Modern AI-ledgers for Accounting and Finance Teams to Replace QuickBooks and Automate the Books with Jeff Seibert
In this episode of Future Finance, hosts Glenn Hopper and Paul Barnhurst are joined by Jeff Seibert, founder and CEO of Digits. Jeff shares his journey from creating his first app at the age of 13 to building Digits, a platform that’s changing the way businesses approach accounting. With a background in tech and a passion for solving real-world problems, Jeff explains how his frustration with outdated financial systems led to the creation of Digits, which now provides real-time financial data to businesses.
Jeff Seibert is the founder and CEO of Digits, an innovative accounting platform that helps businesses manage their finances with real-time insights. Before starting Digits, Jeff co-founded Crashlytics, a mobile performance analytics company that was acquired by Twitter in 2013. He went on to lead Twitter's consumer product team and later appeared in the Emmy Award-winning Netflix documentary The Social Dilemma. A self-taught programmer, Jeff has been a part of the startup world for years and has invested in over 100 companies, earning recognition as one of Insider's Top 100 Seed VCs.
In this episode, you will discover:
How Jeff’s passion for coding started at age 13.
The story behind founding Digits and solving slow financial systems.
How Digits provides real-time financial data to businesses.
Why traditional accounting systems fail and how Digits overcomes it.
Jeff’s vision for AI and automation in the future of accounting.
Jeff shared his journey from building his first app at 13 to leading Digits, offering an insightful look at how modern accounting platforms are reshaping finance operations. His perspective on real-time financial data, overcoming the limitations of traditional systems, and the role of AI in automating workflows provides valuable guidance for finance leaders navigating the future of finance.
Follow Jeff:
LinkedIn - https://www.linkedin.com/in/jseibert/
Website - https://jeffseibert.com/
Company - https://digits.com/
Join hosts Glenn and Paul as they unravel the complexities of AI in finance:
Follow Glenn:
LinkedIn: https://www.linkedin.com/in/gbhopperiii
Follow Paul:
LinkedIn - https://www.linkedin.com/in/thefpandaguy
Follow QFlow.AI:
Website - https://bit.ly/4fYK9vY
Future Finance is sponsored by QFlow.ai, the strategic finance platform solving the toughest part of planning and analysis: B2B revenue. Align sales, marketing, and finance, speed up decision-making, and lock in accountability with QFlow.ai.
Stay tuned for a deeper understanding of how AI is shaping the future of finance and what it means for businesses and individuals alike.
In Today’s Episode:
[02:25] - Jeff’s Background
[07:11] - Why Traditional Accounting Fails
[10:24] - Overcoming Data Challenges
[13:17] - The Long Road to Market
[17:42] - AI Agents in Real-Time Accounting
[21:55] - Overcoming AI Adoption Resistance
[23:20] - What’s Next for AI in Five Years
[25:41] - The Future of Accounting
[30:44] - Transition from Startup to Big Company
[33:56] - Personalized and Fun Questions for Jeff
Full Show Transcript:
[00:01:53] Host 1: Paul Barnhurst: Hello everyone! Welcome to another episode of Future Finance. I'm your co-host, Paul Barnhurst, and I have with me my partner in crime, Glenn Hopper. Glenn, how are you doing? Doing great. Good to see you. Glenn is ready and charged up. He's had his coffee, right?
[00:02:09] Host 2: Glenn Hopper: Yeah, yeah, 11 cups or something like that, I don't know.
[00:02:12] Host 1: Paul Barnhurst: And so we're really excited for our guest today, we have Jeff Seibert. Jeff is the CEO of Digits. So Jeff is welcome to the show.
[00:02:21] Guest: Jeff Seibert: Thank you so much for having me. It's great to be here.
[00:02:23] Host 1: Paul Barnhurst: Yeah, we're really excited to have you. So a little bit about Jeff's background. Jeff Seibert is the founder and CEO of Digits, the world's first AI-native accounting platform. He previously served as Twitter's Head of Consumer Product and starred in the Emmy Award-winning Netflix documentary The Social Dilemma. Jeff was co-founder and CEO of the mobile performance analytics company Crashlytics, which was acquired by Twitter in 2013. Now owned by Google, Crashlytics runs on over 6 billion monthly active users and is the market-leading crash reporter for both iOS and Android. A self-taught programmer, Seibert released his first app at the age of 13 and went on to graduate from Stanford with a B.S. in Computer Science. He's angel-invested in 100+ startups and was recently named one of Insider’s Top 100 Seed VCs. Jeff, thank you so much for joining us on the show. We're thrilled to have you.
[00:03:33] Guest: Jeff Seibert: Paul Glenn, it's great to be here. Really excited for this.
[00:03:36] Host 1: Paul Barnhurst: And before I jump into the first question I had, I have to ask, what was the app you created at the age of 13 and how did it do it?
