Taktile’s Journey with CEO Maik Wehmeyer

In this episode of Fintech Layer Cake, host Reggie Young speaks with Maik Wehmeyer, co-founder and CEO of Taktile, an AI-powered risk decisioning platform redefining how fintechs, banks, and even telcos approach mission-critical decisions.

Maik shares the multi-year journey from MLOps tooling to a hyper-flexible decisioning platform, and how Taktile now powers over 150 institutions. They discuss the evolution of embedded finance, overlooked AI opportunities, and why enterprise deals are where the real money is. This episode is a masterclass on product-market fit, infrastructure innovation, and building a culture that attracts the top 1%.

Transcript

Taktile’s Journey with CEO Maik Wehmeyer

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 Maik Wehmeyer, the co-founder and CEO of Taktile.

We cover it more in the episode, but Taktile is a next-generation risk decision platform built for the era of AI. Their tooling helps fintechs and banks to build, monitor, and optimize automated risk decisions across their entire customer life cycle, from onboarding and credit underwriting to fraud and compliance transaction monitoring. Taktile is used by industry fintech leaders like Mercury and Zilch, as well as enterprise institutions like Allianz and Rakuten.

Maik and I cover the company's background, risk industry trends, his lessons from building at the cutting edge of risk in AI, and much more. This is part one of a two-part series with both Taktile founders, so make sure to follow Fintech Layer Cake wherever you get your podcasts so you don't miss our next episode with Maik's co-founder, Max Eber.

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.

Maik, welcome to Fintech Layer Cake. Really excited for our conversation today. Taktile has been taking over fintech, so very excited. You guys have been building something super interesting. Let's maybe start with some of the basics. Taktile is an automatable, low-code risk decision platform. I'm going to guess that some listeners may not understand what I just said, so let's break that down a little bit. What is the Taktile platform, and what does it help fintechs and financial institutions to do?

Maik Wehmeyer:
Thank you, Reggie, for having me. Super excited to be on the pod. Big fan. Been listening to most episodes. Hopefully, I'm contributing something useful for the listeners today.

Reggie Young:
I think you will.

Maik Wehmeyer:
Taktile, you already gave it a good intro. We call ourselves an AI risk decisioning platform. What does it mean? Overall, the goal of banks, insurance companies, fintechs using our product is make better decisions in general. That's how we built the product. What does it mean in detail?

We want to be the operating system for the most mission-critical decisions within, let's say, a fintech, a neobank, a bank. What is that? That can be a KYC decision. That can be a credit decision. That can be a transaction monitoring decision, a fraud decision, even debt collection. There are a lot of things, and I think what differentiates us from many other companies out there is that we started the company with a very horizontal vision. We didn't want to be the best transaction monitoring tool. We wanted to be something that people can use in a hyper-flexible way to really make sure they can run their most mission-critical decisions.

What actually helps a company doing that is if you know exactly what the decision is you should be making as a business. Normally, the alternative to Taktile is you just hardcode everything into the backend, and you have an API, and it makes a decision. But in most cases, people do not know how to get to the best decision or the macro environment changes in terms of credit. Maybe interest rates go up and then you need to optimize your strategy, or the CEO says, of the bank, I want to become more profitable and you need to start iterating again on your policy. We want to be the platform where people can test and iterate and optimize on their decisions.

Reggie Young:
Love it. We're recording this in May 2025. You and Max started the company in 2020, but you seem to have hit a particularly amazing inflection point 2024, at least from what I've seen Taktile popping up all over in fintech. Let's maybe look back to that 2020 timeframe. Why did you start the company? The other end of that is all the adoption you're seeing right now. Why are you hitting such a good product market fit right now?

Maik Wehmeyer:
It's actually a great observation from the outside. Not many people have done that.

Reggie Young:
I'm sure there was a lot of stuff between 2020 and 2024 when you started taking over.

Maik Wehmeyer:
I can tell you the full rundown. There's a reason why we started the company in 2020, and now it took four years to be on the radar of most fintechs and a year later for me to probably be on the pod and be invited. 

Max and I are both machine learning engineers. Before Taktile, we built a mini type of Palantir, almost like a machine learning consulting company. We started to build models- Max and I were building models for huge companies like Wayfair and e-commerce, any type of bank, insurance company out there. We just tried to hire very smart people and put them on the toughest problems in the world, and we built models, very horizontally, from pretty much every industry you can imagine.

