Cutting Edge AI

#4 From Legacy Systems to Agentic AI: Fabian Heinrich (CEO, Mercanis) on the Future of Procurement

• Angel Invest • Season 1 • Episode 4

Share episode

Procurement is one of the largest and most complex parts of the global economy – over $100 trillion a year – yet much of it still runs on spreadsheets, email threads, and legacy systems built decades ago. Mercanis, founded by Fabian Heinrich, is taking a different approach: rebuilding procurement from the ground up around agentic AI.

In this episode, Fabian explains why procurement was historically underserved by modern software, how his first startup exposed the inefficiencies behind RFPs, supplier comparisons, and contract negotiations, and why large enterprises still rely on Excel despite the stakes involved. We unpack what “agentic procurement” actually means in practice, from AI agents reading hundreds of pages of contracts to running complex pricing scenarios and orchestrating sourcing workflows end to end.

We also discuss the recent inflection point in enterprise demand for AI: procurement teams suddenly have budget, and board-level interest in automation has changed buying behavior, and why heavy industries like automotive, chemicals, energy, and pharma are now at the center of AI-driven transformation.

The conversation goes deeper into the future of work inside procurement: how the role of the buyer is shifting toward orchestration and analysis, what human-in-the-loop really looks like when agents negotiate with other agents, and how procurement systems will increasingly sit on top of, and eventually eat into, legacy ERP and compliance infrastructure.

If the last decade optimized procurement interfaces, the next one is about autonomous systems running the process itself.

[00:00:00] Fabian Heinrich: I personally would love that from 2027 onward, we don't sell any more non-agentic software because I think also the profession of procurement will change and will evolve a lot with the advancement of technology and software.

This is Cutting Edge AI brought to you by Angel Invest with your hosts, Jens Lapinski and Robin Harbort. 

[00:00:31] Robin Harbort: Our guest today is Fabian Heinrich. He's the co-founder and CEO of Mercanis, an agentic AI Procurement Suite. The global procurement market is over $100 trillion per year. And yet the majority of it is still managed using outdated, non-optimal tools like Excel.

But now, Ani equips procurement teams with advanced AI tools and insights to make smarter, more effective procurement decisions. This year, Fabian has observed a sharp increase in corporate demand for AI solutions. Perfectly timed. with Mercanis raising a 20 million Series A this summer, we'll talk about the future of a Gentech AI procurement, the complexity of this industry, different integrated AI strategy into enterprise software, and as well far beyond's vision for this decade. This is the Cutting Edge AI podcast by Angel Invest. Let's go. Hi, Fabian. Great that you are here on the Cutting Edge AI podcast. 

[00:01:38] Speaker: Yeah, great to be here. Um, Jens and Robins, thank you so much for your invitation. 

[00:01:43] Robin Harbort: Yeah. Also, hello to Jens, my fellow co-host. Great to have you here. 

[00:01:45] Jens Lapinski: Hey Robin. 

[00:01:46] Robin Harbort:  So, so Fabian, you are the founder of me, but but isn't the first company you founded, right?

[00:01:55] Fabian Heinrich: Yeah, exactly. I've been an entrepreneur, I think now for 10, 12 years. Uh, initially, um, started with Rocket Internet who made me kind of, uh, to the entrepreneur I'm now. And uh, my first rodeo was, uh, with Scalpy, uh, which I founded earlier on. And I experienced what procurement, um, actually means, and that led me also too founding Mercanis years later. And, I mean, Scalpy was a pretty simple idea. I mean, if you think about sales, you are using tools like ZoomInfo and Lucia to find new leads and Scalpy was exactly that for procurement and finding new suppliers, finding information about new suppliers. The more basically I work with our customers on the product, and the more I got to know them, I've seen well once they have figured out, what suppliers they wanna buy from, the painful process actually started because then they wanted to compare the suppliers, they wanted to have upper to uppers comparison. They were doing those so-called RFX processes, requests for information or quotation processes. The bigger the demand, the bigger the tender and the process.

