Freedom From Mundane Tasks and Remembering the Commodore 64 with Tipalti’s Chen Amit

Episode Overview

Chen Amit is a seasoned entrepreneur and the visionary co-founder and CEO of Tipalti, a leading global payables automation platform. With a robust background in technology and business, Chen holds a B.Sc. in Computer Engineering from the Technion-Israel Institute of Technology and an MBA from INSEAD. Before Tipalti, he gained significant experience as the CEO of Atrica, a telecommunications company that was later acquired by Nokia-Siemens, and as a General Manager at ECI Telecom. His extensive expertise in engineering and business management has been pivotal in driving innovation and efficiency in financial technology.

Under Chen’s leadership, Tipalti has grown into a prominent player in the fintech industry, recognized for its innovative solutions in AI in fintech automation. His focus on streamlining global payment operations has led Tipalti to provide comprehensive automation for accounts payable processes, ensuring compliance and improving financial controls for businesses worldwide. Chen’s entrepreneurial journey is marked by his ability to foresee technological advancements and integrate them into practical, scalable solutions that address real-world business challenges. His passion for transformative technologies continues to propel Tipalti’s mission to simplify and automate financial operations.

Beyond his professional achievements, Chen is known for his diverse interests and hobbies, which include flying small planes, playing semi-professional poker, and a deep affection for classic computers like the Commodore 64. These hobbies reflect his multifaceted personality and his commitment to maintaining a balanced life. Chen also actively mentors young entrepreneurs and participates in various initiatives supporting education and technological innovation. His holistic approach to leadership and life underscores the importance of continuous learning and personal growth, making him a respected figure both in the fintech industry and beyond.

Episode Transcript

Kevin Rosenquist: Hey, welcome to PayPod, where we bring you conversations with the trailblazers shaping the future of payments and fintech. My name is Kevin Rosenquist. Thanks for listening. Today I’m chatting with Chen Amit, CEO and co-founder of Tipalti, a fintech company that provides a comprehensive accounts payable automation solution. He’s an entrepreneur. He’s a small plane pilot, a semiprofessional poker player, and a self-proclaimed computer geek. We bonded over our early days messing around with the classic Commodore 64. He even showed me one he just bought after we were done recording the episode. Anyway, he’s a fascinating person and well-versed in all things fintech. Please welcome Chen Amit. One of the biggest aspects of AI that companies talk about is automation, freeing businesses and people from mundane and repetitive tasks so they can focus on growing and innovating. Obviously, this is something that is at the core of what Tipalti does When you founded the company in 2010, well before the AI craze, as most of us know it, how did you help free businesses from those mundane tasks that bogged them down?

Chen Amit: When I started, my now co-founder and friend from business school, Ramsey connected me with one of his portfolio companies and online ad network that served a small company. I think there were 30 or 40 people serving thousands, maybe tons of thousands of publishers. And what I did was, I shadowed the president of the company. The president was the one who came with the pain. I shadowed him for a couple of days. I saw the ridiculous effort he needed to go through to make payments and the friction and the failures. And later on, there were compliance failures and other issues with the banks. I just shadowed him. I could see where the pain was and then tried to figure out how I solve it. A part of it is automation. Part of it is putting the task in the right hands. The best person to enter the account number and the local nuances of what regulators in a different country require that you attach to a payment that is best served by the person who’s most motivated to enter those details, which is the recipient of the money, the supplier, if you will, the publisher.

Chen Amit: I switched it from having the payer have to enter. It wasn’t so much only a solution for the finance department to be able to be more efficient with how they execute their function. Part of it was just putting the work in the right hands. Tax compliance, AML compliance, payment regulation, compliance. The right person is the supplier. Let them enter it in a simple way. guarded. Protected, asked the right questions. A lot of it was that. The other part here was automation, and there was not so much AI in 2010. Although in a prior life, I was too early in AI. You can talk about it later, but it was more about proper automation, just streamlining processes connecting all the dots, and, ensuring that people do what they are what they work on, what they should work on, and let other people work on the things they should work on.

