Building the Modern Tax Engine with Gavin Nachbar, Column Tax co-founder and Aiwyn CPO

If AI is about to transform tax filing, why do 55% of Americans still trust a human to do it for them?

In this episode, host Reggie Young sits down with Gavin Nachbar, founder of Column Tax and Chief Product Officer at Aiwyn, to unpack what it really takes to build in one of fintech’s most complex and overlooked spaces: tax filing.

Gavin shares how Column Tax built the first modern embedded tax engine in the US, what AI is about to change for the 55% who still rely on a human preparer, and why he’s genuinely bullish — not worried — about the road ahead. They also dig into his lessons on entering a regulated space as an outsider, the unique challenges of a seasonal product, and the hard-won takeaways from going through an acquisition.

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 Column Tax founder, Gavin Nachbar. Gavin founded Column Tax, a modern tax filing platform, which was acquired by Aiwyn in November 2025, where he's now Chief Product Officer.

I wanted to get Gavin on the podcast because as someone new to fintech, he successfully navigated tax filing, a very complicated regulatory space. He's got a lot of insights for founders and operators entering fintech for the first time, as well as lessons on the quirks of tax products and insights from going through an acquisition.

Fintech Layer Cake is powered by the card issuing platform Lithic. We provide payments infrastructure that enables companies to offer their own card programs. Nothing in this podcast should be construed as legal or financial advice. If you enjoy Fintech Layer Cake, please give it a review on Apple, Spotify, or wherever you listen.

Gavin, welcome to the podcast. Long time overdue. We've been trying to set this up for literally years now, so I'm excited to finally chat. You've been building and doing a bunch of exciting things. Maybe the best place to start for listeners is just, plain English, what did you build at Column Tax and what are you now building at Aiwyn? If you're at like a happy hour or something and somebody asks you, what do you do, how would you explain Column Tax at Aiwyn?

Gavin Nachbar:

I'm happy to be here. Long time coming. We founded Column five years ago and built the first modern tax engine in the US. And so if I'm describing it to people, it's we built a TurboTax-like product that banks, fintech companies, investment platforms can embed inside of their app and bundle and offer to their users. And we think that's a better way for tax to live inside of financial apps that people already use. So did that file about a million returns, super fun, sold. Column Tax was acquired by Aiwyn in November of last year, and now we're taking that same engine and applying it to the professional industry. In the US, there's 170 million individual tax filings a year. About 45% of those are the TurboTax-like DIY products, and 55% are actually still through humans. So we're focused more on the latter now, which has been very fun.

Reggie Young:

That's crazy. I would have guessed a much lower percent is still human filed, so big market to attack there.

Gavin Nachbar:

Yeah. It's interesting because in the early ‘90s, tax filing moved from everybody filed with papers to e-filing, filing online. And you saw a pretty linear increase from the early ‘90s to 2018 of do-it-yourself. So it was like another percent every year up to 45%. And then it stayed flat. It actually jumped up in COVID. Fewer people could go into the office to file, and then it stayed totally flat. One of these counterintuitive things we learned, it's actually fairly consistent across income levels. So low-income filers are as likely to file with a pro as median. And then higher incomes, it's almost all pros. But even at lower-income levels, a majority of Americans still file with a professional.

Reggie Young:

Interesting. Yeah, I wouldn't have guessed that.

Gavin Nachbar:

Taxes are stressful.

Reggie Young:

Yeah. Nobody likes talking about taxes except us. So we're here. One of the big headlines right now, AI, a lot of stuff going on in there. Tell me about AI and taxes. Speaking of the pain of taxes, I would love to tell a robot to file my taxes and then never have to think about them. I know Aiwyn just rolled out some exciting news about APIs and MCPs and what you all are building related to tax. So walk me through that a bit.

Gavin Nachbar:

Yeah. And this informed a lot of why we were so excited about joining with Aiwyn, the shared vision around infrastructure and how that will both help consumers and pros, oftentimes through the latter. Tax filing itself, it's three different subtasks. People talk about filing their taxes, and we break that down. The first part is you have to go gather all your documents. So you have to go find your 1099s and your W2. And it's very unique that in the US, that's an individual burden. So you're like, I got to search my email, and oh, I had that expense. So there's the gather part.