And it was just a lot of manual work and a lot of time procurement people have been losing. And in addition to the time, the fascinating thing for me was what kind of dollar amounts are being moved around in corporate procurement on exit spreadsheets? Um, I was doing a bit more research and that have been seeing it's not only a thing for the enterprise customers we had in Scalpy, it is a thing for everyone. And the global annual procurement volume is a hundred trillion. If you, of course there's a lot of public procurement, but if you then go towards the private sector is around $55 trillion every year. And if you then think about, a lot of that is run on Excel spreadsheet it's uh, crazy. 

[00:03:57] Robin Harbort: But what you have done at Scalpy, this was all before AI, right? 

[00:04:03] Fabian Heinrich: Yeah, I mean that was, uh, I mean, what means before ai, I mean, AI goes back to the sixties, right? But it was pre, it was certainly pre LLM and uh, what we did with, uh, intelligent and ML based web crawling, we figured out new, new suppliers, but was certainly light years before the LLM area.

[00:04:24] Jens Lapinski: So what was the, the thing that led you to, to Mercanis? What was the, the insight that you generated at Scalpy or after you left there? So what was the core idea where you thought, okay, so they do this in spreadsheets and now what? Now I can, I can do this with LLMs in a certain way. What, what was the, what was the spark?

[00:04:44] Fabian Heinrich: Yeah, I think even before we started Mercanis, the were no LLMs. Yeah, I mean we, we, we founded pretty much five years ago, the company. So what we've seen, the only solution to that, uh, is like 20, 30 years old legacy system such as SAP Ariba, or like Cooper, like big old systems, which have the look and feel of Windows 95, and we were saying: "Hey, there need to be a really intuitive, easy way, basically consumerizing experience and delivering the experience that people have in the, in the private life, on the, on the apps or the template or on the iPhone, in there, basically B2B software, that it's fast, easy, seamless", and we were always saying, we don't compete against Ariba. We compete against Excel and Outlook, and we just have to be better. And it just has to be easier than Excel and Outlook because those legacy systems, they didn't have any automation. They're just like so to say data input fields, if you think about how old SAP Ariba is, which is still the market leader in procurement software, I think it got initially founded in the US in 1992 or 1994. So I mean, of course maybe they rewrote the code, but it's certainly has some edge.

[00:06:09] Jens Lapinski: that's let's go into Mercanis started there, but then initially in the first, I don't know, couple of years or something like that, it went with a certain speed. Right, but the speed has recently picked up a lot. Yeah. So the acceleration in the business is really significant. I mean, I can say it like that. Without disclosing any confidential information, how do you think about that? What has impacted that? Apart from, of course you've got more software, it's more mature, you understand your customers better, you've got more features, but what was the inflection point in the company if you like?

[00:06:41] Fabian Heinrich: I mean, it's the combination, I think it's the combination, what you said, the maturity of the company, of the product. But I think it's also like, market timing because I mean, you mentioned, Robin, was called with the pre-AI age. And I think it was certainly the pre-AI age when you think about procurement teams.

Yeah, because when we were pitching in 2017, 2018 to procurement people, "Hey, we have a AI-based solution to fight new supplies, they were scared. They were like thinking like, Hey, they wanna take our jobs away." And they were thinking about that fight human against machine. Now I think what. Changed in particularly 2025.

I mean, maybe in the US before, but we are now in Europe and in particular the DACH market. But I feel like 12 months ago there was some kind of inflection point in every customer conversation is, was what benefits can you deliver with AI? Show me your AI capability. Show me the automation with the AI.

I just got an AI budget or I can, I can finally get budget for procurement software after 10 years because my management board has AI on the agenda. And I think that appetite, that hunger for AI driven solutions has completely changed within the last 12 months. 

[00:08:03] Jens Lapinski: Yeah. Interesting. Why do you see the biggest benefit of the AI in the system for the customer?

Like what, what does it, where does it take a lot of, where does it increase options? Where does it take cost out? Where does it increase speed? What you demo to them and they say, oh my god, that's fantastic. 

[00:08:19] Fabian Heinrich: I think you have to bear in mind the complex, I think tasks and topics procurement is dealing with.