[00:03:41] Guest: Jeff Seibert : Oh man, it did. All right. So as with most kids, I was playing computer games in the mid 90s and I actually got bored of one of these games. It was the Mac game Escape Velocity, and I started hacking the game, and that's literally what taught me how to program. So I ended up writing the editor for the game. And so you could have this graphical editor on your Mac and sort of change the game around. And I ended up getting a couple thousand users of it, which was pretty sweet. It was. It was completely free. I didn't make money off this, but it was really fun to prove that I could build something I wanted and ship it out there and like, actually get people using it.
[00:04:16] Host 2: Glenn Hopper: I was gonna say, kids these days are missing out on that kind of joy because they could just vibe code it in their 30s and not learn anything.
[00:04:22] Guest: Jeff Seibert : It is true. I mean, it took me a solid year plus to write. It was a massive project for that age.
[00:04:28] Host 1: Paul Barnhurst: Did you get a lot of help from parents, or did you just kind of figure out the coding all on your own? That's a young age to spend a full year on a project.
[00:04:34] Guest: Jeff Seibert: Yeah, no help at all. I later found out my mom had actually minored in computer science in college, but that was sort of punch card programming. It was not relevant to what I was doing, and I had to figure out everything by myself. I was just like, came home from school, rushed through my homework and just slammed my head against the wall, trying to understand the code every day for a year. But the benefit is doing that. You really understand every single detail of how it works. So I actually think in retrospect it was a huge advantage for me.
[00:05:04] Host 2: Glenn Hopper : I'm going to completely date myself here. But a similar venture when I was let's see, this would have been around 1980. I would have been eight years old. I had a Commodore Vic 20 because we couldn't afford the Commodore 64 and I would do the same. I was trying to write a Dungeons and Dragons game that I thought was going to have artificial intelligence, but same thing. Rushed through your homework and then just pounding away at code, having no idea what you're doing. I had all the magazines, you know, pre-internet days and I'm, I'm flipping through trying to find code I can steal from other projects.
[00:05:36] Guest: Jeff Seibert : Yep, yep, I love it. I mean, I'm truly in love with it. I was coding last night. It's still my favorite activity on earth. It is just so cool to be able to build something that you write once, and it can run a billion times.
[00:05:48] Host 2: Glenn Hopper : Are you leaning into the vibe coding and using the Code Assistant tools I am.
[00:05:52] Guest: Jeff Seibert : So that's what I was playing with last night. I didn't actually write any code myself by hand. I'm just telling Claude what to do, cursor what to do, etc. and seeing how they go. And they've gotten way better just in the past 3 or 4 months. It was pretty bad at the start of this year. It's now it's decent. It's very interesting.
[00:06:10] Host 1: Paul Barnhurst: Kind of. Speaking of vibe coding, I chatted with the guy the other day. I messaged back and forth. I'm going to meet with them in a few weeks. He's trying to create the equivalent of cursor in Excel.
[00:06:19] Guest: Jeff Seibert : Oh, interesting. All right. I want to hear more about that.
[00:06:22] Host 1: Paul Barnhurst: Yeah. So I'll be really curious when I meet with them and see what they've done. And I saw one yesterday that's from a research lab that's talking with the guy that's done some pretty amazing things where they built their own spreadsheet that saves the files in Excel, but there are certain things you'd want to do in their spreadsheet. Then you can import it back. And it's pretty amazing what we're seeing. It's just like, I never thought it would be that simple for some of these apps that I'm like, how long did it take me to figure out how to write that stupid formula?
[00:06:50] Guest: Jeff Seibert : Right? Exactly.
[00:06:52] Host 1: Paul Barnhurst: All right, well, we'll go ahead and jump into some of our questions here. I digress. So, you know, Glenn, it's your turn. You have another area you want to digress into before we get to the question.
[00:07:01] Host 2: Glenn Hopper : Well, I do, but you keep reminding me of the timeline. So I'm going to sit quietly and drink my coffee while you, uh, ask question.
[00:07:08] Host 1: Paul Barnhurst: Now, go for your question. We're all good.
[00:07:11] Host 2: Glenn Hopper : No, I think I think we're fine to move on because we came to talk about we came to talk about digits. So we can move on to that. We can.
[00:07:17] Host 1: Paul Barnhurst: Did. So we have a brave man here. Jeff entered the general ledger accounting software space. So I have to ask what led you to decide to kind of tackle or start a company in this space?
[00:07:29] Guest: Jeff Seibert : Yes. And so keep in mind I have no formal background in finance or accounting. When I started this, um, so digits is my third venture back company. I started a document collaboration tool back in 2000 and 708. We got acquired by box and ended up displaying, uh, powering all document displays on box for the next five plus years. Our core IP was how to render files in a web browser, which today sounds dumb, but back then was actually pretty important. Then I started a mobile performance analytics company, Crashlytics. You said that in the intro got acquired by Twitter. I was then named head of product at Twitter in 2015. And so I saw both the building of the company phase with Crashlytics and my previous one, and then also the big company phase where Twitter had thousands of people. And I was so frustrated by the finances because in my startup experience, you have like, look at the difference between product engineering, which we were just talking about, and finance. In product engineering, you have B testing tools, Google Analytics, live performance dashboards, server logs. I can see exactly who is on our site, what they're doing at any given moment, and it's all real time. And on the finance side, I'm sitting there as the founder, waiting 2 to 3 weeks for our accountant to give me a black and white PNL and balance sheet. And it's like that.