Then we said, okay, now we want to build software. So we applied with Y Combinator with the pitch, and actually just uploaded it last week to my LinkedIn. If everyone wants to see it, it's pretty funny. Maik and Max, five years ago, trying to pitch a company to the YC panel. We wanted to build an MLOps tooling company. Us being engineers, we wanted to help companies to deploy their machine learning models in every industry, for every type of model, for every type of use case. That's how we started. That's how we went into Y Combinator and started to build a very developer-focused infrastructure play.

What then happened was on the one hand, we didn't get that much traction. We got killed by the cloud providers. AWS launched a product called SageMaker, GCP, a product called Vertex, Databricks got something out there. Competition got pretty rough, but we had some fintechs, very sophisticated fintechs, using our product for the ML decisioning. What then happened was that the business teams, the credit teams, the fraud teams, they said, oh, my gosh, I can now test and iterate on something in a production API without engineers. The only issue was that you had to write code in order to do that. If you're a credit analyst in a fintech, you're probably not the greatest software engineer. You just know exactly what you're doing in terms of your domain.

And then we started to build a no-code, low-code environment on top. And then people asked us, hey, we also need data integrations. Can you build data integrations? We said, sure, we can do that. We listened very closely to the market, to the customers, and to be honest, pivoted the company. We threw the first 200,000 lines of code into the trash in maybe mid-2022. It was a big moment. We started to rebuild the full thing and then launched the product only in January 2023. That's maybe two, two and a half years ago, which is why you mentioned 2024, we got on the radar. This is when we started to actually find a strong product-market fit and started to scale the company and now have been running in two and a half years to more than 150 customers. It's been a lot of fun. We've been through a lot of pain, but now it's a lot of fun.

Reggie Young:
I love it. That's quite the journey. Maybe to help listeners understand, what does the typical customer journey look like at a high level for a Taktile customer coming in?

Maik Wehmeyer:
Taktile always operates in the background, meaning if you are a neobank and you apply for opening a bank account and having a credit card, we never show up. We're purely operating in the backend of the fintech or the bank. However, we operate the credit teams, the risk teams, the fraud teams. They work with us on a day-to-day basis. Imagine you're a credit analyst at a big neobank. In the morning, you get a coffee. The first thing you do, you log into Taktile. That is the operating system for them.

A typical customer journey, we're being called every time that there's a critical decision, so opening a bank account as a KYC check. Next thing is you want to apply for a credit card. We're being called. Then you have the credit card and you want to buy something in a supermarket, and we do the transaction check. Every time there's a decisioning piece happening, that comes into play. That's the customer journey of our customers, if that is what you referred to, I think.

Our customers, we normally come in and try to really replace the heart and the core of the business, but never bringing the pure brain intelligence. That is what the very sophisticated players, they want to own it themselves. Of course, we come with templates and we have proposals, but in the end, we would not have the possibility to close a new bank if we would have told them, we just outsource your decisioning. No, we don't do that. The credit IP, the risk IP, the fraud IP, that stays with you.

Reggie Young:
Yeah, it makes sense. Taktile is a really interesting place. We get this at Lithic as an infrastructure-type business where you get to see all these fintechs. You get to see all the things happening in the industry. With that in mind, are there particular trends that you're seeing that are top of mind that you think maybe folks in fintech, they're underestimating or they don't fully appreciate?

Maik Wehmeyer:
That's a great question. I think I can talk maybe about three buckets that we see. Why do we see that? Because we serve fintechs of all stages. We serve banks. We serve very different geographies. I think one big topic is, for us, the shift to more, let's say, contextual and then embedded finance. Historically, there were a lot of standalone financial products. Now what we see, you can embed them directly into your platform, into a marketplace, into a journey. That's not only a fintech credit-related thing. That's also in insurance. Like Allianz is a big customer of Taktile. They have a unit, which is Allianz Partners, which is an embedded play. So I booked a flight on United.com, and the time I checked out, everyone has seen that they offer you an insurance.

Those type of things, they require a couple of things from an infrastructure side that are quite hard to do, but once you've nailed them, it's really profound. You need a lot of data. You need data, which is super real time. And then you got to do the underwriting, you got to do the risk decisioning at the spot where the customer hits you.

But people sometimes forget that many of those customers, even marketplaces, they have data that no bank or fintech could actually get to. Like Shopify Capital is more than half of the profits right now of Shopify because they know exactly what store owner has sold how many goods at which point in time. If you would be like a traditional SMB lender, it would be much, much harder to make that risk evaluation for a merchant. Shopify can do that in the perfect way. We've seen incredible companies like [Odefi], Parafin popping up. They're making use of the access to the data and fast decisioning to actually launch products that would not have been possible 5or 10 years ago. I would wish for more fintech founders to go into that direction and actually start to innovate because the time is right now.