Because if you think about contracts, they don't deal with a two page contract with a PRO marketing agency. They deal here and there with contracts ranging from 155 to 300 pages. So imagine like a funding round documentation, that kind of documents when they talk about their contracts. Also, if you think about pricing sheet, it's not like "Hey, I wanna compare supplier A, B, C, and it's the hourly rate of a lawyer." No, I mean those pricing sheets can have hundreds of line items. So it's quite a complex comparison and I think LLMs are perfect to finally give the automation to really this, this one click show them detailed comparison, give them suggestions, recommendations on such vast data amounts and documents.

[00:09:18] Robin Harbort: Can you maybe give some examples like which specific customers do exactly have those problems? So, which industries or are like, uh, facing those large amounts of input data they need to process with AI in some way? 

[00:09:35] Fabian Heinrich:Yeah, I think it's every kind of industry where you have a certain complexity within your supply chain or where you have a certain amount of your costs and your gross margin and allocated to procurement.

So what does that mean? I mean, is of course not, not the software industry, neither may be banking and financial industry, but I think it's ranging from energy to chemicals, to pharma, to automotive in terms of complexity. So automotive, by far the most, uh, complex industries. So here you have the most complex processes and data structures within procurement.

[00:10:14] Jens Lapinski: It was interesting in, during the fundraising, the one thing that was critical was this one market overview that you added in at some point, right. He said, this is how we are actually positioned. And then it became clear to all of the investors, oh, okay. This is why Mercanis is so cool. Do you wanna talk a little bit about that?

Because that was, for me, that was a, that turned out to be, in some sense, the killer slide of the slide. 

[00:10:44] Fabian Heinrich: We are talking about on one side the, the complexity of the industry and on the other side, uh, the amount of procurement volume, I suppose. How basically very little venture funded software solutions go after that market of, uh, kind of, complex industries, certain amount of procurement volume.

Because if you think about it, like in such industries have just described chemicals, pharmaceutical, manufacturing, automotive, half of your revenue are your procurement associated costs. So, I mean, that is massive. I mean, like, VISF, I mean 60, 70 billion in revenue, if you think about it, 30 to 35 billion in procurement volume every year. Even like in the mid market, we have companies with three 400 million revenue, 150 to 200 million in procurement volume. So if you only make that a bit more efficient in the recession, in the recession times, that can decide whether the company makes a profit or not.

So I think, um, that is extremely critical. It's also extremely complex because if you, as an outsider look, look into that kind of industry, it might look extremely boring and it might look extremely cumbersome to deal, uh, with, uh, with procurement. I think that is also one of the reasons why so many procurement startups have been focusing historically, in particular on the so-called indirect procurement, or in particular on the spend of software companies, which is funny enough the easiest procurement you can imagine, because tell me about the supply chain of a software company and if you then look of all the venture funded procurement solutions, I would even say 70 to 80% of them focus on software companies. Yeah. 

[00:12:41] Jens Lapinski: Yeah, or other digital players, which is, uh, and obviously the largest input into a software company are the people who write the code.

Um, I think when, when we fund companies, 70% of their money is spent on people. Yeah. So, um, or more. So that's basically, uh, how that works. Totally different question. Going a little off topic, but. Because this is your second startup. Let's talk a little bit about team, in the team setup. What have you, what are you doing differently now in terms of how the team is sort of set up or how you, how you're managing the individual. You know, obviously you have co-founders, but there are other people in the company too. How has your view changed on how you think you, you wanna manage your team, how you compose it, how do you structure it? What are the, like the few top things that pop up in your mind and said, yeah, I've, I definitely have done it differently now because I'm doing it again.

[00:13:38] Fabian Heinrich: I think in the first company, I mean like similar to my time in Rocket, I had a bit of a shiny object syndrome. Yeah. So very much kind of blended by the big logos on a CV of someone. Yeah. And, uh, then thought, hey, that guy saw everything because that guy used to be at, I don't know, a glamorous company before.

That's, uh, was a big learning that that's not necessarily the case. So we have a very kind of targeted interview process where we focus a lot on case studies and also a lot on entrepreneurial skills because, I mean, what, what we are really looking for in Mercanis is that every employee matters and every employee is kind of a entrepreneur themselves because it's not like we have 500 people. I mean, we are now like 50 something people. A year ago we were 25 people. So every person matters so much. Every person is changing the needle every day and every week. If we kind of have everyone with a, with a mini entrepreneurial mindset is uh, really game changing for the company.

