April 7, 2023

Episode 229: Maik Wehmeyer, CEO & Co-founder of Taktile

Maik Taro Wehmeyer is the CEO and Co-Founder of Taktile, a global automated decisioning platform that is revolutionizing how fintechs harness data to steer their business. In a world increasingly run on automated decisions, Taktile’s low-code platform makes it easy for credit and data teams to build more accurate decision flows, adapt to change quickly, and, ultimately, improve outcomes. Prior to joining Taktile, Maik worked at QuantCo, McKinsey, and SAP. He has significant experience in the machine learning and financial services spaces, serving as a member of EU Commission’s AI Alliance and the German AI Alliance.

Julian: Hey everyone. This isJulian with Behind. Company Lines. Today we are interviewing Maik Wehmeyer, CEOand co-founder of Taktile, a global automated decision platform that isrevolutionizing how fintechs harnessed data to steer their business. Maik, I'm sointerested in chatting with you today and excited to not only learn about yourbackground as a founder, but also Taktile and how you're just helping, I was, Iwas going through the product and it, it seems, And using data in a reallysophisticated way to, to not only capture a bunch of variables, but also togive that insight intelligently to businesses.

And it seems like even more so, not evenafter the pandemic, but in more recent events of not only this year but latelast year, it's becoming more and more imperative for companies to be verystrategic about the moves that they make, the decisions that they make, andalso the communication and the speed at which they make those.

So, really interested to hear yourinsights and get more, more into the weeds of, of Taktile and what you're doingand who you're helping. But before we get into all that, what were you doingbefore you started the company?  

Maik: Awesome, Julian.Thanks for having me on the show. Very excited to deep dive on, what you do atTaktile and you know how data actually is being used nowadays in world to makemake better decisions.

Yeah. So, what was I doing beforeTaktile? So, it's actually started pretty early 2016. I was still in gradschool, grad school and doing research on machine learning, machine learningmodels and big data was just, the very early times of, of data science, therewere hardly any data scientists out there.

It's also where I met my co-founder,max. He was getting his PhD doing something very similar and we thought we wantto try out. Machine learning actually works in the wild. There were hardly anyuse cases, hardly any people who've tried it out. So we started to approach bigcorporate big banks, big insurance companies, and asked them, Hey, can you giveus like some of your data?

And we will then potentially figuresomething cool. And we'll want to show you on how these novel statisticalmethods, these novel machine learning methods, actually potentially could bringyou some business value. So that's actually what happened in the beginning. Wewent out and started to build models, and then we were doing that for we weredoing that for three years.

Building up with a, with a bunch ofother. Fellows from from uni building up this machine learning consultancy.Yeah. And then it was 2020 and we started Taktile.  

Julian: And it's incredible tothink about the, the fact that businesses weren't using data Now that it's socommonplace for any business to have some data component at any part,understanding it, collecting it, and also making decisions based off of.

When you were first going out theretalking to companies, where did you start with the data? Did, did you, were youcategorizing it and then just kind of built models to, to make predictions. D,describe a little bit of the mechanics when, when you have a big DA data set,how do you start labeling it? How do you start identifying.

What your key signals are to startextracting, important bits of information and what are, what is the processlike to, to do that and when you were first going out?  

Maik: Yeah, it's a, it's,it's a great question and a lot about data is you don't know exactly what's inthe data and no one can tell you before.

So you have to start digging and youhave to start experimenting and then you slowly start to figure out, Going on.Yeah. And it's a little bit like research. You need to understand what's deeplyin the data. And the way we did it most of the times were we were starting torun very structured experiments.

We're trying out stuff you kind of likeiteration, fast iteration on certain like hypothesis that we had and what wewanted to figure out. And by running these experiments, you start gettingsignals. On the potential value in the data. Yeah. Kind of like, that maybesome examples you've probably heard before.

If you think about pricing, dynamicpricing, risk adjustment price aren't always the same. They change all the timeand they do change in a, and maybe looks very unstructured to the outside user.But in the end it's a very structured process behind by, algorithms trying outdifferent vari. Of, for example, prices or price structures.

And by doing that you can really, reallynarrowly measure on on actually what changes of the algorithm would have, whatchange on a user perspective, on conversion rates, on price elasticities and soon and so forth. And that's pretty much the approach we were running.  

