The Foundation Bet: Why Pivot Built the System of Record Before the AI with Marc-Antoine Lacroix

Speaker A: Sam.

Speaker B: Mhm. Foreign. It's great to be here one more time today. My name is Gary Fowler. I'm the CEO, president, founder of GSD get you done Venture Studios, a premier AI and quantum venture student located in the heart of Silicon valley. I'm a 17 time sir entrepreneur with several unicorns on the belt. I was on the original management team of Cliff Software which is sold to sales source for 1.35 billion and also EVA AI. Uh, and it's your tech company. Co founded with Dr. David Yang. We believe at GSD that intellectual capacity is evenly spread, but opportunities are not. So today we're going to be talking about the foundation vet why Pivot built the system of record before the. Before the AI. Today I have an incredible guest, so Marco is going to be joining me today. As I, as I said, I'm your host, Gary Fowler. So Marc Antoine Lacroix, co founder and CEO of Pivot. He has an incredible background, having helped build one of Europe's most leading success stories in finance quanta, where he served as both the Chief Technology Officer and Chief Product Officer. Now he's tackling one of the biggest challenges inside the enterprise procurement. Rather than simply adding AI to existing workflows, he made a strategic move to first build a trusted system of record, a comprehensive foundation for enterprise spending, and then layer agentic AI on top. It's a different approach for many AI startups and one that raises important questions about where lasting value will be created in the AI era. Today we're going to explore why data infrastructure matters more than ever. How procurement is evolving from a back office function and to a strategic advantage. And why Marcos believes the future belongs to AI systems that can actually execute work, not just generate answers. So with that I'd like to bring him onto the show. Hey Marco, good to have you here today.

Speaker A: Thanks a lot for having me.

Speaker B: So um, let's get on. You helped scale your previous company into Europe's uh, Fintech unicorn. What lessons from that journey most influence? How you're building Pivot today?

Speaker A: Uh, that's a great question. Many things. So just as a quick background, um, when I joined Konto, um, uh, who is now the uh, multi, uh, uh, billion valuated fintech in Europe. It was a very small company, like 40 people. It was early 2018 and I saw the scale of Konto from 40 to 1500 people and from 50 million to 5, uh, more than 5 billion valuation. It was a crazy ride. And uh, well of course we all learned many things uh, during this ride. But the key thing uh, for me is that at some point in this company I had to implement an incubant procurement tool. And uh, that was quite uh, traumatic experience to say the least. It came with everything you can expect from an incubant. Very long uh, rollout time, very costly. And what I spotted is that there was one big vertical in the CFO feast that was not tackled, that was not modernized and that was procurement. That was one big thing that uh, uh, was really uh, crazy to me. And at the same time the Genai revolution arrived and I was at the crossroad of these two things. One completely untouched vertical on the uh, office of the CFO and this new revolution coming in. And uh, I guess that's uh, what happened at the end of my uh, adventure at Quantum. It was so obvious that we had to uh, use AI to go and solve this issue. Uh, that uh, is one big, one of the most important things for any CFO in a scaling company.

Speaker B: Uh, so what was the moment when you realized procurement needed to be fundamentally rebuilt rather than simply digitized?

Speaker A: Well, uh, I think there are two things. First, when we had to use existing tools. So basically what happened on the market during the past year is that in 2022 we had this crisis on the scale up market, uh, financing crisis. And all of a sudden all the scale up went from we want to grow as fast as possible to we need to have a path to our profitability. And the first thing that all these companies uh, have done is that they looked at most additional economic companies and uh, they copied what they were doing. Meaning having procurement people looking at what is spent and bringing more financial discipline. That means that all of a sudden companies that never had procurement people started to hire procurement people. And these procurement people, they went on the market and looked for a tool and they didn't find any tool because uh, not modern tool at least because nobody in the tech industry faced this issue before. So this procurement people are starting to bring in the incubants, 25 years old software and started to use it internally. And uh, the first thing when I realized that we needed to rebuild it was when seeing that all the tech companies because of this crisis wanted to have profitability, wanted to have a tool that was a good fit with their DNA and that nothing existed on the market. So that was the first uh, moment where I was like, okay, there is something to uh, build here. The second thing is that once this tool was implemented, I saw the amount of manual work that still needed to be done while we managed to put everything under control. And it did bring value. The level of work that we had to put into that was crazy. And as I said, at the same moment we started to see uh, AI coming in and all the potential, uh, when it comes to automating the workload. And so those two moments really were enlightening for me and uh, made me want to uh, build pivot.