That's one. A second one comes, I think, an interesting observation from building tech over the last two or three years. A lot of banks and lenders, they still run on very old legacy risk engines. I do remember when we raised our Series A, we brought the pitch that we want to build a decisioning platform, not an MLOps tool anymore, but a decisioning platform. The big question from investors that came to us and they said, oh, I can see you closing your seed stage company, a Series A company, maybe Series B company, but you will never be able to close an advanced player, like a Nubank, like a Remitly, like a Mercury. That's so core to what they do, and they've built so much over time themselves, they would just never switch.

I didn't know at that point in time. There was an unproven hypothesis by us. I was convinced that if everyone builds something in-house, then there must be a good opportunity for a new category. But just because people have built something in-house doesn't mean you can productize that. That is where some people like, oh, they're trying to find problems to solve for companies, but then they underestimate how hard it is to build a product that actually can be used by everyone in a similar fashion and you have a high code reuse.

I think what has happened now over the last two, three years, the whole category of decisioning has really hit the road, the good teams re-evaluating their tech stack every two to three years. Like in Nubank, they said, look, there's a sunk cost fallacy, which they're super aware of even if they've invested a lot of things, they're super open to go out and just see what's out there. Maybe it was the right decision in 2015, ‘16, ‘17 to build something like that, but maybe now a player has built something new. Those are the pioneers, and they helped us a lot to grow.

But then I'm still surprised by the number of banks and lenders that haven't really understood that. What can you do nowadays with dynamic ML-driven underwriting? How can you use cashflow data? How can you behave data? How can you use non-traditional data? And that's the next frontier. I think that is, for me, a big trend where I do see a lot of under-appreciation still in the market for good decisioning infrastructure. People should get ready for the wave of AI to be able to actually implement that. Everything which is built in-house and legacy, it's normally much harder to keep up with the speed there.

That would be my second one, and allow me to talk about the third one because AI is in everyone's mouth. It's super hype, but I think many people confuse gen AI with chatbots. A lot of clients, also big banks, they come to us and say, how can I use gen AI? There are incredible companies out there that use gen AI as chatbots and they try to get support out of the way. This is a very different category.

But if you look at the PhD level type of decisioning, which is what we do, that's risk, that's fraud, that's credit, then the question is, what can you actually do with AI, and how can you use it? I think the misperception is it's hype, but the final decision, in many ways, is not done by gen AI, by the LLMs. LLMs are not really good at structural data. They're not really good at making credit decisions because they're just trained on the public internet, and the public internet doesn't have that much credit default data out there. People keep that to themselves.

So what we see really, really, really powerful is combining gen AI with traditional statistical methods, meaning in the first place, if I'm a B2B lender, I first have to get the data from an unstructured format into a structured format. Once I've done that and I have a structured format, you can run it into a linear model, and it brings you incredible results. We do right now see a lot of traction in the B2B space and bigger banks that just try to get access to gen AI in order to have the first point of information retrieval in the decision process on gen AI. And then afterwards, this is being fed into a decision engine to actually make the final decision. So yeah, gen AI is very hyped, in many cases a bit misunderstood in the risk world.

Reggie Young:
Yeah, that's fascinating. It's a brave new world with all the AI potential. Beyond Taktile's current offerings, where do you see the future opportunities? What do you see as some capabilities that you want to expand and extend to?

Maik Wehmeyer:
It's one of the things which is so top of mind right now in Taktile, so I'd love to talk about it. I think the advantage that we as Taktile have is that we do not come from a traditional credit or risk background. Many of our competitors, they let fraud teams or risk teams in big fintechs, and then they say, oh my gosh, I've built something here, which is so cool, I should sell it to other fintechs. And then they start to build a product which is very, very focused on their historical problem space. Max and I, we've not been in credit and fraud all of our life, we've just built ML models for every type of player, every type of industry. We came in and designed the product in a hyper-flexible way, that you can theoretically use it for every type of decision.

What now has happened in the last two quarters is that we've closed customers that I didn't even expect that we should be closing that fast. A big bucket is insurance, because if you think about a decisioning in terms of a data workflow, data comes in, their models, their rules, you make a decision. And you have to do that for claims management as you have to do it for credit underwriting. Of course, it's a very different decision, but it's very similar. So we have a lot of traction right now in insurance, claims management, insurance pricing is a big bucket for us. I expect that, knock on wood, by the time of we one day IPO, I expect that to be maybe third, half of our revenue.