Kevin Rosenquist: So then, as AI and machine learning became more readily available, how did that change how you guys operate?

Chen Amit: So there are areas in Tipalti that have used AI for a while. And then the obvious ones. One is OCR, so just optical character recognition and a layer of AI on top of it today with large language models, the quality of the results is just way superior to what you could have accomplished before. Before LLMs, all you could do was try to scan the words and the letters deduct the letters, and maybe try to build a word out of the collection of letters. Today you are the sentence, you are the context. You know it’s an invoice. The level of accuracy is orders of magnitude higher because you can deduct the context of what you’re scanning from a document. So that’s one place where we’ve seen great results. Another place other places are LLMs are great at context. So when you make a purchase request, there’s the context of the request When you get an invoice, there’s the context of the invoice. An LLM is great at aligning the two contexts and saying, oh, these two purchase requests that said XYZ and the invoice that said something different are one. So when we try to match invoices with purchase orders or invoices with ledgers the fact that now we have the power of context through LLMS also improved us We used to match automatically around 30% without human intervention. Now we match 93% without human intervention. So orders of magnitude. Now we’re also starting to step into copilots of sorts and helping the novice users by guiding them through using Tipalti and helping the expert users through using Tipalti. So that’s the next step. For the current step, we’re about to launch a few products on that front.

Kevin Rosenquist: When did you realize that the large language models and AI were going to make a big change for Tipalti and give you some tools you didn’t have before?

Chen Amit: I think once I understood that the LLMs are really, or at least, a part of the value of LLM is the context of the text and we have a lot of text coming from the interaction with the users, with invoices, with purchase orders, with product descriptions. There’s a lot of text around what we do. It was obvious that there were a lot of opportunities within. And I think many companies did hackathons We did our own few hackathons around AI, and the ideas were just you know very inspiring We implemented many of those. The idea of matching invoices to purchase orders came in in one of those hackathons when teams spent half a day and were able to create matches and orders of magnitude better than what the product that we’ve been investing in for so long was able to do, so that was the definite eye opener.

Kevin Rosenquist: On another note, that I read this right? Are you a pilot or were you a pilot?

Chen Amit: I am a private pilot. I got my license in 2000. I got it in Israel. Most of my flying hours are in California. I flew to Mexico and flew around.

Kevin Rosenquist: That’s cool.

Chen Amit: Private pilot. Small airplane pilot.

Kevin Rosenquist: So do you still do it?

Chen Amit: The last time I did it was when COVID hit. So, there’s a large investment to get to be a pilot, and then you need to make certain investments, time investments to maintain it. Once I stopped maintaining it, I needed to do it in the same place. So I’m commuting between Israel, California, London, and Toronto. Before COVID, it was California and Tel Aviv. So I would spend my evenings and weekends flying and it was easy for me to maintain my currency. That’s how it’s called. Now that I spend time all over the world, it’s hard for me to have enough time in one location to maintain my currency. Maybe maybe I’ll be able to get back to it. But March 2000 to 2020, when COVID hit, was the last time I flew an airplane Wow.

Kevin Rosenquist: We gotta get you back in a plane What drew you to want to fly? What was the motivation?

Chen Amit: What drew me? Well, it’s a child’s dream When I was building model airplanes, I was reading airplane magazines and all of that, and I was always fascinated with flying. In 2000, I decided to take a break from work. I was kind of burnt out from a busy ten years and I needed some time off. The next day, after stopping work, the first thing I did was, go to fly school and spend the next four months, or six months to get my license. That was a childhood dream come true.

Kevin Rosenquist: That’s great. That’s awesome that you were able to make that happen.

Chen Amit: Yes. And good memories, and exciting experiences while flying.

Kevin Rosenquist: Are you a thrill-seeking kind of person in general?