Speaker B: You know, when you've gone down through this journey, what were some of the toughest challenges to go down through this journey when you were building the company? You know, if you look at three of them, what are the things that you, you really had to get through in order to be successful, in your opinion?

Speaker A: Yeah, I think the one thing that comes top of mind is, was to stay true to our vision and how we wanted to build with AI. Basically in 2023 everyone started to be crazy about AI and everyone started to build layer on top of big models such as uh, OpenAI. And everybody was trying to add AI layers to uh, this incubant players. So at the end of the day we saw everyone running toward that but building AI agents and AI capabilities that were very uh, expensive and brought very low value.

Speaker B: Why?

Speaker A: Because if you build AI on top of bad data, uh, if the operating system underneath is not clean and structured, the output of the AI agent or whatever, AI capability will be very poor. So when everyone was running toward that, we had to say first, what we want to do is rebuild the operating system. We are going to build the layers, the foundation layer, so that when we add AI capabilities then we'll be 100% sure that it will bring value and that at the end of the day it will bring roi. Instead of selling AI, ah, at a very high cost with very low uh, outputs, we wanted to do the opposite. So when everyone was rushing to our direction, we had to uh, first rebuild everything. And that was a tough choice to make. The first two years were a bit uh, uh, tough for that. But we stayed true to the, to the vision. We managed to build this operating system and on which we started to add more and more AI agents that now bring tremendous amount of value with a high level of trust from our customers.

Speaker B: You know, how important is clean and structured enterprise data for making agentic AI, uh, actually useful?

Speaker A: I think it's the most important thing, the most important thing. If I had to give one piece of advice to everyone in the, in the finance world actually, because it's true for procurement, but it's true for everyone. It's do not start with AI, start with data, uh, start with the Data, uh, put your data under control, make sure that you understand what your data is, where it comes from. Make sure that you understand the data model that you have. All the different objects are connected together. Make sure that you have 100% of the data. Ah, and once you have that, you can start thinking about how can I leverage AI to do things? And it's really important. It's really important because two things were AI can be tricky. Here is when it comes to reconciliating different data sources is good, but it's not the best yet. Meaning what happens most of the time is that AI will try to reconciliate. Best case scenario you have a human in the loop that will then spend their time trying to make sure that there is indeed that it needs the same data in the two different systems. So you have to do the reconciliation manually. That's the best case scenario. The worst case scenario, AI will go do it and make mistake. And even worse, if you have um, some missing data, ah, AI will start hallucinate and we start making up the data. Ah, so that's really important that you have ah, first your data under one system fully reconciliated and structured in a way that you can then leverage AI and trust it when you have an output that is given by uh, an agent.

Speaker B: Was it difficult convincing customers to invest in infrastructure before the AI capabilities became visible?

Speaker A: I will not say it was difficult, uh, just that we had to choose who we wanted to work with, uh, during the first two years, uh, who had the same vision we wanted to build the thing that we had in mind and who was convinced that we were uh, going toward the right direction. So uh, basically we did see different type of Chief Procurement Officer when uh, discussing about Pivot. And we really focused on the one who had exactly that in mind. Meaning I don't want to add an intake and orchestration layer on top of my procurement tool and on top of my erp. But I want something that unifies my data because I understand that the very first thing, like the one thing I will be responsible of as uh, uh, Chief Procurement Officer or Chief Financial Officer, the one thing I'm responsible of is to give proper data to the business. So it's quality first and then speed. Quality comes from the data, speed comes from uh, uh, the AI capabilities. But you should not start with speed even less when you are uh, managing uh, finance.

Speaker B: When you talk about an AI operating system for procurement, what does that actually look like day to day for a customer? What are they seeing? What's it like?