[00:08:50] Host 1: Paul Barnhurst: I.
[00:08:50] Guest: Jeff Seibert : Get.
[00:08:50] Host 1: Paul Barnhurst: That.
[00:08:50] Guest: Jeff Seibert : The month's. The next month is almost over, right? Like, people are just getting their June numbers today. I don't care about June anymore. That ship sailed. I'm now focused on August. And I was like, why? Why is finance not real time if user analytics are real time? And at first I thought it was just a small company thing. It wasn't. Twitter had the same issue. I went to corporate finance and I was like, I want to run an event for our team. What's our Q1 budget? And their literal response was, oh, we haven't run those books yet. Give us three weeks. And I was like, you have 100.
[00:09:23] Host 1: Paul Barnhurst: People in.
[00:09:24] Guest: Jeff Seibert : Corporate finance. What are you doing? And so it was literally that naivete, but sort of obsession with it that caused me to start digits in 2018, and we fell all the way down the rabbit hole again. I had no background in finance or accounting. I read two accounting textbooks cover to cover. We hired UCLA's professor of introductory accounting to give our team a private class. We tried to get up to speed as quickly as possible and we realized it was a software problem. It's not the accountants or the finance folks, it was that the fundamental ledgers, the software, were just not built for the modern era of high frequency business, high volume transactions data coming in from all these different sources. And it was such a pain, it created so much tedium. So that was the founding premise for digits: let's build a new modern GL. Back in 2018 we believed we would be ML native, machine learning native. Obviously the whole world has changed, it's now called AI, etc. etc. but that was the premise from the beginning.
[00:10:24] Host 2: Glenn Hopper : When you were thinking about machine learning there, what were you? What were you thinking about using to train the algorithms at that point? If you had, you didn't have the data at that point and you saw the vision with it, but you didn't have the data at that point. So I'm wondering how you went about that.
[00:10:42] Guest: Jeff Seibert : Yes, that was the key first challenge. And so what we ended up doing was we released an expenses dashboard in 2020. Right as Covid hit, all these small businesses were panicking about like, what would this do to their business? And so on. All of a sudden, everyone started really paying attention to their accounting and finance. And so we built a product that sat on top of QuickBooks and visualized your spend and launched this for free. And we actually got 17,000 people on the waitlist. We ended up getting thousands of people using it. And that gave us the initial sort of transaction data set that we could start really training models on. Good.
[00:11:21] Host 1: Paul Barnhurst: A good idea to do that. Good way to get that data.
[00:11:23] Host 2: Glenn Hopper : It always surprised me that the incumbents but you know, the net suites and the QuickBooks of the world, they just aren't, they don't have that startup ability to pivot and do something like this. But they're sitting on so much data. And I've always thought for years someone is going to come along because if you're a small business and you're using QuickBooks and you've got a dry cleaning business with three locations. You don't have big data, you don't have anything to train on. But if somebody could do what, like what you did and then train the algorithms on it, then I'm kind of just surprised the incumbents haven't done it. But I applaud you for seeing the ability and reason to do that.
[00:12:02] Guest: Jeff Seibert : So they have an architectural challenge. This is the real thing we changed is you have to look at the atoms of the data set you're dealing with. And in finance the atoms are the transactions. But if you look at QuickBooks, Xero, NetSuite, Oracle, all of them, these are relational databases. It's a relatively outdated architecture. They are holding these transactions as just text strings. And so when you say take an Uber ride, it just says Uber trip, blah, blah blah. Is the code right? None of those ledgers know what Uber is. They're just seeing this as text. And you type in the transaction into QuickBooks, you close your browser, log out. Nothing's happening with that data right until you come back. Log back in, run a report, and then it prints out what you typed in. And so that was the fundamental insight is we had to change the atoms of the architecture. Digits is not a relational database. We built our own data store. Digits is a semantic object graph. So it's basically every vendor customer, bill, invoice, your chart of accounts, etc. these are objects in our system. And the transactions connect those objects. That allows us to build this sort of semantic understanding of your business. That's how the models get so good.
[00:13:15] Host 1: Paul Barnhurst: That's cool. That makes a lot of sense. I'm curious, when you started, do you think it would take as long as it did to get to market? That's a no. By laughing.