Julian: Yeah, I can imaginethat the data is not always clean.

And what do you do in the sense whendata or your missing components of data, do you just kind of put a, some kindof I guess, like a variable that kind sits in its place? Or what do you do inparticular to. To make sure that the model doesn't get broken. And when, whenyou, when you're working with the big data set and if you have missingcomponents to it.

Maik: Yeah. So, one, onething every, every person who's worked with, with data has realized, real worlddata is always very messy. Yeah. It's not, it's not a Kaggle Kaggle challengewhere you get a well structured and nicely set up data set and you can startrunning your, your models on it. Real world data is always messy and it's a bigpain to clean it up.

So you know that that's part of thework. That's part of the work. But I think, apart from only like cleaning upthe data set, the very interesting piece and also kind of like the inside thatwe then got over these years was what is actually the right. Infrastructure to,to do that on. Because you can do something in your notebook and you can try tosome to like do something, very manually.

But in the end, most teams want toactually control their decision flows on their own and want to do all of thaton their own. And that's kind of like, also like kind of the interesting pathof letting, trying out building machine learning models for other companies atsome point in time. Two things that they, we figured out and the one was theyactually want to own all of the.

Themselves. Like if you go to a bank, ifyou go to an insurance company and you are, you're thinking of risk models,it's the heart stuff. Like so deeply important also from a regulatoryperspective, but also from a business perspective that you own these models.And then the second thing is that a lot of the work in the end, when you buildmodels and if you think about decision flows, if you think about decision logicgoes actually into building the infrastructure that you can start experi.

And iterating very fast and they canneed, give the tools and the shovels to people that they can actually startrunning on their own and claim flake. That's what actually led us then to toalso start the company in 2020. Cause we saw how much value could be created bydelivering this infrastructure to then internal risk teams, data science teams,business teams that actually are responsible for designing these skillsnowadays.

Julian: Yeah, it's incredible.About, the whole transition to, it's either no-code or low-code types ofplatforms. And looking through talkative, it seems like there's, there's thisinteresting I don't know if it's an org, it is like a composition of, of notonly hardware, sorry, not hardware, software, but also enabling teamsinternally at other clients to be able to use that software.

Now, when you're making theserelationships and you're building these relationships with these companies, howmuch. How much is involved in the onboarding process, being that you'veminimized it, I'm sure, to such a infrastructure foundational level on, on thecode side, but also they need to build sophisticated models on top of that,what's the onboarding process and, and what's the learning curve look like fora lot of these companies who are implementing talkative?

To, to start giving them insights intotheir business and, and and, and start to kind of help their decision making atTaktile.  

Maik: Yeah, it's, it's veryinteresting. I mean, like, they are, I would kind of like separate between twotypes of, of, of users there. Yeah, the one is the teams that, have been, havebeen, working on, on, on models, on, on decision flows for work for quite someyears.

More like the very existing verysophisticated players. And then they're the young, the younger companies,they're just starting out. It's a young FinTech, they wanna launch a creditcard for immigrants in the United States, and they just, they're just kickingoff the tires. And, if you talk about, if you talk about the first categoryit's always quite a lift and shift.

Yeah. Like if you, if you, what I saidearlier, if you rip out the heart of of a bank or an insurance company, it isquite some work actually to get the decision flows onto, onto a platform likeours. On the other hand, if you're just starting out, then the onboardingprocess is actually quite fast because what we do is we actually have sometemplates for these companies to start out and running from the.

And then, instead of them trying to,define their first credit policy over the, I don't know, a period of 12 months,they then can do it in three to four weeks. Because a lot of the things alreadythere, infrastructure is there and and the flow's already there as well.  

Julian: Yeah. Why, why FinTechin particular when, I'm assuming, and I think a lot of us now assume that, thatthat data is pretty much just, collection of variables.

Building these models is essentiallybuilding hypothesis on, on top of that to, to then either validate orinvalidate them based on, some kind of assumption. Why FinTech? What inparticular got you excited about this industry and, and, and, and this inparticular versus expanding into another horizon?

I'm sure maybe there's some plans to doso, being that, low code platforms are just, I think it, they're really kindathe next step as, as the front end became, kind of like Webflow and all theseframeworks and structures, backend technology is now doing that with Bubble andall these no-code platforms.