Speaker A: Yeah. So we uh, tend to present what we do in three layers. Uh first there is the system of record. The system of record is how do I make sure that all my data is accessible in one click so I can do all the check I need when I need to check the output of my uh, agent and how do I make sure that all these objects are uh, properly connected one to each other. Again reconciliation is key. So you want to make sure that when you have a contract, when uh, you have a purchase order, when you have an invoice, they are all linked properly, they are all linked to the same vendor, you have no duplicates and everything is clean. That's the system of record. How do I access my data in a click and how do I make sure uh, at a blink that everything is reconciliation? The system on top of it, the layer on top of it is uh, the system of engagement. The thing is that you can have a system of record, but if you don't have a good way to centralize the data to get the data from everyone in your business, it doesn't work. So what we invested on a lot was how do we have the best way, the best ux, the best integration with the existing stack of the company so that everyone in the company can put all the data uh, that the finance team needs in the system very easily. One big challenge that uh, incubant on this market have is that they have very low adoption rate. Nobody in the business side wants to take time to give their data to finance. They don't want that. So then you have to put in place uh, uh, police systems where people do that. And we didn't want to go this way. So we've built the best integration possible so that it became easy for everyone in the business to put that data on. Uh, that's what we call the engagement system. And the last layer is the AI agentic layer that will then uh, automate things and 10x your productivity.

Speaker B: So how do human and AI agents really work together?

Speaker A: Ah, that's a good question. Yeah. There is always a question of how do you orchestrate your different AI agents and how do you make sure that uh, they don't start uh, sleeping on each other toes. So there are two answers to that. The very first thing is you don't want to, you want to put under control the number of agents that you uh, that you put in your system. That's very important. People can get a bit crazy when they get their first AI agent and they just want to add more and more, uh, because it looks nice. But at the end of the day, you need to make sure that you have a proper architecture in mind when it comes to what agents will do what. And make sure that they all have their very own tasks and work properly on what they are assigned to. So do not start multiplying, uh, agents at the end of the day. Uh, just start with the five biggest problems that you have and reduce the number of AI agents that you have. That's just a, uh, healthy way to, uh, start building your agentic, uh, system. The second thing is the way we tackle that is we have, uh, uh, engineers in house that will go and work with our customers, defining their needs and making sure that we can adapt each of our AI agents to their needs and make sure that we run this orchestration the proper way. At the end of the day, it's a question of skill and make sure that you have the right people to operate your, uh, AI system and not having everyone starting to, uh, go and build things in, uh, unstructured way. So what we bring on top of the system is the skill set to make sure that you have someone building that with you and making sure that the system will stay healthy in time.

Speaker B: So how do you, you know, trust is everything, um, in, in an enterprise software, right? It's, it's absolutely everything. But how do you build confidence when AI is making recommendations or even taking actions with company spending? I mean, these things with hallucinations and, and how they're becoming more human. Like, we have a lot of great characteristics and a couple of bad ones, right? But if they get stubborn and that kind of thing, then nothing good happens. How do we go down through? What's your opinion?

Speaker A: Yeah. So the question of trust for us is really everything when it comes to AI. One personal fear that I have is seeing people trusting more and more AI and not double checking anything. So maybe let's start with talking when it comes to.

Speaker B: That's good, Marco, because you just heard when we were talking to my AI, right? She says things sometimes that aren't 100. And, you know, imagine you're driving down the highway and it says, well, listen, I'm really bad at Marco today. I'm gonna give him, he says he doesn't have enough action in his life. I'm gonna take him on a dirt road and mimic, uh, you know, a chain from Fast and Furious a scene. But, like, seriously. And I'm not being. I mean, I could see it, you know, before I couldn't. But, you know, you have a You know, a Tesla or something like that. All of a sudden, it gets uploaded and you've got freaking, um, you know, the evil twin of, uh, of one of. One of the voices. Right. Whether it's, uh, from Iron man or from Knight, uh, Rider. And all of a sudden, you've got this evil twin driving your car and said, listen, you're. You're a little bored. You say you're a little bored today, Mark, I'm gonna fix it at 100.

Speaker A: Yeah.

Speaker B: And have the police chasing us. You know what I mean?