[00:13:24] Guest: Jeff Seibert : Yeah. Yes and no. Um, so we did. We knew that building ledgers was an incredibly long journey. I mean, my literal first product doc I ever wrote when we started the company said, this will be a many year journey. So we knew that, uh, we did fund the company. For that, we ended up raising 100 million pre-launch revenue. It did take probably a year, two years longer than I think we all would have hoped. Um, of course, as with any project, the first two years are great and you're so excited you're three and four. We're challenging because it's like we're way in it, but we're nowhere close to the end. And then year five, we sort of started seeing the light at the end of the tunnel, and it came together and we started serving customers and a beta product on it and so on. Um, so yes, it was a long journey. It's just hard to build a ledger. If you look back at the history of accounting software, it's taken these teams a long time and massive effort to build just this complex of a piece of software.
[00:14:19] Host 1: Paul Barnhurst: Yeah, kind of an example of that. I was talking to the founder of one of these other, you know, ERPs out there, and they've told all their investors, look, it's at least a ten year journey here. You know, this is not a quick turnaround, quick sell, quick exit, if that's what you want. Go invest elsewhere. You got to be in it for the long haul. And there are some huge deep pockets. Established players.
[00:14:41] Guest: Jeff Seibert : 100%. And the advantage is if you can crack into it. It's a truly massive market and it's a very sticky market, right? Quickbooks has held 80% market share since 1992. And so if you can break in, you have a great business.
[00:14:57] Host 1: Paul Barnhurst: Yeah, it's a lot like the CRM market in the sense of there's very much consolidation, right? If you talk about small businesses, QuickBooks and zero are 90% of the market. And then there's everybody else. And you got a similar thing in the mid-market. It's. Net suite. And everybody else you get to the big it's Oracle SAP and the rest is relatively small. I mean dynamic sage I mean there's others. But yeah they're pretty sticky and pretty dominant players and all all across the market sizes.
[00:15:25] Guest: Jeff Seibert : Exactly right. Yeah.
[00:15:27] Host 1: Paul Barnhurst: So I have one other question. And I know the answer to this, but love just for you to share kind of with our audience. You really leaned into machine learning for classifying, you know, for booking the journal entries. Why not use some of what we're seeing in gen AI? Kind of, you know, technically, what's the reasoning behind that approach?
[00:15:45] Guest: Jeff Seibert : So we do use some gen AI, some LMS, but the way to think about this is accounting is fundamentally not generative. You do not want an LLM hallucinating your transactions, your chart of accounts, the numbers. Remember, these LLMS are just fundamentally not good at math because they are eager to generate an answer. They are not computing an answer. And so what we do, and what we sort of have been pioneering over the past five years is custom domain specific predictive bookkeeping models. So these train both on your individual books and across the entire set of businesses on digits to really understand how a transaction is booked. And that allows us basically to say, hey, if your business has seen this transaction before, we understand how you booked it, we will mimic that. And it's effectively 100% perfect. These predictive models cannot be hallucinated by the architecture that you have. It's not possible if your business hasn't seen the transaction before. We fall back to our global models that look at all types of businesses and how they book things, and then basically translate that back into your business. And this is the power of the semantic knowledge graph. Those are also extremely highly accurate. Then you have this third tier where let's imagine a new business just opened up down the street. No one's ever been there before. No one's heard of it. What would your bookkeeper do? Right. Well, your bookkeeper would probably start googling it and try to figure out what it is. We have AI agents that are powered by LLMS that do exactly that. The agents go and Google the transactions, and we use them for zero shot classification. So there's no training data there. They make the best guess they can based on all the information they can collect about that transaction. And then once you as the business owner or the accountant, go and close the books. Confirm that. Boom! It gets baked into the predictive models immediately. And so then every next one is perfect.
[00:17:42] Host 2: Glenn Hopper : I was about to say that with that, with the agents going out and working and knowing kind of the state of where agents are right now, that human in the loop is so important. But it's, uh, yeah, that's that's great though. And it's so how like in preliminary looking at that, how are the agents doing it at their classification.
[00:17:59] Guest: Jeff Seibert : Yeah. So we've run two sets of benchmarks over the course of the spring across a large data set, almost 18,000 transactions, a hundred different businesses, and we're now up to about 98% accurate. Um, so it is incredibly, highly accurate. And we actually our latest benchmark was against human outsourced accountants. So we took our agents and our data set. And then we went and hired 12 outsourced accountants and gave them the transactions, gave them the chart of accounts and asked them to book the data. And what shocked us is they got 79% right. So call it 80% right, 20% wrong. Lems. If you just give it to ChatGPT, they're at about 70%. So 10% worse. We're up at 98% y. And what the data seems to be saying is this is the challenge of having an accountant or bookkeeper who does not know your business. And this is the challenge with a lot of the outsourcing that's happening and so on. There's a bunch of quality issues because they may know accounting, but they don't know that industry, that business, their patterns, etc.. So it becomes very hard to book the transactions. That's how you should think about LMS. These foundation models train across all the accounting knowledge of the entire internet. They know nothing about your individual business. And so the key thing we unlocked with our predictive models is they know and memorize your individual business. That's why it can be so accurate.