Chen Amit: I guess Flying aerobatic flying, some aerobatic flying. Not a lot, but I experienced that as well. Scuba diving, safaris. You probably can say that.

Kevin Rosenquist: A little bit of danger?

Chen Amit: Played poker. I played Semiprofessionally for a while.

Kevin Rosenquist: Did you play semiprofessional poker?

Chen Amit: I was earning money. Not enough to sustain me, but I was a profitable poker player for about a year between jobs. But they say it’s a hard way to make an easy living and it’s one of those things that are better kept as a hobby. So it still is a hobby for me When I travel, I need to find a way to spend my evenings or weekends when I’m away from home. I usually look for a nice place to play poker. And last night did. I just did that in Boston.

Kevin Rosenquist: Awesome. That’s cool. So you got a degree in computer engineering in 1990. What drew you to tech? And did you see back then that computers were going to change the world?

Chen Amit: So I’ll age myself now. That’s okay. I started coding well before the personal computer came. So I started programming on a pocket calculator called the Texas Instruments 58. Look it up. I saw it in the Computers Museum in Mountain View and then the Vic 20 and Commodore 64. And if any of that makes.

Kevin Rosenquist: Oh, I had a Commodore 64. Me and my brother used to read the stuff from the book to me, and I would type it in and see if we could get the game to work.

Chen Amit: I just bought myself a Commodore 64.

Kevin Rosenquist: No kidding. That’s awesome.

Chen Amit: I have it on the shelf right here. So I was writing professionally. Code on the Commodore 64. Educational software for, an educator, an advanced educator that wanted to use programs. That is from me at my when I was 16, 15, 16. I also cleaned cars when I was young. But quickly I moved into making my living from writing software at the age of 15 and 16 Wow. And then I think the personal computer came out when I was 20 or 21. So, I was a kid, I was spending my nights writing in assembly, writing drivers on the Vic 20 and car 64. So I was it was a hobby for me. And I went to undergrad just to continue with my hobby. It was obvious to me from a very young age that I love it, I want to continue doing it While in undergrad from month one, I continued to work and pay for my living through programming. So it still is to some extent a hobby. You say I do what I love, I enjoy every, day and, it was always part of my life.

Kevin Rosenquist: When did you become interested in or at least aware of the power of machine learning?

Chen Amit: Of machine learning? So in 2003 or 2004, I met a person who had an idea and it was all new to me, but that’s how I took my first step into machine learning, and he was an analyst who was able to find trends in multidimensional cubes. So he was in retail. His focus was retail, and he was able through hard work to find it. So you have a multidimensional cube that says for each product that it was a dairy company and like each product would have weight, would have a type of product, would be sold through a certain distributor in a certain price point, in a certain geography. There you’d have a lot of attributes. And then if you cut it in the right way, you’ll see that in this city, this type of packaging has started to fall off. So he was looking to find deviations from normal patterns in multidimensional cubes. He does that the hard way just through hard analysis. And then we learned about AI and machine learning We took a professor from Tel Aviv University to support us. It was before the models that were published by Google and Amazon and the availability of the tools and resources that you have today with the cloud and with the tools that these large companies provide for AI and ML, we had to do it at the ground level and was challenging. The company still survives, it still operates. And probably we were 6 or 7 years ahead of our time because it required more research than development and we couldn’t fund the research, so it pivoted to be more of a reporting around these domains but that’s how I got introduced to an email. And then the big players put AI and ML tools in the hands of everyone as they do with the cloud and all these tools. And that was my 2003, 2004, probably a little bit too early.

Kevin Rosenquist: Yes, a little bit, a little bit. So let’s talk about Tipalti. In your 2023 Year in Review letter, you mentioned that raising growth financing during a time when credit is tightening is a challenge. But you guys were still able to raise 150 million to continue innovating. Your customers undoubtedly find a lot of comfort in that Why is credit so tight for companies that are looking to grow?