Then there's preparing that. So take your information and get it into some sort of software or get it to a person. Call that the prep part. And then there's the calculate and file. So there's actually like three distinct parts. AI/LLM-powered workflows have made the prep and calculation way faster. Last week, we launched Claude Connector for Aiwyn, where you can actually go drop all your tax docs into Claude with the Column now Aiwyn engine hooked up and actually do the whole prep part all in one, as if there's a person interviewing you as you go with the power of Claude there. It's really good. I go play with it. It's fun. It blows our mind that that combination works today.

And then AI/LLMs are very good at taking structured language and writing code. The tax law is structured language, that's formulaic, and so it's accelerated our team that writes the formulas quite a bit. So very helpful on the latter two tasks.

Reggie Young:

I love it. I will definitely be checking out Claude for my taxes next year. Love that potential. What are the impacts do you expect from a- I got to imagine, sitting from the product team vantage point, you're looking at AI and you're like, dear, God, there's so much we could be doing with this beyond just that Claude integration setup. What are the things you are excited about on the horizon for AI and taxes?

Gavin Nachbar:

Today, if you sit with a tax pro- we said 55% of Americans file with a pro. That's where the majority of time, effort, blood, sweat, tears goes. Actually, the human part of preparing taxes, much of that is not- when we talk with pros, it's a lot of time on things that they don't want to be doing and isn't adding value. And so it's like I'm reading a PDF and I'm typing in to a software UI. It's really painful and tie-outs and everything that goes with it. So I'm excited because I think a lot of that will become a lot easier, and that'll mean, A, pros can spend more time doing value-added things like things people actually want, and B, I think it'll increase accuracy.

There's an interesting study that I remember early on reading at Column. When you look at a sampling of different tax returns prepared for the same data, and this is done by people going to different tax prep places with the same data, you get many different results. I think it'll help increase consistency and accuracy when so much of it is done by machines that are better at the task than we humans are.

Reggie Young:

I love that. I've done some comparisons before of setting up my taxes and some of the online free tax filing things, and they spit out different. Some of the legacy, older ones spit out different answers. It's wild.

Gavin Nachbar:

There are many stories- it's so fun learning these when you get into a new sector. There's many well-known stories of companies that actually, within the industry, have multiple engines across different product areas. And so in theory, if you filed with one versus the other, you actually do get different answers because different people coded up the logic in different ways, and the tax code is so wide and sprawling and interpretive.

Reggie Young:

I love that. Yeah, I love your AI answer generally. I think there's a lot of headlines right now about this large company is laying off X percent because of AI. And it's actually they're just saying AI so that it helps their public market valuation when really they just over-hired or made bad choices and now we're paying the price. There's this incorrect narrative if you pay too much attention to the headlines, whereas the real narrative for AI is, let's take the work we hate doing and get rid of that so we can focus on the interesting, exciting work with our time instead.

Gavin Nachbar:

I've never met a tax professional, CPA, EA, tax pro, that when we say, this will make- this is a good example. We'll meet them- and it's really hard to tie out returns and review the tax calculations with the source docs. We've never met someone where we're like, hey, we can help you do that more efficiently. And they're like, they're worried about their job. That's not what they consider their job to be. They have so much value beyond that, and so they're like, heck, yeah, automate all the things I don't like. That sounds amazing. So I'm very bullish.

Reggie Young:

Even CPAs hate taxes. It's great. One of the reasons I wanted to get you on the podcast is because you came from self-driving cars at Waymo, and then you decided, hey, I'm going to go into this fintech thing, and not just any fintech but tax, which is an area of fintech that very few people say, let's dive headfirst into that. And you did it pretty successfully. I think you're a really interesting case study.

We see this sometimes. It's founders that come to launch a card program, and maybe they're new to fintech, and so you get, here's the five questions that you have to lead them through to help them learn the basics. And then you get other founders who come and they have that background. There's deep background in fintech of the corners to see around all that sort of stuff. So it's different models. A lot of innovation comes from founders that are new to a space, and it is a necessary thing we need. And I think you navigated that pretty well.