So that's insurance. Then very interesting, energy. We have a couple of huge energy companies now utilizing Taktile. When the first sales agent from us came to me and said, Maik, I got this huge energy company. I was like, why does an energy company need risk decisioning? I don't get it. But the fact and the truth is that in many countries, when you have a contract with an energy provider, even if you cannot pay anymore, they still need to deliver energy to you, meaning there is a credit default risk where you're not lending money, but you're lending an asset, which the energy provider has to provide even if you don't pay. That is an unexpected big vertical for us.

And then maybe a more obvious one, but it was not obvious to me as well, is telco. iPhone by now costs you 1,500 bucks. When I'm T-Mobile and I sent that thing to you, I first want to make sure that, Reggie, you actually exist and there's no fake address. And second, even if you do exist, I want to make sure that you actually can pay the monthly payment because it's not your phone. You're using it for two years until it's yours. So telco is also becoming a quite really, really strong vertical for us.

I would sum it up at every asset-based business requires complex risk decisioning. If that is the North Star, you probably will find new industries that are somehow related, even if they're not obvious in the first look.

Reggie Young:
Yeah. It's always fun to see those unexpected cases come out of the woodwork. For Taktile, do you have to do a lot of LLM training that differs by the type of if it's telco versus fraud for fintech? Does it materially differ?

Maik Wehmeyer:
Not as much as one would think, because in the end, there's still a consumer. There's still Maik. There's still a Reggie who's being evaluated. It doesn't really matter if you're getting a credit card, if you're getting a phone. The evaluation of you as a risk asset is similar. I think the more you go on market, the more data actually is there, which is super, super powerful.

I'm not a big believer in companies telling you they've pre-trained models. A lot of people in the market do that, and they say, oh, we've got these out-of-the-box models for credit and for fraud and for everything. I've not really seen many of them working particularly well, especially me being in industry for such a long time of building models. In the end, it's so dependent on what you want to achieve as a business and depending on what data you can have. That is what you have to train and optimize..

The more you go upmarket, the bigger the company gets, the more interesting it is actually to have very customized models, and they just become so much more powerful than anything that anyone can give you out of the box.

Reggie Young:
Yeah. Interesting. Interesting. Makes sense. Looking back at your journey over the past five years, what's one thing you wish you'd known when you started Taktile that you think would be valuable for other founders? I guess, in particular, for founders who want to leverage LLMs in whatever they're building and unleashing.

Maik Wehmeyer:
I would talk maybe about one observation, which is a fresher one, with regards to your question about LLMs. And then I have two things that are actually interesting for founders in general as an advice, which I would have appreciated to have.

With regards to LLM, I was just in Napa, close to San Francisco and the wine region. Y Combinator invited to a retreat. They're doing an incredible job to keep helping you over time if you grow much bigger. They organize an event for AI in general. They invited Sam Altman from OpenAI as he ran Y Combinator for a while.

We talked to him for the weekend, fireside chats, everything. What I really appreciated there, we all have to operate under the assumption that the models will be possible to do everything. So when you found a company now, sometimes you try out a pilot, a prototype. Okay, is that a problem I can solve with the latest, whatever, Anthropic, OpenAI model, Mistral model? Then you come to the conclusion, oh damn, that doesn't really work. It's not good enough yet.

That's not how you should think about building a company in that field. You should always think about, imagine, could another 5%, 10% more accurate? Do I then have a product that solves a real business need? I can guarantee, after that weekend and seeing what's already possible, what's not even public yet, I would now found a company in that space, assuming that it can do everything. I think that's much more specific to LLM case, but it's hard to sometimes have that imagination of everything will be possible when you need to try it out and show value the first day founding a company.

That's the one. And I think two general remarks I wish I would have appreciated more. I got both of them as a founder, but I just didn't understand them. So I try to emphasize it now as much as possible. The one is, which is also a Y Combinator mantra, which is build something your customers love. And I was like, okay, man, that's so obvious. Of course, I build something that the customers love. And then I said, my group partner, Michael Seibel, he said, no, you have to listen to your customers on a day-to-day basis.