Chen Amit: So there are multiple reasons. First of all, there’s a high level of scrutiny by all investors, credit or equity. All investors just put a high level of scrutiny. The investors themselves have alternatives. One alternative is what the government offers. There’s interest. You can get risk-free and interest-free from the government. So when you invest or when you lend to a company, it needs to assume the risk and the return needs to pay off. There is a concern about the viability of companies, right? So if a company is still burning money it has to be on solid footing. Not all companies that are burning money and most startups are burning money. Many of them will not survive the crisis. Many of them will not come out of whatever period we are in right now as winners. So, the bar is higher, the bar is higher. There’s a flight to quality, if you will, and the big players, the solid players, the skilled players, those who, as they say, reach the escape velocity We are one of them. They will get the resources to continue to build and grow, but not everyone will.

Kevin Rosenquist: And what makes Tipalti or what puts you guys in such a great position to raise capital?

Chen Amit: The metrics eventually at our scale and numbers talk. And it’s not about me sweet-talking anyone into putting a $60 million.

Kevin Rosenquist: You know winning in a poker game or anything like that?

Chen Amit: No, I’m not playing that level of stakes. It could be interesting to do that. So there are some very unique metrics that we are not only best in class, we are very unique in our metrics. So I’ll start with not the first metric to discuss, but the most unique metric to discuss, which is customer retention or churn We have a 1% gross annual dollar churn. This is unheard of. I don’t know of any company that has better channel retention, if you will than Tipalti. And there’s a lot that this one digit one, there’s a lot that this digit it tells about Tipalti. It says that we’re doing something very valuable for our customers. Otherwise, especially in this economy, customers would share. It says that there aren’t equally viable alternatives. Because if there were viable alternatives, then just the nature of human beings would be to select some different, to churn to be disappointed like we cannot be perfect so right some will switch to those, and some will switch for price We are not the cheapest. by no means. Maybe we’re the premium-priced solution out there. So when you have 1% churn with everything else I mentioned, it means that you’re doing something extremely valuable and you are well protected comparatively. On top of that, we have a huge addressable market. There are a million customers We target the mid-market. Mid-market is 100 to 1000 employees. The office of the CFO. There are millions of customers out there and only 5% of them have a solution, whether it’s Tipalti or someone else, there are a few other solutions. So when you have the win rates, we have, when you have the retention rates, we have, when you have the address, the green field, that’s ahead of us, that gives a lot of comfort to investors. And then the sales economics, the unit economics, the margins, everything else. It’s all the metrics that investors care about and we accept it.

Kevin Rosenquist: I mentioned in my opening how as AI becomes more prevalent, more and more companies are searching for ways to use it and to help their workflow. And a lot of times automate tasks. Do you find that you’re getting more companies coming to you looking for ways to automate just because they read in the news how AI helps automate? Is there something like do they come to you like, hey, what can we do here? How can we help this?

Chen Amit: I’m not sure that the AI is already the driver for that. But automation is cost reduction. Labor reduction? The financial pressures and the need to be more efficient that the economy is pushing all of us and our customers to be are among those. So when you have fewer resources, when you cannot grow your resources, maybe you have to shrink your resources. You’re looking to replace labor with automation, and we save 80% of the labor associated with the tasks that we automate around finance operations We free 80% of the labor and we accelerate processes We accelerate financial close by 25%. So we reduce the direct transaction costs. So when a customer of ours pays a supplier, the cost to pay is lower with us than it is with their band. So they are looking for automation. They are looking to free their labor to do other work than these manual, mundane, risky, labor-intensive tasks that we automate for them and reduce costs where they can. I think our role in Tipalti working with finance leaders who are not always tech-savvy or not always as geeky as you and I may be to mediate AI for them to make it accessible for them, to help them benefit from AI in the finance world, wherever we can, it will be harder for them to do it on their own. So our role in Tipalti, that’s how I see it, is to mediate it for them and to make the value of AI accessible, easily accessible for these finance leaders.