Maybe let's start focusing on building in tax. The IRS, like I was saying, is not a common agency that many fintechs grapple with. How did you approach that? I took a bunch of tax classes in law school. Loved them. Hard classes, but loved them. The tax code is this like giant Byzantine structure of logic trees with some abstract implications. How do you start as a founder new in tax and attack that problem of turning that into a product?

Gavin Nachbar:

There were two things that I think we did right. We tried to be very humble coming to it because we were new to both fintech and tax. And a lot of people have tried and a lot of really smart people have tried to do some of the things that we were doing. And so the first thing, there were just great people that we met. We tried to be historians. There were two or three companies that had tried building tax engines in the last 10 years. And Michael and I just obsessively met with the team members. We were like, tell us about it. What worked? What didn't work? And so it was just a lot of learning from what had been done.

Most people we met were like, this is impossible. You can't and shouldn't do it. We didn't take that conclusion, obviously. We spent five years working on it and still are. But it was being a historian, learning, sitting down, understanding, and not coming from a place of I know better, really take the lesson, and we learned a lot. We benefited a lot from that.

The second was, someone said this to me, I forget which person, but fintech founders were so generous to me early on, so grateful. Someone said to me, get to know your regulator. You're new to regulated products, you're going to get to know them. And I thought that was really good advice. We got to know the IRS and state agencies early on, found them to be incredibly helpful. There weren't many companies trying to build new e-file products, and they were helpful and helped us understand. There's great trade groups.

So we tried to get to know the industry. And then on actually doing it, we just built methodically. Michael, my co-founder, gets all the credit for this. I thought he did such a good job of, okay, we're going to start with just W2s and one credit. Okay, now we're going to expand. Now we're going to expand. And every time we expanded, the first year we did 10,000 filings- first year, we did 10 filings. Then we did a hundred. Then we did 10,000. Then we did a hundred thousand. So it was like every zero you added, we learned a lot. It was a combination of people and just doing it one piece at a time.

Reggie Young:

Yeah, I hadn't thought about this before you just framed it that way. Tax is such a wild product to build in because it's once a year. It's not a regular ongoing. You're getting product feedback. It's like, you got four months that your product’s basically in market and getting feedback and debugging or doing whatever and figure getting customer feedback, like all of that stuff. And so it's crazy from a founder strategic planning perspective to look at. Okay, what are we doing this year when four months of the year our product is actually being used? That's wild.

Gavin Nachbar:

I would not recommend it. It was not fun. We did it and continue to do it, but I think it's actually a huge downside. And when I talk with founders who are doing seasonal businesses- and this happens also in unemployment with enrollment. So there's a bunch of businesses that have seasonality. Intensely, retail also with Black Friday, less pronounced. Tax is so pronounced. So we would have zero people in our product one day, and then there's a barbell.

So you have about 40% of filings are going to come January 1st to February 15th, and that's just a rush. And people really want their refund. Generally in that demographic, if you file in January, you want your refund right away. Then it goes to a nice simmer in March, and then April 1st slowly, and then, bam, crazy way to build. The people that did infrastructure work at Column were incredible because they had to deal with zero usage and then a lot of usage really quick.

Reggie Young:

I'm thinking about in cards at Lithic, when you need basically perfect uptime or as close to perfect uptime is because if my own credit card doesn't work, I'm going to question using it again. But on April 13th, the reliability of somebody's taxes isn't working. I imagine those must have been crazy days.

Gavin Nachbar:

Yes.

Reggie Young:

What are some of the harder or thorny or product problems when working with tax filing? What might surprise the average fintech listener?

Gavin Nachbar:

This is probably folks that do consumer fintech. I've spent more of my career focused on B2B. So I learned a lot when we- Michael and I joke, we accidentally founded the consumer part of the company. You need a UI in order for people to file taxes. But that wasn't our original thought, which I could talk more about. The thing we learned is I actually think something that a lot of people learn eventually as they get into more complicated products, we had this idea that the more simple we could make it, the better.