But why FinTech? What in particular wasexciting about this industry in particular?

Maik: Yeah, I, I'd say firstof all, financial services is a beautiful industry for, for data, banks andinsurance companies. They've been gathering and storing data in a structuredformat since. Decades. Yeah.

In contrast to many other industries,they're just now starting to realize what you can do with data. But thebackbone of a bank and insurance company is using data to make more informedand better decisions, yeah. And start to segment risk in a more precise mannerin order to make to make a good margin.

So that's where, in general, whatfinancial services is super super compelling about. And then, when we started Taktile,beginning of 2020 all these neo banks came also up. Mm-hmm. It was a full. Offintechs coming up, not only in the US but also in, in Germany, in France, inlike Africa.

We have a bunch of customers in Africa,in India, and a lot of like, there was a full new industry coming up and if anew industry comes up, these people also need software. So kind of like we werethen starting and approaching this market from one content perspective, whereactually is our software very useful?

And then also where is actually the wordgrowing into? Very interesting customer base. And that's definitely happenedin, in in the FinTech space in the beginning of 2020.  

Julian: Yeah. And what werepeople doing before Taktile, before they were understanding, and, and, andbuilding, whether it's policies or other things based on the data and, and whatthey were seeing historically for themselves, but also within, within themarket and kind of the broader landscape.

What were they doing before they hadsomething like Taktile?  

Maik: So our, let's saybiggest competition is build your own. So normally people start buildingsomething on their own. Yeah, and I mean that, that's exactly also what we sawwhen we, building models for, for these companies over years before we startedTaktile, was that there was no real good infrastructure out there.

And that's what people always led toactually building something in-house. Normally looks like someone, the domainexpert, the credit analyst, the actuary, the underwriter she put something inExcel. The rules, if you're a student, you're not gonna get a credit card. Ifyou're below 18 years old, you're not gonna get a credit card.

And that is then being sent to thebackend team, to the IT team, to the engineers and data and hard codedsomewhere in the back end. And now you have a problem because, What you alsoasked earlier, what we talked about making good decisions. It's aboutiterating, it's about experimenting. Sure. And if that transition between thepeople who actually wanna make the decisions and until it ends up in the realworld, takes a couple of weeks and it's very rigid, you cannot experiment.

Yeah. Like you kind of like you are, youare, you are dancing in shade. You actually want to, like, interest rates go upand you want to like change something and you have to call it, but no, we don'thave time for you for the next six weeks and you cannot change it. And thenyou've changed, you put something live and maybe it was not the right thing todo, what you expected that it would do, and then you have to correct it in a,as fast as possible, but you cannot do.

And that's exactly, what our low-codedecisioning platform is actually doing. It helps it helps these teams to justiterate faster. Yeah. Without being blocked by engineering.  

Julian: Yeah. And what aresome of the things that, you would see companies iterate on? Is it, changing,internal policies, allowing, you mentioned something about not letting acertain type of lender say borrow from that institution for.

But what kind of iterations are theyhaving to make as they kind of, look at the, the broader market or themselvesinternally? What, what do you see that are, that are common that, thatcompanies need to iterate on and, and move quickly with?  

Maik: Yeah, I mean, greatuse cases, everything around risk.

Judgment, credit, underwriting. And whyis that? Also like one reason why we picked FinTech is because I deeply believeinto the value of, fintechs being founded and them making better decisions becausethey have new data and they can harness the data in a more precise manner. Forexample, like, when we raised our series A 20 million.

Quite some money in the bank account,and we wanted to get leasing bikes for our employees. And then we applied andwe got rejected because our, let's say, on Dunham, Brad Street score was just,it's outdated. They use a very old mechanism like making these decisions. Wewere rejecting us like, Hey, like we, we have 20 million of cash in the bankand we were just on five bikes.

I think that should be possible. Andthey like, no, no, it's not possible. You cannot get. And then because they,they just don't use the data which is available out there. And then they areyoung companies, and I'll, back to your question, the use cases there Moss forexample, it's one of our customer, it's a very successful B2B challenger bankfrom Berlin expanding on different countries in Europe and they use bankingtransactions Yeah.

In order to make underwriting decisions.And, that's beautiful because it's not an rigid old score, but it's actually,they look. They look at what goes out. And in our case, there was a 20 millionfunding round. We're not spending that much money. And of course they gave usa, they gave us a good score.