I mean, of course that has a certain side effect. We had already, I think like within the last years. Couple or maybe three people leaving because they then decided to build their own companies. But I mean, that's okay. That's, uh, part of the story and part of the journey and, uh, what they can give us is incredible with that, with that entrepreneurial mindset.

Even if they didn't leave after two or three years and build their own companies.

[00:15:18] Jens Lapinski: I had management team meeting at Angela Invest yesterday, and I, it's, they're always quite intense and I think the reason is because every single one of them that sits there has been a CEO of a company before. So, so you have basically lots of alpha men and women around the table.

And everybody is sort of like, you know, if you've had a CEO hat on before you develop a certain attitude towards doing things. You know, you become very, very, uh. You have a problem, you go for it. Right? You debate it quite vigorously. What kind of, what style do you have on the management side?

Are you more on the, are you, because of, what I found very interesting is that on the one hand you say, oh, it's founder mode. You need to be in the details a lot. And then on the other hand it's like, well, all the, all the people or the seniors or the team leads, obviously they want to do their own thing. What have you found? How do you bring those two together? How do you think about that? 

[00:16:16] Fabian Heinrich: I mean, I think if teams or departments are running extremely successful, then I am, I am pretty, pretty hands off and, and give that to the entrepreneurial person. 

But then there are certain areas which are changing the needle and which, uh, I think is also on my shoulders then to drive the needle here. And that's then a thousand percent founder mode. I think at the moment, it's basically two things. I'm focusing on founder mode. One thing is our agentic AI team, but then I am kind of supervisor senior product manager as well, because that really, um, is the vision where we are heading.

And I think if we get the things right today in the agentic AI product team, then uh, that is the success for me in 2027, 2028 and beyond. So I think that's where I can make a difference as a founder today. Similarly, I am, I'm here and there in the, in the trenches, let's say with our new VP sales when we recruit, uh, rockstar salespeople because the new sales talent also matters for the, for the future success.

[00:17:24] Robin Harbort: Can you go a little bit more into detail? What's your vision on this agentic AI procurement? So like how do you want that procurement looks in a few years and what is actually now possible with this buzz word of "agentic AI"?

[00:17:41] Speaker: I think very good question. I mean, I told you where we are coming from for, from those like let's say cumbersome old school, uh, systems where you need to input things. And I think, uh, in 2023, 2024, we were at a state where you had really cool UI UX driven systems, where you had automation.

That's where we are coming from and I think now where we want ahead and where the ultimate vision is, it's not only automating or freeing up time of the, of the, of the procurement people. It's at the end of the day, augmenting them that at the end of the day you have one procurement manager who is managing hundreds of agentic procurement people. Yeah. 

If you think about it in the future, you have a very skilled, maybe even like product, managerial, technical guy who's then the, um, procurement category buyer of the future, and he has then his kind of army of agents. Who has then maybe one agent to negotiate a contract, another agent to reach out to its suppliers, another agent to do him different pricing scenarios, um, and so on and forth. 

So basically that kind of augmentation. That's where we wanna hand in. And I mean, I personally would love that from 2027 onward, so at least from mid 2027, we don't sell any more non-agentic software because I think also the professional procurement will change and will evolve a lot with the advancements of technology and software.

[00:19:29] Jens Lapinski: What do you think that will look like? How will that work if you speculate? Because we won't know exactly, because we don't know the capabilities of the agent systems quite yet and how good they will be and, and how to actually orchestrate them correctly. But if we take, say, in two or three years in the future, which we can kind of see, what do you think is sort of the, the next obvious thing for you? The next obvious step. 

[00:19:54] Fabian Heinrich: I mean, it depends from what kind of perspective you see it, but certainly, I mean, what I already mentioned, the profession of a, of a buyer, of a procurement person, they, they change a lot, we get a lot more like technical driven in order to, to build the workflows, to manage all your, all your agents.

And they also get a lot more analytical. If you think about the vision, I mean, yeah, I mean, I don't know if that happens in two, three years, but then probably the procurement negotiation agent will talk to the sales SDR agent on the other side. The agents will, will, will talk to each other. So, uh, certainly interesting then if the procurement agents will outperform the sales agents at some point.

But, uh, certainly, a race here to, uh, to keep up with. 