Kevin Rosenquist: From a compliance standpoint, does automation help with compliance? Just because you don’t have to worry about human error or nefarious actors or anything like that, do you? Do you find that automation is a positive? I feel like a lot of people might maybe nervous about automation because of the compliance aspect of it.

Chen Amit: So there are two-way areas to answer this. One is how our customers benefit from us and ensure that they are compliant, and how we ensure that we are compliant. So for customers who are not compliant, savvy, not in global payment regulations, the nuances of taxes around the world, all of these issues and even the complexity of US taxes W9, WA 1042 1099 all these and it’s not necessarily the thing every finance leader is proficient in We are because that’s our job. So we have people, product people, engineering people, compliance people, and dedicating others themselves to ensure that the product provides the compliance that our customers need or that our customers need. And by that, we provide a relief for them. Customer number one, I mentioned to you, my co-founder of Tipalti and the president that I shadowed, if you remember.

Kevin Rosenquist: Yes.

Chen Amit: The point where they pivoted and said we need the product now. We were building an alpha-beta kind of engagement at some point they said, we need tomorrow, you need to go live tomorrow Where it was when they failed the compliance issue with one of their banks. And the issue was that they were an ad network that makes payments all over the world. And the publisher in some remote country. It’s not a very close relationship between an ad network and that it could be an individual person in some remote country. And you’re required by law to vet that this person is not on a blacklist. They didn’t know that they needed to do that We built it into our product even in a version in the very first version, and they understood it and they switched to us and we were able to overcome their compliance issue. That’s called customer number one already today. All these types of customers have suppliers around the world and today almost every company, about minimal size, has some relationship across the border.

Chen Amit: So for them, we are the solution they know they trust us, and that we are doing the work that they are required to do in order to be compliant as it relates to us and at risk. And we have millions of suppliers that were addressing tens if not hundreds of millions of transactions. That and tens of billions of dollars that we process. And we require automation. It’s not something that you can solve with labor. Labor handles the exceptions. But in the very early days, we developed a fuzzy logic to do this matching to match the blacklists and the names that we have. It’s not always 1 to 1. Sometimes the names from different languages can be written in multiple ways. So you need to find a way to do the matching in a so that, that that you whether you have a hit or don’t have a hit. So we use automation a lot because of scale. And then with what we build, we’re able to provide safety to our customers.

Kevin Rosenquist: You work in a lot of different industries with a lot of different types of companies. One that caught my eye on your website, being a former musician is that you work with companies that have to pay royalties to artists. How did that become a use case for Tipalti’s product?

Chen Amit: So It’s kind of a version of the very early that how we started to partner. So we started I said that we started with ad networks, but it was the crowd or the gig economy at large. The first two customers were ad networks, but the third customer was already kind of a crowdsourcing if you will. It’s a Seeking Alpha. Seeking Alpha is a website for financial content, and the writers are all over the world, and the advanced writers get paid by Seeking Alpha. And they needed us to make payments to those writers. And we kind of iterated and continued. And then game developers and then streamers and then musicians as well. So Twitch is a customer, and Roblox is a customer. And, we have several royalties, music royalties, customers. It’s similar. I don’t know if calling it gig economy is relevant or is the right description. But it’s when you have masses that put content out there and get paid through content, many of our customers are like that. I mentioned a few of them, and it was an evolution from the very first few customers that we understood that one of the huge pain points these types of customers have is that they may be small companies, but they interact with thousands if not more of a potential PRS around the world. And that comes with complexity and compliance, complexity and payments complexity, and taxes and currencies. And we can help there.

Kevin Rosenquist: That’s cool. That’s cool. That’s funny that I didn’t realize that that’s kind of one of the early things that you guys did with your product which is to help the gig workers of the world.

Chen Amit: That’s how we start started.

Kevin Rosenquist: That’s funny.