[00:19:25] Host 1: Paul Barnhurst: Ever feel like you go to market teams and finance speak different languages? This misalignment is a breeding ground for failure in pairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet QFlow.ai, the strategic finance platform purpose-built to solve the toughest part of planning and analysis of B2B revenue. Q flow quickly integrates key data from your go-to-market stack and accounting platform, then handles all the data prep and normalization. Under the hood, it automatically assembles your go-to-market stats, makes segmented scenario planning a breeze, and closes the planning loop. Create air-tight alignment, improve decision latency, and ensure accountability across the team.
[00:20:33] Host 2: Glenn Hopper : Paul and I talked to software firms all the time. New software companies that are doing amazing things that you just couldn't do. Like I was saying with the incumbents and they just their whole data structure and the software and just where when you're that big too, you're, you're nervous about pivoting and doing anything because if, if you know your QuickBooks is hallucinating, your whatever, how many millions of customers they have are going to lose their minds. So they don't, they're just they're overly cautious at that point. And I'm so we see all these new technologies. And then I talk to everyone from enterprise level companies down to, you know, a lot of private equity backed companies. And they're everybody's got this. There's a FOMO of not jumping on the AI wave and getting left behind. But more often than not, I'm saying it's um, it's that fear is overwhelmed by their fear of the technology itself and their lack of understanding. And I think you can. We've started in some of our talks, uh, telling uh, you know, leading with let's go ahead and write your company's obituary. What would kill you if you didn't, uh, if you didn't survive the next 5 to 7 years and just tried to get companies to think, like, I understand we're finance and accounting people. We're risk averse by nature. But at the same time, there's this tsunami of technology, and it just seems like the adoption is slow. What are you saying? And how would you advise those companies that are. I don't know, this seems untested. And you know what? What are you seeing in the marketplace and how are you guys countering that?
[00:22:05] Guest: Jeff Seibert : Yeah, the big analog is the cloud transition. So if you look back to 2010, right, so many folks are actually just coming out of it because it took the profession a decade, sometimes 15 years to go through that cloud transition. But could you imagine today doing everything only on desktop software with no collaboration? Right. It's insane.
[00:22:25] Host 1: Paul Barnhurst: And so here.
[00:22:27] Guest: Jeff Seibert : Like you're not. You're not moving your data. There's no more huge data migration. It's more just are you willing to adopt these new AI enabled tools? And I totally understand that folks are nervous. They have questions on what it is going to do. But I think if you put yourself five years out from now, can you picture yourself not using anything AI powered? I just don't think you're going to have a successful business. And so.
[00:22:50] Host 1: Paul Barnhurst: What.
[00:22:50] Guest: Jeff Seibert : We're seeing it like obviously there's a bell curve of adoption, speed and so on, but there are more than enough folks who I think see that realization. And they are aggressively trying these new tools.
[00:23:01] Host 2: Glenn Hopper : As fast as everything's going. You say five years out, I think five years is, you know, five AI years, you know, they're faster than dog years. I mean, what do you think? Uh, so if people are you know, I'm not saying jump on every bleeding edge product, but having a plan trialing, you know, jumping into this. But I mean, can you like, since you're very close to this and you've been developing in space forever, what's your. I guess I'm asking you to get out your crystal ball here. But knowing the piece of technology, what do you see in five years? Do we have AGI? Is it? Is it just more reliable agents? Less hallucination? Where are you? Where are we headed?
[00:23:37] Guest: Jeff Seibert : I think we'll definitely have something close to AGI. It won't be a singular moment where it's like, oh, that was AGI. It's just going to be a progressive increase in the quality of these models and agents. And the analog I would draw, because in the finance world, in the accounting world, there's relatively few companies training. These models were among them, but it's a relatively niche domain specific industry. If you look at coding, there are so many companies working on coding models, it's incredible. And even six months ago they were not that good. And I again, I've been coding for almost 30 years now, like my life is coding. I was pretty skeptical, particularly last year. I was very skeptical because I was like, the output is terrible today it's decent And like for that change to happen in six months is actually sort of mind blowing. And I can only imagine where we'll be at the end of this year. And so if you take that velocity and then apply it to finance and accounting, I think in the next couple of years, bookkeeping will be basically solved perfectly. And then there'll be companies progressing into the more advanced stuff, like how do you auto manage accruals and schedules, how do you basically automatically reconcile everything in real time? It's really going to start pushing the bounds on what's possible.
[00:24:56] Host 2: Glenn Hopper : What's really interesting with accounting. I mean, it's complex, but it's inherently rule based. And so there's, you know, gen AI that you can use for, you know, providing insights and commentary and everything. But I mean, everything that you're addressing with, uh, with your software is it's not guessing. We know, you know, we know how to recognize revenue. We know, you know, how to run depreciation and all that. So it does. And there's so many people who conflate uh AI and automation and agents. And there's a lot of confusion around there. But I think leaning back to what you started with earlier of the deterministic machine learning models and just rule based stuff that's going on in there and having the human in the loop, I mean, it does. It just feels like the future of accounting.