And that's, in the end how you can makebetter decisions. So, and I think there are a lot of these very, veryinteresting examples, how financial inclusion is being enabled by FinTechbecause they can now have access to novel data. And because they're using moderndecisioning approaches that allow them to really use the data and make and makegood decisions.

Julian: Yeah, it's in, it'sinteresting. It also gives me a little bit of a PTSD from when I was applyingfor apartments and they had to, they had to review my, my, my my income fromthe last few months. In New York in particular, they're, they're a littlestringent on that. So, a little bit, a little bit of shock there, but no, inregards to like this whole neobank, all these new banks that.

Are are coming, live and are able toaccess different parts of the population, whether it's because of, like yousaid, more, more up to date underwriting or, they have a little bit more of aninsight. How does that change the overall, not even just lending market, butall the, the financial institutions in general when they're able to make, notonly more intelligent decisions, but quicker decisions and allow access todifferent sections of the population that, may have been Or for some oneparticular reason or another, how does this change the landscape of whetherit's lending or risk?

As these companies build these newmodels and start to understand more about the intricacies, and not onlydecision making, but but population and, and where you can, take a little bitmore risk than than you were traditionally, how does that change, how has thatchanged in the last three years?

But how does that change in the nearfuture?  

Maik: It's, it's veryinteresting because I mean, the market is just being segmented into muchsmaller population, right? And by kind of like new companies coming up and theyhave a hypothesis and they say, well, I deeply believe an immigrant like mecoming from, Germany moving to New York.

Should that person actually get anapartment? Should that person get a credit card? And they think, yes. ThinkMaik should get a credit card, because we have a way on, how to evaluate him ina more, in a more precise manner. And I think that will win a lot of marketshare for the types of personas I am.

And the same example goes on and for SMBlending, for auto lending. And that's only the us if you look at kind of likeother customers of ours in, in, in Africa we have a couple of customers inNigeria. It's very interesting. And, and, and there. There was hardly anyfinancial institution that served certain parts of the population before.

There's no access to finance. And by, byallowing, using, for example phone data, if the people allow, allow the, thenew banks to use them. Yeah. If you look at bank insurances actions, you can,you can really segment the market in a, in a way, which, wasn't done before. SoI do believe that there is, and we, it's also very much proven by the numbers.

There's quite a big influx of customersto the neobanks away from the bigger institutions. Yeah. And, but now it's, nowit's on the neobanks to also prove that they can build a profitable business.And but but I see very good numbers at, at many of them.  

Julian: Yeah. It's impressiveto think about how much, what institution or future lender or what, what haveyou, how much they know about the population.

It's interesting to think about thelittle intricacies, whether it's considering someone's background, if they'retransitioning from another country to one another one as well. But even in morenot de not decentralized places, but places where there's less essentializedinfrastructure. How they, how they, can.

Underwrite, keeping track of otherpieces of information that they know about the population that they can makepredictions on. That's in, that's incredibly impressive to think about thelevel of access people will get. And I'm curious to learn more about Taktile inthe sense that, tell us a little bit more about your traction.

How many companies are you working with?How many people have you been able to. How impressive has your model gotten?What's the traction that you've seen recently and what are you particularlyexcited about in this coming year and and and so on?  

Maik: Yes, so I mean, westarted, we started in, in, in, in 2020 Ren to the white Combinator summer,summer 20, and then launched the software beginning of beginning of 2021.

Yeah, you. Back to back to ripping outthe the heart of a company, you better make sure you've built something veryrobust and stable before you actually allow someone to actually run, run onyour software. And traction has been absolutely fantastic since we raisedseries A mid last year.

We saw a 300% increase in in revenuesfour x of our customer base grow. We have customers now in consumer, which isabout 50% of the business, b2b, and the other, the other part, it's use cases,buy now, pay later. It's a lot about, working capital. We do have insurancecompanies on the platform by now using the software for, for their pricinginfrastructure.

So it's, it's, it's enormous to see how.Traction lately picked up in so many different kind of like use cases on theplatform because it just kind of like for me shows and shows more that theworld is run more on automated decisions. Mm-hmm. But people just didn't havethe infrastructure to really work with this uncertainty out there in themarket.