That actually ended up not being true. It didn't help with conversion in the same way that we had hoped, meaning when people are going through a complex financial transaction, there's a desire to understand what's going on. And giving them the detail to understand what the credit they're getting is and why it's not the number they expect is just a trade-off with making it simple, and tax is so complicated that we always had this tension- our designers were amazing in figuring that out, but there was always this tension between simplifying and giving people enough information to understand what was happening in such a complex world. So we learned that that was really hard.

Reggie Young:

Yeah, interesting. Yeah, makes sense. Money is at somebody's personal bank account. I was just talking about their personal spending cards, they got to work. I imagine you want visibility. It's like, these are my taxes, and I have a liability if I don't understand them. So I should probably understand them. Makes sense.

Gavin Nachbar:

And there's so many permutations. An example that comes to mind, I think this is our first year. This is a funny story of when you build in your first season. We had to put a phone number down with the IRS. We put my cell, which was, in hindsight, a first-time founder move that I'll never do again. So we put my cell. Long story short, my cell ended up on the PDFs of all the returns that we were filing. And January came. If you went to go find the cell number on your PDF, you were the most upset user, or most upset filer. And so I started getting a lot of calls in the end of January. I was like, where is this coming from? But it was also amazing to just hear from people.

An example, got a call from a woman who, the previous year, she had gotten, I think, a $5,000 or $6,000 refund. And this year, she owed $5,000. She was sharing her story. She didn't have that money lying around, and she was sure that it was a calculation error. We went through, pulled up the data. She had twins, I think, who had just turned 19 and were no longer eligible for the child tax credit. And you're like, oof, that is really tough. One permutation that impacts this person a lot, but you don't want that in every person's UI. That doesn't impact most people, thus the permutations and tax make the experience so hard. This is why I think LLMs are so good for this task, because they can be personalized and tell that person about their situation. That's an example of where you're like, you want simplicity, but you want that person to understand why there was a $10,000 swing on their taxes.

Reggie Young:

Yeah, I love that. Good example. Giving her the context in one filing, you should be better armed the next year. Interesting.

We talked a little bit about this, but would love to drill in just more generally to your experience as a fintech founder and getting up to speed. I love your framing of be a historian. There's a lot of folks who are like, I'm going to start a neobank, and then they find out about all the neobank- I'm going to start a neobank for XYZ specific niche, and then they find out that multiple people, founders have tried that over the past few recent years and failed, and there's lessons you can learn there.

Are there other sort of tips? Because, again, I look at you as somebody who dove headfirst into a very quirky fintech area and navigated it well. Besides just being a historian, talking to other founders, were there other things you did that help to successfully get your feet under you in fintech?

Gavin Nachbar:

A lot of it was people. People plus reading, I guess is how I think about it. There's a really great set of people who write in fintech. I think this is probably more true in other industries, but I actually feel like it was more true- so I started working on this late 2020, early 2021. Fintech was pretty well documented. I'll give you an example.

Sophia, who founded Ansa, wrote a great book on payments. I knew nothing about payments before I met Sophia. I probably wouldn't have read that book if I didn't get to know her so well. And that was a great way to learn about payments in the US. Tax is sort of entangled with payments, not, depending on which part of the world you're in. But it helped me understand all the payments people that we were working with better. So it was like people plus writing.

There was also a wave- there were a handful of companies that were doing payroll APIs that were growing coming of age in a similar time as Column. I learned a ton from them. We partnered with a handful of those companies. And so it was like getting to know those companies. We co-sold with a lot of those folks. We partnered with them. But a lot of it was just getting to know people who were building in similar spaces, and I found fintech to be an incredibly friendly community. I was quite impressed.

Reggie Young:

Yeah, that's awesome. Yeah, it's a good community. Sophia's great. Folks should go check out her book if they haven't. When folks join Lithic, we typically give them Anatomy of the Swipe, and also Sophia's books just give some good crash course on the 101 of fintech, explains quite well.

Gavin Nachbar:

Yes, people and reading, that was definitely the recipe.

Reggie Young:

I love it. As you talked about Column Tax was acquired by Aiwyn at the end of 2025, you had a great LinkedIn post recently on some of your top lessons after going through an acquisition. Run me through some of those top reflections.