Chen Amit: And at some point, like 2 or 3 years later one of those, I think it was an ad network customer, reached out to us and said I spent 15 minutes to pay 3000 publishers and then half a day to pay all the 50 regular suppliers I have. Can you do your magic on the regular suppliers as well? By that time, we were already kind of deep into research of traditional accounts payable. And this was just a trigger. And we started working with that customer. And that was 10, 11 years ago. And by now, it’s the largest part of our business.

Kevin Rosenquist: So general sense of being a computer guy what areas do you see AI and machine learning having an impact that maybe isn’t talked about as much, both in fintech and and just society in general?

Chen Amit: I don’t know if it’s not talked about. I wouldn’t proclaim to say it’s not talked about. I can tell you how I look at the use cases in the body. I always get surprised by innovation We just came out of hackathon again. All kinds of nice, interesting, crazy ideas that we will.

Kevin Rosenquist: Nice, interesting, and crazy ideas, I like it.

Chen Amit: Yes. From a product perspective, there’s a use case for the novice user, and there is a use case for the expert user. The novice user of Tipalti would just think of an employee who needs to purchase something. They make purchases once, twice, three times a year. They don’t do it on a regular basis. They don’t know Tipalti well enough. The best way to interact with us would be through a chatbot. I need to buy myself a headset. Great We used to buy headsets, buy whatever Bose, and Will this version that we just bought work for you? Just have a dialog with the bot and then get the request routed. After all the information was collected in a conversation. The alternative is to fill a form with all kinds of fields that the employee may or may not know what ledger code, what department, what I don’t know what department, and who ledger code. Can you answer that for me? So that’s how we will interact with the novice user. On the opposite, the expert user might have questions that a simple. I’ll give you an example of a simple question how much did we spend on Dimension in London in Q4? Very short sentence that describes exactly what I want to ask. If you want to get the answer, you probably need to download three to four files, consolidate them, aggregate and pivot Vlookup and whatever, and get to an answer. It’ll take you 20 or 15 minutes. But an LLM, can parse the question to a few API calls and get the answer within seconds. So that’s the other use case where we’re building both use cases. The novice user is about to come out in the next few months, and the expertise also. It’s all in the next few months. So these are kind of type of copilot, but not generic copilot answers. Just trying to be very precise in the exact value of the exact use case that we have for the customer.

Chen Amit: And there’s a lot in the behind-the-scenes or not only behind the scenes, but there’s a lot outside the product. So the obvious one is customer support. You get the questions and the questions you get from customers and users. It’s the 80/20 rule. 80% of the question will be around 20% of the problem areas. You can put AI to work there and get 80% of the workload automated to AI. We’re doing that. Educating our own employees is a complex system. So when you need to understand something internally, you are a support person, a solution consultant, an engineer, a developer, asking an LLM that sits over all of Tipalti’s knowledge, the documents, the presentations, the support tickets, everything, the code itself. You can get an answer quickly. Very accurate with all the pointers to the right documents. So, these are some of the areas where we take advantage of.

Kevin Rosenquist: How close do you think we are to AGI or artificial general intelligence? I like to ask my computer nerd guests these questions.

Chen Amit: Yes. I wonder if you think how LLM works. It’s guessing the next character and it’s predicting what the next word is. It’s not intelligence. It’s building an answer based on your knowledge. So I’m not sure that that’s in itself intelligence. I don’t know what to call it, but I don’t think that’s the AGI we fear or not only fear, but we transform the world but it provides a ton of value. You see, every version, 404. 404 is so much better than 4Nm. Now with audio and video. It’s phenomenal. but is it general intelligence? I’m not sure. We may need yet another breakthrough to get to real intelligence.

Kevin Rosenquist: Well, by all accounts, Sam Altman is working on it. It sounds like he is.

Chen Amit: He is and he has a lot of money and a lot of resources and a lot of smart people around him. And he’s doing wonders. Know that.

Kevin Rosenquist: For sure, for sure. Well, Chen great talking with you. Thanks for being here, I appreciate it.