Yeah. And we try, to. Teams develop abetter relationship to uncertainty. And we just see how actually, like if youcan really enable these people by having, a structured low-code localenvironment, it it picks very well up. So that's great. That's where we areright now. And now going forward you know how, how the, how the game goes.

It's about growing. And at the moment forus it's, it's interesting cuz it's very, very clear that the product isobjectively better than everything which is out there. And now for me, it'sabout how can we in the most efficient way get adoption on the product. Yeah.Which is for our case.

Going up market. So we are now startingto sell too much bigger institutions. Mm-hmm. It's not only physics more, it'snow the big incumbents and, that's a different animal. Yeah. And that's goingbe back to, back to your question in terms of, goals. For this year. It's,it's, it's, it's about whale hunting and going up market as fast as possible.

Julian: Yeah. And why, why upmarket not expand to other industries?  

Maik: That's a, I mean, weare discussing that. We just, we are still discussing that and, finding. Andincreasing your degree of, of product market fit is again, a constantexperimentation, but we feel very, very much, home in the financial serviceindustry and we've proven how much value can be generated in terms ofadditional profitability with our software.

Yeah. And going them from. A big FinTechto, an mid-market bank to an incumbent bank. It's, it's actually very, very,very close and similar, if, if I think about the P and the s, there's a productand there's the market and it just fits so well together for us at the moment.

Moving up there, sure. Works quite well.If you now go to use it for telco pricing or commerce pricing. It's a differentmarket, yeah, I could, I could well see how, how, by the time of IPO there willbe potentially use case in these, in, in these industries. But for the moment,we want to be laser focused on financial services and keep closing morecustomers there for the moment.

Julian: Yeah. And, andthinking about, whether it's obviously we have a lot of macro. I don't wannasay issues, but concerns in, in the general public about not only the economy,but also what that means in terms of transitioning from, whether it's fiat tocrypto. There's all this talk and all this conversation about not only how youcan get money, but where that money is gonna be distributed and, and how it'sgonna circulate within, at least for the us our economy in particular, and Iknow other countries are dealing with changes to their infrastructure as well.

What does, what does that mean in termsof the risk that, that you see a Taktile having? Is it yeah, I, I'll just leaveit there. What are some of the biggest risks that you, your company facesmoving forward?  

Maik: Yeah great question.And the thing, that's, there's always risk, on a young company like ours.

There, there, there's always risk. But,especially on the, on the macro side we now coming back to what we were talkingabout before, which is, moving up market. We're selling right now. A lot, a lotof fintechs and I deeply believe in the future of of these companies, but ofcourse Goldman Sachs or Bank of America, a little bit more stable than someseed series A companies that just raised money in 2021 and one when it wasquite easy to raise money.

So I think that's also definitely. Onerisk to our business that the industry we're betting on is not growing as fastas we want them to grow. Mm-hmm. And by that, our growth would be would beslowed down a bit. But that can be mitigated by diversifying your or. Ourportfolio of customers as well.

Yeah. Are starting to sell to more, alsoinsurance, has very different macro risks than, than banking in itself. And bybeing able that our software can be used and serves as the backbone for theserisk teams, we are making use of that and going into, into insurance and thenbanking, we going up market.

So that's a strategy for the moment to,to mitigate potential from the horizon there.

Julian: Yeah. And, and justthinking about also, with, with not, not even just the underwriting, but makingbetter business decisions or, or decisions overall. How could your technology,what, what are your technology say at.

Svb, were they able to, to use Taktile?Is it a technology that would've helped them steer in the right direction? Orat least understand maybe some trends, some, some kind of greater understandingwithin the macro, macro market? I'm, I'm curious to hear your opinion onthat.  

Maik: Yeah. Yeah. I don'tthink that would've been a good use case for, for the platform.

Why is, why is that? Because, It was, itwas not so much about them making individual underwriting decisions, automatedunderwriting decisions in their portfolio, but it also had a lot of effects of,rising interest rates, their internal strategy. To invest, money from theirdeposit into long-term maturities.

And let's say that specific SVB problemwould not have been, let's say, one of the classical decision flows you see onTaktile, that's normally what you know. How do you onboard with a creditunderwriting KC policy? How do you onboard a customer make you know, with asmuch data as possible, harness the data?