[00:25:46] Guest: Jeff Seibert : Yeah. The mindset shift is right now, accounting firms across the country, across the world, end of every month. They're like, okay, it's time to close the books, right? And like, let's gather all the data. Let's make sure the bank feeds are updated. Let's start reconciling. Like you're the human is guiding that process. That era, I believe, is quickly coming to an end. You're going to have agents say the books are closing. Here is what I need from you. And this is how we've architected digits. We have an inbox where they just surface things that they can't decide on. And so if there's a transaction that's low confidence, they're like, yeah, we could book this a couple different ways. We throw that out because you don't want the agents making a mistake. We just surface that for human review. And so I think finance is going to turn from a sort of doing, doing all the tedium mindset to let me sit back and review everything and approve it. And great, the books are done, I can move on to the analysis. These models are also really good at analysis, as long as you stop them from doing math. If you give them the tools like we built a calculator. It sounds dumb, but you build a calculator, give it to the agent, you build a financial modeling engine, give it to the agent, then they're really good at analysis too.
[00:26:57] Host 1: Paul Barnhurst: Crazy times for sure. Very exciting though.
[00:27:00] Host 2: Glenn Hopper : Well, Paul, uh, we're coming up on time, but I did want you.
[00:27:05] Host 1: Paul Barnhurst: To ask your favorite question, though.
[00:27:07] Host 2: Glenn Hopper : And. Yeah, um, I do, uh, I can't remember if it was before the show, if we were talking about it on air. But I do love talking to finance and accounting and tech folks who also kind of lean into the creative side. And I know you had the involvement with, The Social Dilemma documentary, but you also were an associate producer of a documentary called Chasing Coral, and I'm just it was also picked up by Netflix. I'd love to hear about that. And what got you interested in how that was? Just part of sort of in your investing mode of that, or was there something about the film that drew you to it?
[00:27:45] Guest: Jeff Seibert : Yeah. So one of my passions outside tech and coding has been climate change for a long, long time. Growing up, before I got into computers, I wanted to be a scientist and just, like, study nature and so on. Um, and so I actually got involved with another film first called Chasing Ice that was picked up by National Geographic and was an associate producer on that and then helped make Chasing Coral and then The Social Dilemma. The interesting story, what many people don't know is this is the same director producer. He's a documentary filmmaker who's really focused on climate change films. And so Chasing Ice was about the glaciers melting. Chasing coral was about, sadly, the Great Barrier Reef dying and being declared dead by science. And The Social Dilemma started as a climate change film. Because what he realized is what he was like, I'd made these documentaries. How is there still so much misinformation about climate change? Like, we showed you the footage of it dying, and he started researching that and realized it's actually social media that was to blame. It was spreading misinformation and mis-educating people about the truth. And he's fallen down that rabbit hole and started looking into the algorithms and the motivations of these social platforms. And that became the social dilemma, which then did, in its final form, lose any sort of climate change aspect to it. But like that the bigger issue is that unfortunately, these social platforms are polarizing people due to their very nature of how advertising based, user generated content businesses work.
[00:29:15] Host 2: Glenn Hopper : Fascinating. I did not know that backstory, but it's I, you know, with documentaries, I'm sure the director. You start with 4 or 5, 600 hours of coverage and you have to find the narrative in there. So it's interesting that the component was lost. And I wonder if it seems like the director would think, man, that was my biggest movie, biggest platform. I sure would have liked to have gotten some climate change in there, but it ended up being so important. Yeah.
[00:29:41] Guest: Jeff Seibert : Ultimately, The Social Dilemma was a massive success. Over 100 million families watched it. It spawned a bunch of congressional hearings. The protagonist, Tristan Harris, testified before Congress on a number of occasions. So it did create a lot of motion, which was great.
[00:29:57] Host 1: Paul Barnhurst: That's fabulous. If you know, sadly enough, I have not seen that one. I don't have Netflix. And so I'm going to have to watch it now. I'm going to have to go to someone's house who has Netflix because I've heard it's good. So now you're making me watch it. Knowing that backstory even more.
[00:30:11] Host 2: Glenn Hopper : Paul, I'll share my log in with you.
[00:30:13] Host 1: Paul Barnhurst: Oh, Here we go. Perfect.
[00:30:15] Host 1: Paul Barnhurst: Yeah. Put that out there. Uh, Netflix. All right, so I have one more question I want to ask. And then we're going to move into our AI generated question section here. So you previously sold two businesses to much larger organizations. Can you talk a little bit about that transition? Right. You went from the small company startup to Twitter, as you mentioned, thousands of people. You talked a little bit about some of those frustrations, I think a little bit before and a little bit on air. But would you just, you know, love to learn. What was that transition like for you going through that twice from your baby, so to speak, to now I'm part of a I'm a cog in a bigger machine.
