch sole props because it's a hard problem. And one of the- I guess call it fringe use case but becoming more common is think of a TikTok influencer. Arguably their money is sole prop money. They may have an entity, they may not. They also might be 15, and so how are they getting a bank account their parents are signing off on and they have no revenue history. They definitely have an online presence, but how do you do the risk analysis on a TikTok profile versus here's a blue-collar business that has a website in their local community. It's fun fringe cases.

Kyle Mack: You talked about the case of the earliest, most fringe unstructured company. Even as of six months ago, and I should probably know the answer to this today, but even if you go to the website of the largest banks in the country and you go to open a bank account through the digital channel, as of six-ish months ago, you could only open a bank account online if you are a single-member LLC. But if you are a multi-member LLC, you couldn't do it, because just the added complexity of the UBO setup and needing to verify, it's multiple people. When you have single-member LLC, it's most likely that the person applying is the owner. But as soon as you branch now, the complexity means there's going to be a ton of document work, and so there's just pushing that to branch. So there's opportunity everywhere.

Reggie Young: Yeah, a lot of opportunity for sure. I know you've been thinking in terms of business verification in terms of rails, which I find is an interesting framing because I think rails and stablecoins and cards and ACH and wires, those sorts of rails. But in business verification, I know there's increasing this idea of verification Rails.

Walk listeners through what that means in the business onboarding context.

Kyle Mack: Yeah, you listed some of those off, like rails in the context of payments, what are the tools that we have to move money. And I think about for identity, what are the tools that exist to help relationships and identity move across all the different places? Within payments, there needs to be reliability, it needs to be fast, standardized, interpretable. Identity is very similar to that.

What we're trying to build at Middesk really is a unified identity layer that connects the key parties that all need to work together to grow the economy. For me, this is a small business, the business owner, it is the government who is a necessary party in this, and then it is the platform, which could be the bank or the payments company or the lender or whatever it is. What we're trying to build is the underlying infrastructure that connects all of that stuff. 

Someday, I hope, if we do the thing we're going after well, if I'm a business owner, I should have an account with Middesk. If I change the address of my company, like we move to a new office, I hope Middesk someday is the place that people go to make that change. And then we automatically propagate that change to any of the governments they're registered with, like any of the banks, the financial partners, and really think about how to make this a fully connected rail that allows their data and intelligence to move across all of those connected parties instantly. That's how we think about it. And the different products that we're working on are starting to build that network of setting up government accounts and managing those accounts. Obviously, we have all of our banking and fintech customers who are trying to engage the small businesses, and maybe eventually we start thinking about how to give some of those tools to the business owners themselves.

Reggie Young: Yeah, I love it. Verification is not the most sexiest area in fintech, but I think there's so much interesting stuff happening in verification. There's this pipe dream of portable identities that I know a lot of crypto has been investing in, like how do you put identities in the blockchain and all that. I don't know, it's exciting. It's not a sexy space, but all the rail stuff you just walked through, there is a lot of really interesting strategic stuff that has super impactful experience on SMBs and others.

Kyle Mack: I keep going to that idea of picks and shovels business. I agree there's plenty of sexy places of fintech today, and you're talking stablecoins. Within stablecoins, the number one use case today is B2B payments. The reality is identity is still a thing, and managing any on-and-off ramps still require needing to get through KYB. So I'm all for the sexy versions of it, and I love the idea of being I want to help those companies unlock their missions by making it as easy as possible for them to build the products they want, with the experiences they want, enter the markets that they want, and not have to think as much about how to do all of that while managing risk and meeting their compliance requirements.

Reggie Young: Yeah, definitely. I'm all for picks and shovels businesses. Yeah, I'm recording this from San Francisco. I think about Levi's, the jeans. It's like that's picks and shovels and jeans and the gold rush, and the businesses are still around hundreds of years later.

Awesome, Kyle, this has been an awesome conversation. I have my standard wrap-up question for a last one. What have you been thinking about a lot that you think folks in fintech aren't thinking about enough?

Kyle Mack: My team, when they hear this, will probably laugh at me for saying this, but just bear with me. Do you ever hear when you talk about the context of hiring people, do you guys ever talk about the idea of like T-shaped people? Do you know this? It's like the idea of being you want people who are generalists and can touch many things but have depth in one very specific thing.

Reggie Young: It's been a while since I've heard it, but yeah, it is interesting, 

Kyle Mack: Bear with me here. So I've been thinking about-obviously we sit in this world of identity and risk of this fintech infrastructure world. And so something I've been thinking about is there are tons of companies in our space, and in pockets of it, it's quite crowded. And I think what that means is the budget, it's diffuse across tons of companies. And I think that the reality is that has made it difficult for there to be real breakout businesses. There's plenty of good businesses, but real winners? Because the budgets and the spend gets chopped up across so many people.

Okay. So then what I think is happening is so then companies push to not be a single point solution, but then they push to move to more of a platform product because we want to try to have ways to grow the market and be able to take over more of that market. The reality of the way that most fintech infrastructure identity and compliance businesses do this is they wrap and resell a bunch of other products. This has just been super top of mind for me.

So then I go to this idea of a T-shaped business. People start doing one thing, and they're really good at it. We are great. I'll put a super fine point on it. Non-documentary KYB of US-registered businesses. I think we are the best in the world at that thing. But then people start to bolt on these other products, and they start to build the T.

The reason that I mentioned that, and I think for the people listening, any of the operators, it's interesting to think about, is like when you go to the companies you are working with to help with identity risk and compliance, especially the ones that have a very wide breadth, a wide offering, being really clear on what the thing that they're great at, because they've had to build the expertise at it, and what are the things that they've bolted on to be able to grow the product suite, that's an inevitable thing, like people have to move to be a platform. But I think it's especially important for risk and compliance because if you're going to go to them and looking for guidance and advice.

If you want guidance on KYB, you should really talk to Middesk. We are very deep on it. That is our T. But if you go to companies that are wrapping other solutions and ask them about best practices around KYB, like the guidance you're going to get will be fairly surface level. And so I've been thinking about a lot from Middesk company strategy, and I think it's relevant for folks that are operating because it's important to understand the strategies that the companies you're trusting with to help you automate this stuff, the strategies they've taken to build their business and what they're great at.

Reggie Young: I love that. That's funny. I was thinking a lot, as you were talking, about stablecoin infrastructure right now, and they're this nebulous. I'm thinking over stablecoin companies. I know it's like what's the long leg of their T. It's too early for I think a lot of them to have differentiated. But that is an interesting sort of framework for thinking about, not just verification, not just stablecoins, but a lot of businesses like that.

Kyle Mack: Yeah, we'll see if it sticks. I've been on this thing. That's why I say my team will laugh because I've said it probably a hundred times in three weeks. We'll see how long it sticks around for.

Reggie Young: I love it. It's a great idea. Great framework. Cool. If folks want to find out more about Middesk or find out more about you, where should they go?

Kyle Mack: For Middesk, it’s middesk.com, and for me, you can find me on LinkedIn. I’ve got an emerging X presence, but LinkedIn is probably better.

Reggie Young: Awesome. Kyle, thanks so much for coming on. This has been a great conversation. 

Kyle Mack: Thanks for having me.