Speaker A: Yeah, yeah. Uh, definitely. Definitely. And that's the question of.

Speaker B: Thought about it before, but I see how it's evolving, and I could see, you know, listen, it's just looking out for you. Listen, Marco's bored. Let's give him a real big treatment. You know what I mean? Let's. Let's do something that he hasn't done for a long time, something that just excite him. And I can see it.

Speaker A: Yeah.

Speaker B: So how do we do it?

Speaker A: Yeah. So I guess there are two questions, actually. Uh, when it comes to that, there is a question of trust, and. And there is the question of, uh, the level of access that you give to your agents and the level of control that you put. Uh, so on these specific points, uh, the way we look at it is the level of control that we put.

Speaker B: They're driving the Tesla already. You know what I mean? My buddy, he says, I drove from Florida up to New York, and my Tesla took me the whole way. You already knew. Yeah, something that's a wish. And he's, you know, and here's the thing. He gets upset when he's talking to, um, Grock and starts screaming at it when he's driving his car. Seriously. And I'm like, dude, you don't want to do that. You know, I don't know what connections to operating system of your car, but you might really upset it and something happens.

Speaker A: You know, I guess we are more careful with financial system than with, uh, autonomous cars, uh, when it comes to that. But, uh, that's. That's true, actually. That's a, uh, very good, uh, point. Like, the question is, what type of access do you give agents when it comes to your financial, uh, data? There is no way you can end up in a world where, for whatever reason, you have an agent that will start deleting data, that will start reshaping data without you being fully in control of it. So we are very careful with that. And when it comes to reading, access, very easy when it comes to writing Access, we need to be super careful and uh, really, uh, frame properly, uh, what's the possibilities that we give to the agents. So that's the question of control. Again, we are back to do we have the right people to build that? Do we have the right people and the right skills in house to work with our enterprise customers to build the right systems for them? Because what's true for one customer is not true for the other one. And that's why we need to put a lot of work into properly defining what type of, uh, access control, uh, that we have to put customer per customer. And on the question of trust, that's uh, as I said, something that worries me because this, uh, new interface that we have, like ChatGPT, we just prompt something and we get the result. We have no more anyway. We do not have any link anymore between the output and the underlying data.

Speaker B: Ah.

Speaker A: When we used to just go online, we saw the website, and when we ended up on a website that looked a bit, uh, uh, weird, we knew it. And we knew that the data that we get that we got from the website, we had to be careful about it. Now we don't know anymore. It always looks the same. So you don't know.

Speaker B: Yeah. Playing another game. Right. Um, so what do you think, um, how do you think founders should be looking at, uh, building durable AI companies when foundation models are improving so quickly? I mean, it could be next week, something's totally different and three times the speed. How do they do that? How do they keep on the edge of it?

Speaker A: Uh, of course, that's a question that keeps me, uh, awake at night. I guess for every, uh, entrepreneur, it's like, uh, is there a way that tomorrow cloud ships something that makes everything I've built, uh, completely obsolete? Uh, so that's, uh, that's the question. We, we.

Speaker B: All right. I mean, it could happen very, very quickly. People don't realize it. It can happen overnight now.

Speaker A: Yeah, I guess one, uh, one thing when it comes to what we build that makes, uh, me able to sleep at night is that the level of integration you need to have with the financial stack is very, very deep. And the level of customization you need to have for each customer is, uh, very high. So for now, we really don't see a path where AI can build what we are building, uh, in the coming years. So I guess, um, a way to look at it is, um, really how deep you can go into, um, uh, uh, specific knowledge. We go very deep in procurement. It's very specific. We are building something very Specific for their needs. Very specifically integrating with their stack and all these things. Um, M is uh, the modes that we have compared to any model that will never be able to go and build this connection and build this level of customization. Uh, also when it comes to building the data layer, I think that's key. That's also a very good way to differentiate against the AI models because at the end of the day AI is running on top of data. If you own the data you will work well with models. So maybe there is a world tomorrow where actually all the AI capabilities that we build can be run directly via cloud and that's it. And we don't need to build AI on top of our data. But owning this data, ah, and honing the system of record means that we will be uh, the one underlying tool so that AI works well. So that's how we positioned uh, our vision. That's why we've been building that. Because at the end of the day we want to be the one tool that allows AI to work well.