[00:30:53] Guest: Jeff Seibert : Right? This is such a great question because so many founders only focus on that moment of exit. It's like, oh, I was acquired. And they don't really appreciate that they're going to spend the next number of years with their team or not at this larger company. And every single deal is so different. Like, does your team get blown up and you're scattered across the wind. Do you stay building your product? Do you get reassigned to work on something else? So my advice for founders is you need to be really crisp when you're talking about an acquisition on what's the strategic goal of the acquisition, what will you really be working on after it closes, and what does that look like, and who will you report to and what sort of position in that company will you have? Because the truth is, you're right. You're giving up control of your baby. But more importantly, your team actively did not go work at a big company, right? Like your team likely joined a startup because they wanted a startup experience, and now you're joining a bigger company, and you need to help transition all of these folks into this new role and environment that likely has more meetings and more overhead and probably more drama, just because it's bigger and a lot of stuff's going on.
[00:32:01] Guest: Jeff Seibert : And so how do you navigate that? So for both of our acquisitions, the way I did it is I tried to keep everyone very focused for some initial time period, and that time period varies based on the deal. But with Twitter, we basically negotiated that, hey, for the first six months minimum, nothing would change. We would keep our office, keep the team, keep the product focus, keep our roadmap, etc. and that would allow folks to sort of integrate slowly and learn the dynamics of Twitter and learn what was important. And then we would discuss, okay, how do we augment our roadmap and sort of evolve. And that actually worked very, very well. And we actually had folks from Twitter apply to join our team. It was just an internal transfer. We were owned by Twitter, but they wanted to come over and have sort of a mini startup experience within Twitter, but then also help share the knowledge and the culture and so on. And that was really effective at sort of bridging that gap. So my, my biggest summary would be like, be intentional about it. Don't just sort of rip the Band-Aid off day one, have a plan to integrate. And that'll be far more successful.
[00:33:07] Host 1: Paul Barnhurst: Yeah I've I've seen some that the plans have been terrible and it was bad on both sides. I was not part of one deal at a company I worked for, and I ended up being the finance person. Got to oversee the company after the deal had happened, and they decided to make 90% of the value in the earn out, but they wanted everybody to work together. We didn't, you know, you didn't align incentives.
[00:33:30] Guest: Jeff Seibert : Here.
[00:33:30] Host 1: Paul Barnhurst: Right? It was just a constant fight. It was not fun at all.
[00:33:34] Guest: Jeff Seibert : Oh yeah. That is tough. Yeah I unfortunately I mean Twitter while I was there for four years, bought 50 companies in that time period. And I would say just overall in tech, something like 95% of acquisitions are a complete failure. Um, and most at Twitter I would say were a total failure. But there were a couple that did make a big impact on the business.
[00:33:55] Host 1: Paul Barnhurst: Not surprised to hear you say that. All right. So this next section, we took your bio, the questions we had about your LinkedIn profile. Let it go out on the web. So I have some really interesting questions. It's like I said, come up with some personalized, fun and slightly quirky questions. Yeah, there's.
[00:34:13] Host 2: Glenn Hopper : Some weird ones in here this week.
[00:34:14] Host 1: Paul Barnhurst: Those are some of the weirdest ones I've got yet.
[00:34:17] Host 1: Paul Barnhurst: Oh, boy.
[00:34:17] Host 1: Paul Barnhurst: Usually they're pretty good. So we'll see how this is placed. Here's how I ask my question. You get two options. You can use the random number generator. And it will pick a number between 1 and 25. Or you can pick the number between 1 and 25. And I haven't even read these questions. I've only looked at 1 or 2 of them, so I don't even know what's here on the lotto.
[00:34:34] Host 2: Glenn Hopper : I just saw their weird ball there.
[00:34:36] Guest: Jeff Seibert : I'll call out some numbers. Let's go.
[00:34:38] Host 1: Paul Barnhurst: Eight, eight. All right. It says I don't know what I found. This gives a quote of Lincoln free play. Not sure what that is. So I'm going to read the question and see if you know it says you use Kotlin and not Python. Do you secretly love surprise tech stack parties?
[00:34:59] Guest: Jeff Seibert : Oh, this is such a great question.
[00:35:03] Host 1: Paul Barnhurst: So this is true.
[00:35:05] Guest: Jeff Seibert : So at Digits we made the decision at the beginning that we had to be type safe, end to end. We're building a financial product. You cannot risk having type conversion errors. And I mean, NASA had these issues right, converting between different units and so on. Total disaster. So we chose a type safe tech stack where it is TypeScript in the browser, which is very common these days. Our web tier is Golang, Google's language. This is still relatively uncommon outside of Google but is a super powerful language. And then our back end started as Java. But Java is sort of clunky and we modernized it to Kotlin, which is the Android language. Everyone writing Android apps uses Kotlin. We run it on the servers and it's been great. And so it gives us type safety, end to end, and a really modern tech stack that our engineers have loved.
[00:35:51] Host 1: Paul Barnhurst: Got it. It did good on that question. I wasn't familiar with Kotlin, so I'm reading it. Let's see what he answers to this one. All right. Glenn gets to ask if he does it a little differently.