Open banking, accounting, credit scores,put them together and make a very qualified decision. So I think, as much as Ilove our software, we would not have been the one who would've saved svb.  

Julian: Yeah. And ifeverything goes up, what's the long-term vision for Taktile?  

Maik: Yeah. I mean long-termvision.

So Taktile is helping risk teams torevolutionizing on how they harness data to make better decisions. And I think.We deeply believe in the world being more and more run on automated decisions,but as where we stand right now, we know that business teams and risk teams,they're not actually able. To run with that speed and they're not able to buildand iterate on their decision flows, and they're still very much reliant on, onguesswork, and they're reliant on engineering, but that is not the modern wayon how you can actually approach data.

So that's exactly what we want to solve.And if, if, if the world is going like into the direction that I'm thinkingit's going tech Taktile will be, will be a success. And we'll actually growinginto banking, insurance companies in the future as much as possible.  

Julian: Yeah. It's sofascinating thinking about how things are, are changing, at least in, in ourcountry and in the us.

And how much more sophisticated thesemodels will have to be to understand someone's, ability to, whether it's payback a loan or take on some type of risk with insurance and things like that.It's becoming more and more complicated, right? It's not, it's not so cut anddry anymore. It's not a lot of people having a, a W2 employee, then they havesome kind of credit score.

They have some kind of long, longhistory with, with that you can, that you can really kind of track and graph.It's a lot of def, definitely. Or I guess, I guess in your opinion, Lot morepatch. Is it a lot more patchwork of, of different variables that were notbeing considered previously based on, the kind of the, you mentioned it with,with companies at least, but I guess on the individual level are there a lotmore variables that, that banks and all these lenders are having to considerwhen, when building these models and when servicing their  


Maik: A hundred percent. Ahundred percent. And they are we are, we'll soon, we'll soon announce One ofour, our, our new products where we actually connect to the most advanced dataaggregators and, and vendors in the world on open banking, on accounting data,on phone data, payroll data on tax data, on, the credit bureaus.

Yeah. And there is so much. Little, yousaid, little like patchwork. That's correct. You mean like so many coolcompanies are trying to like, get more data and us being the operating system fordecisions where these decisions are actually being orchestrated in the end, theonly thing which is missing is having all of them available directly.

That's what we're building at themoment. And we we're gonna announce that to the press in the next couple ofweeks and who's like part of that alliance that we're building there? So it'sexactly about that. It's not only about. Help to make better decisions. In theend, you need to have the variability, you need to have the data.

Yeah. In order, in order to come up witha good signal on on whatever you want to actually predict.  

Julian: Yeah. And how doesbeing connected with the biggest data aggregates, does that, I'm curious, doesthat change how. We have an understanding and does that change behaviorsbetween different countries and, and how they, whether the, I guess, I guessI'll stop there.

How does change, how does behaviorchange between countries and how will you be able to kind of notice commonbehaviors, having more data available, having more of an aggregate to be ableto feed the machine and, and challenge these models? How, what are the changesor what are the, the differences you see between countries, but also how isdata gonna help?

Maybe key in some similarities thatother countries can use between themselves?

Maik: Yeah, I, I, I thinkit, it just allows a lot more mobility, not only for individuals, but also forcompanies. So, by there, there's a very cool, there's very cool company, calledNova Credit, and they actually translate the credit scores from countries likeMexico or Germany or Switzerland into if someone wants to move to the US intolike a more standardized.

Inventor score. So like they're doing alot of work there on allowing mobility of individuals and being able to haveaccess to financial data. That's, one example. On the other hand, companies,like, like, like Plaid or Ting or, or Coda, they're actually aggregatingbanking data or like coded accounting data, which then allows companies to,expand very quickly across different countries, via neobank.

Like Revolut. Now you wanna launch in 10new countries. Do. Be the one building the connections, integrations to theaccounting systems of each, each nation. Or you just want to connect, to Coda.And sure, by being connected there, you get everything in one developerfriendly api, which is where structured, which is where documented.

So, a lot of these, a lot of thesecompanies out, they actually allow mobility and by that kind of likehomogenizing the the, the, the boundaries between, between countries forfinancial data.  

Julian: That's incredible.Like this next section I called my founder at faq. So I'm gonna go with somerapid fire questions at you and, and then we'll see where we get, if that'sokay.