I mean, I talk to my customers, but I think what he meant is being obsessed with your customers, not 10%, not 20% of your time, but 80% of your time. It's every founder's job to found product-market fit. That's your main job, nothing else. It's so cool to hire. It's so cool to close partnerships. It's so cool to do podcasts. There's a lot of cool things you like doing, but the only thing that matters by the end of the day is building something that your customers love because then you have product-market fit and then you will succeed. So I think just trying to emphasize this very obvious thing to be more obsessed about it.

And then another one, which I think is also very specific to me as a person but also to Taktile as a company, is creating a culture that allows for the best people to really join and stay. Okay, such an obvious one again, but I don't think people appreciate it as much as they could. If you find, not the top 5%, not the top 10%, but the top 1% of people in the market, and you can actually create a home and a culture for them to think it's the best place for me, for my career to be at right now, and let them succeed and not let them leave, reduce the churn, that just changes everything of the company.

An incredible engineer, you need to find them. You need to hire them. You need to onboard them until they're being productive. Just such a long time. Again, when they leave, people underestimate how expensive churn is for a company. So rather, spend another great offsite, another great feedback session, another great one-on-one with the best people to make sure they really think they shouldn't be working anywhere else in the world. So yeah, those are the things that were always obvious to me, but I didn't follow up on them to that extent when I started the company.

Reggie Young:
Yeah, I love it. Those are all great insights for sure. I have two wrap-up questions. The first one, we've talked a lot about AI in this conversation. What do you think is overhyped in AI right now and LLMs, and what do you think is underappreciated?

Maik Wehmeyer:
For me, underappreciated is that many people do not think about how to actually get the LLM into, let's say, an enterprise, into a business. We think a lot about these experiments that you can do, like how strong is an LLM in itself, but now imagine you're being Bank of America or you're being Allianz. Let's take Allianz. And here's Anthropic. It's the latest model. They're always being benchmarked, like how accurate is it? How is it compared to OpenAI models? By the end of the day, they're all so good that they can already deliver so much value. But until this API makes it into the claims management process of an insurance company, no one really talks about that.

People are thinking a lot about, how can I improve the LLMs? How can I feed in more data? There's so much- 90% of the articles I read about gen AI and discussions I have talks about how to build better models. The models are already so good that they're so much better than humans. It's so many tasks. People should really think about, how can I actually bring them into banks, into insurance companies. That's hard because there's legacy infrastructure, and there's an API no one can use. If I'm the claims manager at a big insurance company, I cannot use an API. I need a solution.

That is one thing which we're trying to enable and trying to help and working very closely with two of the leading LLM companies, a lot more things to announce in the future. I think that's going to be one of the most exciting things for us as a company, to actually help getting these LLMs into solutions applications that people actually can use.

Reggie Young:
I love it. I'll keep my eyes peeled for those announcements. Awesome. For my typical classic wrap-up question, what have you been thinking about a lot lately that you think folks in fintech aren't thinking or aren’t talking about enough?

Maik Wehmeyer:
We're an infrastructure player that sells into fintechs and to banks, insurance companies and telcos and energy providers. I'm not the biggest fintech expert. The way I would answer it, I would talk about all fintech infrastructure folks that are in a similar seat like I am, which there are a lot by now out there.

What many of the people I think are not thinking about enough is getting their head out of the ecosystem into the old industry. When I moved to New York two years ago, I started to be sucked in by this fintech bubble. I was like, oh my gosh, I didn't know anything, nothing. Someone I was talking to, oh, have you heard about this provider? I've never heard of them, because it's just not where I come from. I had to catch up there.

But what many fintech infrastructure players do, they just sell to their friends because that's where they are. There's always this next fintech, all this cool Ramp and Brex. I want to sell to those. But I think the big- this is all peanuts. It's an incredible first stage of segment to sell into because they're super modern and they help you to build a good product. But I would wish for fintech infrastructure founders to really think about the old industry, think about the enterprises, because this is where you just land million-dollar deals. So hard to land a million-dollar deal at a fintech. But go out there. The United States has the biggest enterprises in the world. Some of them- I don't know, one of the insurance companies we're selling into right now, $50 billion in revenue, $5 billion in profit. $5 billion in profit, they don't care if they pay $1 million or $5 million. It doesn't really matter. I think this is like getting out the heads of this fintech world into where the money is. I think that's one thing where I think- would be my take on things that fintech folks are overseeing.

Reggie Young:
Awesome. I love it. Maik, thanks so much for coming on the podcast. If folks want to learn more about Taktile, just go to taktile.com, T-A-K-T-I-L-E dot com. I look forward to having you back on the podcast at some point.

Maik Wehmeyer:
Cannot wait. Thanks, Reggie, for having me.