But I think like certain relationships with suppliers will be so critical. Also the future and also with more and more supply chain disruptions that you need the human in the loop. Yeah. Because at the end of the day, every business relationship is ultimately driven by humans. Yeah. And I think all that kind of change of technology and the advancements of technology is freeing up a lot of time of the procurement people. So I also think they could focus a lot more. On the kind of relationship building and the development of certain key suppliers, which will then also help them in the future to overcome maybe certain supply chain crisis or supply chain shortages.

[00:21:32] Jens Lapinski: We invested earlier this year in a company that's doing, basically building a platform for agentic payments so that the agents can actually pay and get and and receive cash. Yeah, because this is one of these where... How do you think about that in your context? Because you, you have a lot of volume that gets paid.

I mean, it's a huge part of the global economy that is in the, in the space that you are in effectively. Yeah. All of the procurement of all the chemicals, et cetera. That, that's, that's obviously, it's pretty significant. Do you think this will actually get to the point where these agents, I mean, in a supervised way, but they actually do handshake, they, they do deals and then you say, okay, execute.

Then there's actually an automatic transfer of cash, or it's scheduled in certain ways. It's gatewayed. It's not the way in which it's like, oh, there is a printout of A PDF. It gets sent to the finance department. The finance department, you know? Do you think that actually this whole intersection of, of procurement and finance.

How do you think about that in, in, in say, five to 10 years? 

[00:22:40] Fabian Heinrich: Yeah, I think it depends on the process. I mean, like, we started with the sourcing piece, like that process, uh, is called source to contract. And then now we, we have been evolving and still evolving into the, the P2P, the procure to pay. And I think here, uh, will be I think even within a year a lot of the process is being executed by agents. Such as matching the purchase order to the goods received to the invoice. Yeah, that kind of three-way-match, uh, that is still fairly manual, that is still being done by very dated ERP systems that you can do very easily with agents, but I think after the second p in the P2P to pay, that will still, uh, be handled by the enterprise resource planning systems like the big Oracles or the big SAP systems or, or the respective banking softwares. I don't think that procurement actually touch those payment topics where the money actually flows.

On the other side, I feel like procurement analytics will go way deeper than, than we know because so far they had kind of savings targets where they just looked, okay, I pay price X here and now I wanna, to pay price y this year. If we think about it in those saving targets, you can also incorporate cash flows, Sconto in Germany is a big mechanism which plays a role within the contract negotiation of procurement, but also has a big impact of cash flow, on the organization and of the supplier. So I think there's certainly a lot of touch points, where kind of procurement, uh, works in a symbiosis with, uh, with finance. And I think it's definitely interesting how that will evolve in the, in the future, in particular with agents and in particular will be interesting to see how the traditional ERP systems will evolve, um, in terms of agents, because I mean for, again, for the software and the tech companies, the grade view, ERP systems, I mean, such as Isovalent, Let's see how that will evolve for the industrial enterprises. 

[00:24:57] Jens Lapinski: Yeah. We've also invested in a company that's doing, I think right now on the, the website, let's say it's vibe coding for enterprise, but it's, I would actually say it's hard coding for enterprise. It's literally just, it's literally the, the agents are automatically writing code to connect all of these different systems that you have and partially replace them. 

And so in a sense, you build and there is an overlay, not just of individual systems. So some companies are building overlays over an ERP or a hospital information system, or a, you know, CRM or a, we've got one that's doing it in the hotel space, so that's the property management system or the pause or whatever it's called.

Um, so that you have AI overlay systems for all of these. We've invested in all of them, pretty much I don't think we've, we don't have one that we haven't done yet, but they're also the ones that are then building an overlay that touches almost everything, or it actually connects up all of those different disparate systems, right. Um, so I think. In that sense, the, the procurement system that, that you've built obviously sits on top of partially of different systems. So you touch different parts of the organization, hence the question, you know, because right now lots of these things are siloed. Oh, it's in this, and then it needs to get transferred into this other system.

Yeah, and that's partially still quite a bit of manual process, lots of people involved. It's error prone, it's slow and it's actually complete nonsense because you can't do, it's all unidirectional. You go from A to B to C, but actually what you want to do, particularly when you optimize, is you wanna see this much more holistically.