Let's start The first question. I'd loveto open it up. What's particularly hard about your job?  

Maik: For me, I think thehardest thing for me is to zoom in and then zoom out again. Mm-hmm. know, LikeI need to think about the five year vision and if we're gonna expand more intoinsurance or banking.

And the next the next moment. We have tolike, think about a very specific feature on the platform and how we are gonnadesign that, that it's most intuitive thing, and kind of like, I think that'svery challenging. I lo I love that, personally to, be in a position that I cando that.

But on the other hand, it's sometimesyou go from one meeting to the other and be completely in two differentdifferent words there.  

Julian: Yeah. What's somethingas a founder, you, you found yourself spending a lot of time on that you didn'texpect to spend so much time on?

Maik: I think what I did notexpect to spend so much time on is it's actually iterating on product marketfield. I think when you go into a company, you think you write a business planand then you. And it sells or it doesn't sell, but it's not, it's not that,it's not that easy, kind of like in why Combinator, they always just, theydon't talk about pmf, they talk about the, a degree of product market fit,yeah. And it can be low and it can be high, and it can be very strong and itcan be even stronger. And, and, and, I, I definitely didn't expect us after,where, where we are as, as, as, as, as a company. 50 people raised that much.From the outside, it, it's, it's it's, it's an insanely successful business,but and still we're still spending so much time, and each day you have to getup and you have to think about what your customers actually love and how wecan, keep building that.

So I think I didn't expect that to besuch a continuous, continuous effort even, even after these years.  

Julian: Yeah. Did you alwaysexpect to be a founder or was it something that you found yourself? And andkind of propelled yourself into,

Maik: I think, no, I think Ialways wanted to, I wanted to build, I wanted to build stuff. I think, that,that, that's always been clear since I was young. I wanted to, to build stuff.And, I also, like to, to push an endeavor forward and to win. And I think, thatthat's all we have in tech in our hands.

Now there's competition but we're going.And, that's, that's definitely a motivation between, across the company thatmotion of, winning the market at the moment.

Julian: Yeah. And particular Iwould love to hear some predictions. What are some, what are some grandhypothesis that you have about the overall kind of financial system, as you seeit today?

And then what are some hypothesis thatwe're proven right, whether it was customers or yourself, whether it was earlyon or recently. That were surprising to you that, something that, that you didnot expect, whether it was the use of textile or the predictions that it wasgonna make?

Maik: Yeah, I think that,one hypothesis on financial markets are efficient in some way, but you know, ifyou, if you have more information than the others, and if you have like betterways on onto evaluating data, you can actually make quite some money. Kind oflike a profitability within financial services.

Not, is not. Lack, but it's very muchdriven by, by, by data and, and by making better decisions. And I think that'sone hypothesis, which we know we can see every day on the platform that thisvalue is coming out of out of the data. That's the one. And I think on oncustomers, and we had the hypothesis that people, business stakeholders want tobe completely enabled.

By local environment. Not everyone cancode. Not everyone is a Python program up. Do people actually care or do they,do they just want to, put everything in Excel and send it to someone? Right.And, let them do the work of implementing it. We, we, we, we thought that,decision makers want to be empowered and it looks like that they're, that theyhighly, enjoying being unblocked and being, being empowered by now.

So I think these are, that, that were,that were proven right. From our perspective, maybe one that was we were, we,we thought, 2020, that the world is being completely ruled by AI and machinelearning. It's not only that, our decisioning that many cases, it's verypowerful, but many decision flows are a combination between a machine learningmodel and logic and holistics and, and, and proxies.

So it's, yeah, not everything, not the.Runs on ML in, in financial services. Yeah. And there was definitely one hypothesiswhen we actually started out building more a machine learning platform and thenclarified the value proposition to being a decisioning platform, which mightconsist of some ml, but it's also embedded into rural systems.

So that's, I think we were quite wrongin the beginning and it, it took some. We got that out.  

Julian: Yeah. What, what, whatmakes it not able to standardize? Is it that human behavior is so, and, andhuman kind of circumstances are so a different, intricate, or what in particular,why, why is, why is it not allowed?

Not, not, not allowed, but why is itunable to rule over all decisions and, and kind of consider everything and runalmost automatically without any. Interjection.