[00:36:02] Host 2: Glenn Hopper : Yeah. So just since I created them I have a I output one and we were talking about humans in the loop earlier and some of these questions seemed weird to me, so I did. I've got a human in the loop thing here, but I just ran, so it's supposed to give me a random one. I said give me a random one. And which one you think is the best overall question I'm going to ask you, do you want the random one, or do you want what AI thinks is the best question out of the bunch?
[00:36:29] Guest: Jeff Seibert : I mean, we have to go with the best one, right?
[00:36:31] Host 2: Glenn Hopper : All right, let's see. Let's see. Best overall question according to ChatGPT, because it taps into your leadership style and company culture. It invites a story that blends tech, personality and creativity, and it's likely to yield a memorable, entertaining response that highlights digits' internal magic. So the pressure's on right now. It says, you said never tell an engineer they can't build your product. What was the best unofficial engineer challenge digits ever inspired?
[00:37:02] Guest: Jeff Seibert : Oh wow. Owl. That's really interesting, so I gave that advice in the context of crafts Linux, my prior business, which was a developer tool. And the mistake a lot of dev tools companies make is they're like, we're so good, we built you this dev tool. You could never build this horrible marketing strategy, right? It's just a challenge to their customer base. And all these engineers are like, ah, I could build that and then they don't use it. Um, so that was the context of the advice digits. I would say one of our, our most successful moments was we, I talked about the architecture earlier, basically relational database versus this semantic knowledge graph. We did indeed start the company with a relational database, because you need to use something to get up and running. And about a year or two in it was just grinding to a halt. It was so slow because we wanted to do this complex data analysis, and you're doing like 14 different joins across all these tables to try to load a single dashboard page. The dashboard was taking like five 10s to load. Slope. And so we. And we're also a fully remote side story but so we decided okay we need to bring some of our top engineers together and figure this out. And so we all gathered at one of our engineers houses in Boston and literally whiteboard it on his kids toy whiteboard in his backyard for like.
[00:38:21] Host 1: Paul Barnhurst: 4.
[00:38:21] Guest: Jeff Seibert : Or 5 hours and just went through the problem and all of the challenges and all the potential data architectures. And literally at that moment in that backyard, we came up with the actual data model we used today, and we flew everyone back home. We built it over the next couple of weeks, and it worked. And I think just the power of getting everyone together so focused on a problem and having really experienced engineers who have seen a bunch of scale across the internet, these folks worked at Twitter and Google and so on, was able to give us an insight on how to build it. That was just now, I think one of my proudest moments, like the crowning technical achievement of my career, has been working with that team to make that.
[00:39:00] Host 1: Paul Barnhurst: Super.
[00:39:00] Host 2: Glenn Hopper : Cool of it and the power of whiteboards, right? Just having everybody together and being able to see the blue sky. Whiteboard and just map it all out and. Yeah. Love it, love it.
[00:39:09] Host 1: Paul Barnhurst: You know, even though these were different questions, it turned out pretty good.
[00:39:12] Host 1: Paul Barnhurst: Yeah.
[00:39:13] Host 2: Glenn Hopper : We're giving Jeff all the credit for that. We're not giving AI the credit for that.
[00:39:17] Host 1: Paul Barnhurst: I was thinking the same. Jeff, the great interview. I know I'm going like this every week for AI. I mean, we'll show you 1 or 2 others for fun and then we'll let you go because I think you'll find them kind of interesting. Here's one. You've intentionally never used TikTok, so if its digits had a theme song, what would it be? That was one that was on here. Let's give you one more just so you can see the questions. You dropped out of stealth mode with digits after five years, did you ever feel like you'd get arrested for too much stealth?
[00:39:49] Host 2: Glenn Hopper : Bad when they try to be. Funny.
[00:39:50] Host 1: Paul Barnhurst: If you missed a little bit. But you. I know I love it, I.
[00:39:54] Guest: Jeff Seibert : Love it. I mean, quick tidbit on that. In general, I do think stealth mode is a mistake for the vast majority of startups and founders because you're not getting advice, you're not benefiting from the folks you talk to and getting their input and context for very specific businesses like digits. There was no point in talking about what we were doing because we were building a GL and it was still three years out. And so it's just like we have to stay focused on that journey and stay heads down. And the time we would spend talking with external people was just a distraction from finishing the core features, I believe it.
[00:40:26] Host 1: Paul Barnhurst: All right. Well, yo, thank you so much for joining us. It's been a real pleasure. We've loved having you on the show.
[00:40:32] Guest: Jeff Seibert: This was awesome. I really appreciate it. Thank you both.
[00:40:35] Host 1: Paul Barnhurst: Thanks for listening to the Future Finance Show. And thanks to our sponsor, QFlow.ai. If you enjoyed this episode, please leave a rating and review on your podcast platform of choice, and may your robot overlords be with you.