Yeah. And so I think that hence my question, uh, on the payment side, but it could relate to any of those systems. How many different systems at the customer end does Mercanis actually currently touch? 

[00:26:47] Fabian Heinrich: I mean, it sounds like a lot, but the main system detach is always the ERP system. Sure. Because that's kind of the heart of every organization.

Um, so, so that is the key for the supplier master data that is the key for the purchase order data. That is the key for the payment data and the, the analytics data we need later on. However, with, uh, recent regulatory events over the last years, compliance systems have become also more important, so. I think most of our clients use either one of those three big ones, Integrity Next, Free Wave, or Sapiens.

They have such an impact also these days on the procurement activities because, uh, you wanna buy and source in a compliant way. You wanna respect or you have to respect first and foremost or the ESG criteria when you source from, uh, from suppliers. So we are also fully integrated with, um, Integrity Next and, um, and Free Wave for example.

[00:27:53] Jens Lapinski: Interesting. Robin, last question. 

[00:27:53] Robin Harbort: Yeah, I would like one more question. What is your thought on AI agents navigating. Through those legacy software themself. So do you see also like a scenario where you have the ai which runs through SAP Ariba because it was trained in a virtual environment on the software. And so like the question is, does those legacy data storage just continue to exist and the AI was trained to use it? Do you think this is is possible? 

[00:28:27] Fabian Heinrich: Yeah, I think we need to differentiate on one side, on the procurement legacy systems or the other side on the ERP systems, I mean on the procurement legacy systems, there are, um, already in the US also some here, a lot of players, which build a layer. They said, Hey, we take a shortcut. We don't wanna bid like Mercanis a full, uh, procurement suite. We just sit as a layer on those, um, legacy systems. I mean. I think it can go both ways. Maybe people say, Hey, we stick with those, uh, legacy systems with the 20, 30-year-old code and just put lipstick on the pick and use the layer.

I heard mixed feedback how it works and how it doesn't work. I think. Either way, you are better off with, with having like a state of the art procurement suite, uh, where like agents and, and the suite comes out of one hand. But in order to prepare for both, uh, scenarios, we build our agentic system with a middleware.

So we will be able to connect to any legacy system, also to to scale our agents in 2026, even, even faster. If you think about, it's also a different value proposition because if you go more in the enterprise, well, what we are aiming to do then it gets sometimes quite political because imagine you, Robin, have introduced six years ago SPA Ariba and everyone in procurement is hating SAP Ariba, then all of a sudden someone has the idea. Let's introduce Mercanis. Then when introducing SAP Ariba, you burned already a million in cash, and you paid every year already. I don't know how many euros or millions in licensing fee, so probably it's not a good idea to say, well, that was a bad idea. I wanna introduce SAP Ariba, let's introduce mechanics. But if the pitch of the new solution providers, Hey SAP Ariba greatest tool. We know sometimes it can be hard. Maybe it's on the user. Just let it be sitting there. We have our agentic system layer. We put the layer on top. You can be celebrated as the new innovator.

You don't need to commit a mistake. Then it's easy to pitch. In political enterprise. So there you go. 

[00:30:58] Jens Lapinski: That's basically it, which is why, why we think the entry into all of these enterprise systems is you build an overlay and then there's an opportunity for the overlay to increasingly push down and, and eventually say, why do I even need everything that's underneath? 

[00:31:20] Fabian Heinrich: And then you eat more of the value chain step by step.

[00:31:21] Jens Lapinski: Yes. Yeah, I think that's a, that's a likely scenario and it's then gonna be interesting to see how the legacy system providers will respond. Or how they also interesting to think about how they should respond. Yeah. But it's, um, anyway, that's a, that's a different topic. Cool. Robin.

[00:31:39] Robin Harbort: It was great speaking. 

[00:31:40] Fabian Heinrich: Thank you very much.

[00:31:41] Jens Lapinski: Fabian. Thank you very much. 

[00:31:43] Fabian Heinrich: Thanks a lot for the conversation. Was a huge pleasure.

[00:31:50] Robin Harbort: Thanks for being here. If you enjoyed this episode, support us by leaving afollow and share the Cutting Edge AI podcast. See you next time.