Maik: Yeah. I think becausesometimes a rule is just enough for, as an example if I don't want to give acredit card to someone who's younger than 18 years old, Item defined a ruleyounger than 18, no credit card.

I don't need a, I don't need a machinelearning model to be fitted on thousand data points of, of students that are,below 18 and defaulting on their, on their credit card in order to make theirjudgment. I know that, and I wanna make that business decision. So I would justwant that rule to be implemented without, any inaccuracy.

It's a very accurate rule and, andsometimes you know that, that is, sometimes, just the answer. ML is verypowerful, but, Them not necessary to make the effort to train models on data.Sometimes humans know very well what they want and they just need a, just need sometools in order to implement that.

Julian: Yeah. I have a lastfew questions here. I'd love to ask this one cause I love how founders extractinformation from the knowledge that they ingest. Whether it's early in yourcareer or now, what books or people have influenced you the most?

Maik: One book I highlyenjoyed it's called, it's called winning by, by Tim Grover. He was the the, thepersonal coach to, to Michael Jordan and, and Kobe Bryant. Mm-hmm. used to be apro-athlete. When I, when I was younger and I. Lot of I take a lot of energy outof sports analogies and I think, pro-athletes have, have experienced very, veryhard and enduring times.

And that book is, it's a great assemblyof, the journey that MJ and, and, and Kobe actually went through and, justreading how hard they fought in order to, win their championships. You can takeso much out of that and apply that to, to our business. And I, I really, reallyenjoyed that.

And then finding analogies for, forTaktile.  

Julian: Yeah. Yeah. I lovethat. And last little bit, I make sure, I like to make sure we didn't leaveanything on the table. So is there any question I didn't ask you or anythingthat, that you wanted to speak on that you didn't get a chance to? Anythingcome to mind?

Maik: You know, O open topicand not, not something I want to I want to hold a speech on, but I think, jet gp t is, is really coming, coming into the media. GT four just got releasedcouple weeks ago and I, I think we have to find a dialogue. On what that technologyis doing to financial services.

Mm-hmm. So what are the use casesactually that are being that are being positively impacted by that? Because inrisk, risk is not about being a little bit precise, it's about being as preciseas possible. And, we talked about having data sets, which are in many cases notpublic. The internet is public, so Sure.

Like they can Oh my, I can train theirfinancial models on that. Lot of default data of banks and claims data ofinsurance companies not public. So it's much harder for these foundationalmodels to be precise for risk application. And I think it's just importantthat, that people don't overestimate the power of these models to, to the, toan industry which is not perfectly made for that.

We do have some very interestingapplications in mind, what you can do with it. And, you will, we keep youposted Julian, on what's what we will release there in the future, but it's,it's not. On risk prediction. I wouldn't give that, I wouldn't give that to toGPD four.  

Julian: Yeah. Yeah. It'salways, it is always good to, to keep some kind of healthy skepticism with,with any new technology and, and, and kind of realize where its capability isversus how much, you have to input in there and have to then analyze what isgiving back to you to, to really make, make sense of, not only the informationyou're receiving, but if, if whatever information you requested.

Was, in that sense accurate to, to whatyou, you wanted to get out of it. But Maik, it's been such a pleasure. I knowwe're over time here and, and it's been such a pleasure to chat about not onlyyour experience as a founder, but also what you're working on at, at Taktileand how honestly, this, this intelligent decision making based on data hasreally, really opened up rather than people think it, it, it limits anythingor, or goes in in one direction or another.

It kind of opens up the space in a lotof different ways. Not only. Giving say, lenders access to differentindividuals that they can service or testing hypothesis. It's so exciting tosee where this technology can expand and, and be a fan of, of where it will go.So last little bit is where can we find you as a founder?

Where can we find Taktile? Give us notonly your websites, but your LinkedIns, your Twitters. Where can we support youand the company that, that you run in?  

Maik: Julian, thank you somuch for having me. It, it's been great fun. Everyone reach out to us on, on,on LinkedIn or on, on our website, Taktile.com.

If there any questions, you please sendus a message. Message and I would love to, would love to continue theconversation I had today with Julian, with, with everyone out there withinterest in our space.  

Julian: Amazing. I hope youenjoyed yourself, Maik, and thank you so much for being on the podcast today.

Maik: Thank you.  

Other interesting podcasts