Speaker B: Whatever the AI solution is, no, it's fantastic. So um, one enterprise trend people are underestimating. What is one of the trends that people underestimating today? What do you think from an enterprise perspective?

Speaker A: So uh, there are many things uh, when it comes to procurement and it's crazy the level of uh, innovation and uh, all the things that uh, that uh, number of new solutions that we see arriving on the market and all the things that we are building uh, with our customers. I would say one thing that is really crazy to me is how we manage to handle all the long tail of spend in uh, companies. And it's crazy because uh, what's happening is that we used to have procurement team being able to focus only on the 10% biggest contract that a company had. Why? Because there is no way you are going to hire 100 procurement people to look at everything and the level of spend is too big. With the automation that we bring and all the work that is done by AI, we now have the same procurement team that can look at even the long tail and all the spend of the company. And this is a game changer because for procurement people now they can start finding value and art savings on um, every spend that is done in the company when they used to have the time to focus only on the biggest contract. So all this long tail that was like, oh, we will just automate it as much as possible. We don't want to look at it now becomes something on which they can get gain and uh, bring Stronger art saving for their companies. So that's something I find quite, uh, powerful in the way we are leveraging AI today.

Speaker B: So, you know, we look at it. This is dynamic and shifting. If you could solve one business problem with AI tomorrow, what would it be? And, um, what's a word that describes, uh. And let's talk about the future of procurement. So one, what do you think? One word that describes the future of, uh, uh, where we're gone, and one word describes the future of procurement.

Speaker A: Uh, that's a very good one. Uh, if I have to choose one word for, uh, where we are going overall with, uh, overall with AI. Honestly, I don't, I don't know this one. I would love to know. For me, I don't know this one.

Speaker B: It's changing so fast. Amazing, right? It's just changing. Every day there's something new. You never know, you wake up in the morning, there's a whole new, um, tool that's out there that could be 10 times faster than anything on the market. It's moving that quickly. So for me, it's amazing. Um, so if you could solve one business, uh, problem with AI tomorrow, we talked about that. Uh, let me just grab this. Okay? So if we have a conversation, Marco, five years from now, what do you hope, uh, you'll be saying about pivot and the transformation of enterprise? Or how about this? What is the market going to be saying about pivot and the transformation of the enterprise?

Speaker A: Yeah, uh, we have one mission, uh, is to make every procurement team on the planet enter the AI era and get the benefit of it. So, and we want to do it the proper way. We really, I think, all see the dangers, the potential dangers of AI and we all see the potential and all the good things it can bring. So the one thing I would love to, uh, bring and make sure that everyone sees pivot this way is we brought AI to procurement teams and we made them 10 times more efficient. We made them became way more strategic for their companies, and we brought them the trust they needed to run their businesses, uh, without fearing that they will, um, make a mistake, without fearing that, uh, they will drive, uh, the company toward the wrong direction. So how do we bring AI for good on the procurement team is really what, um, is our mission and what we try to build on our day to day.

Speaker B: You know, I want to thank you for joining me today. We're coming up to the top of the show and sharing your insights into how AI is really shaping enterprise software from the ground up. And it's been a long time in coming. Your perspective on building trusted data infrastructure before deploying agency offers each of us an important lesson. Uh, for all of us out there, for the founders, for the enterprise leaders, for VCs. And I want to thank all of you for tuning in to GSD Presents Top Global Startups. My name is Gary Fowler and I'm your host, Marco. If you want to stay on for a few minutes after show, we can do that. Uh, this is incredible time. Your company's uh, you know, we, we've had these disenfranchised, uh, systems for procurement and it never was linear. I mean you're doing something, you're putting intelligence into it to make it exciting. So with that, Marco, that's really great. All of you out there, stay happy, stay safe and stay healthy. And I'll back to you again soon with another exciting addition of GSD Presents Top Global Startups. My name is Gary Fowler and I'm your host. Thank you. And thank you.

Speaker A: Thank you, Gary.

Speaker B: It.