Gavin Nachbar:

It was really hard. That was the number one. Everything else is derivative of that one. M&A is really hard. It's totally worth it. We're so happy with getting to work with Justin and Levi and Pat and Tanner, the team, our founders at Aiwyn. Totally worth it but really hard. I think probably the big one was there's no such thing as nice-to-have M&A was my big takeaway. We worked with a really good advisor on the transaction who had this phrase, M&A gets done when there's a metaphorical hole in the office of the acquirer and they have to walk around it every day and you fill the hole. That's the only way you can get a transaction done. I thought about that a lot as we were going through it. It ended up being a really long and challenging process that was totally worth it. But there's no such thing as nice-to-have M&A was definitely my takeaway.

Reggie Young:

I love it. Yeah, It's not an easy process. I worked on a few product in prior roles, products being sold off. Not an easy process, a lot of hours, very intense, usually short period. I like the whole analogy. That's a good way to view, like, strategically will this make sense?

Gavin Nachbar:

Right. There's this interesting thing happening now where you had a wave of companies founded in 2020 through 2022, like pre-ChatGPT, more SaaS and a fintech company, fintech payments, et cetera. There's an inflection moment for a lot of those founders. It's hard to hit the growth that AI companies are doing today. It's the fastest growth in the history of humanity. It's an amazing time. It's like the coolest time to be building.

So you have to get on that wave, continue to raise venture. You have to find a way to get profitable so you control your own destiny or you have to land the plane. At least a lot of people I talk to are learning how hard that third step is. And so thus lots of decisions for founders coming up in the next couple years. And path two is attractive, but actually it's a trap in a lot of ways because you can do that forever if you get profitable and stay. I'm a big fan if you have an opportunity to join forces with a great company, which I felt like we did, taking advantage of that.

Reggie Young:

Yeah, love it. Got my standard wrap-up questions. I'm pretty sure you have some good thoughts on the first one, which is, what's something you're thinking about a lot nowadays that you think folks in fintech or tech generally aren't talking about enough?

Gavin Nachbar:

I sort of think this will come back around, but there was a really big wave of excitement around embedded fintech and financial companies, and it breaks out into a couple categories. There was embedded financial products for consumer. That's like inside of a neobank. Let's say there's vertical SaaS, so more opinionated verticalized embedded products, and then infrastructure, Lithic obviously being a critical one there.

What's interesting is I feel like there was a big wave in 2020 through 2022 simmered down. I actually think there's a bunch of learnings that we can take, I definitely have from building Column, that I think will be very applicable in the next five years as I think more companies will try to go with agentic and AI-first sort of experiences and as the cost of software goes down. I think it'll be a very interesting new wave of embedded opportunities.

Reggie Young:

Yeah, I think about that a lot, like AI driving the cost to write software down. Obviously, you still need good engineers who can see around corners and know the right questions and right prompts and all that sort of stuff, but it does naturally raise the question of what's your moat if somebody can build the software that you built pretty quickly. We're starting to see some interest in embedded finance perk up, and I think it's going to be even more so in the next 12 months. So it'll be an exciting time.

Gavin Nachbar:

I think so.

Reggie Young:

If folks want to go find out more about Aiwyn, where should they go?

Gavin Nachbar:

Two places. You can go to our website, aiwyn.ai, and then you can also go to Claude and connect the Aiwyn Tax Claude connector and prepare your own taxes with it. So multiple ways.

Reggie Young:

I was just going to say, unfortunately, I think this episode will be out after tax day, but everybody who's listening, I know you're going to go file your taxes with Claude next year.

Gavin Nachbar:

That's right. And you can go check- however you filed, whether it was with another DIY or a pro, go check what they did. Drop that in to Claude with the Aiwyn connector and give them a rating. It'll rate them 1 out of 10. So you can see. And if you see if you need to file an amendment, so not all done.

Reggie Young:

I love it. I know what I'm doing right after we finish recording. Cool. Awesome, Gavin. This has been a great conversation. I appreciate you sharing all your practical wisdom from building Column and now building at Aiwyn.

Gavin Nachbar:

So fun. I'm so glad we got to do it.