Strategic Sourcing Was Always a Compromise

Hi, welcome to a new episode of Algo Ego. We're going to try a slightly different format here today. I'm going to lecture something I've done at least a few hundred times in my life. I had a lot of things on my chest that I want to get off. I've written about these topics extensively in recent months on a sub-stack on LinkedIn. And here's a spoiler alert, they build on each other. And they're specific to AI and procurement. If you're looking for something a little bit fluffier today, the impact of AI on the job is market or your kid's education, this is probably not the episode for you. If you are interested in going deep on the future of procurement, AI, what it all means, and at least one humble person's opinion who's living this every day, you probably tuned in to the right place. I promise not to pull any punches. These are my opinions, not gains opinions, or any other organization I work with. Let's get going. So I'm going to answer three questions today. The first is, what is exact purchasing? Exact purchasing is a thesis I came up with and created a sub-stack a couple of months ago to really build on what Peter Crailick first wrote about 30 years ago, 40 years ago now in HBR. When he wrote that, purchasing must become supply management. I've got an updated version of that, and I wanted to introduce my thesis. It ties into AI, but it ties into a lot more. Secondly, I want to talk about how this outlook around AI, including exact purchasing, changes where legacy SaaS providers are positioned. And when I say legacy SaaS providers, I'm really picking on the bigger suite companies, like CUPA, Jagger, SAP, Evaluate, and others. I want to talk about this generically at first. Then, and this is the part that will likely get me in trouble with lots of friends in the industry. I want to talk about what it means for a few vendors, specifically CUPA, Jagger, SAP, Evaluate, and Evaluate. I have different opinions on each, and again, these are my opinions. I have absolutely zero-bone in this. I have no economic consideration, no investment in any of these companies, nor am I competing against them in my startups. In fact, in many cases, I could see partnering potentially in the future. But I do want to share my opinion based on having been an analyst and consultant in this industry for nearly 30 years, and having not had the pedestals since I left spend matters to given the opinion on vendors. And now, I can truly be unfiltered, which is a lot more fun anyway. So, let's get going. Let's first talk about exact purchasing. In 1983, Peter Crailick came up with, really, the Crailick matrix as it became known as. Before then, purchasing was very much a transactional process. And what Peter Crailick came up with when he was a consultant at McKinsey, writing us an HBR, was essentially a matrix. The matrix would segment your spend as an organization. And when I say spend, obviously, this is very specific to procurement. This is not about finance, this is not about supply chain. This is about how you buy. In the upper right hand bucket, the matrix he came up with or quadrant he came up with was really strategic spend. Strategic spend had high risk and high profit impact. And Crailick essentially said, "Here, you want to diversify your supply, build partnerships, secure long-term contracts." This was stuff to put it bluntly, it didn't want to f with. Critical spend categories where partnering was quite important. Now, if we move on around the clock on his matrix, the next category leverage was about low risk spend with high profit impact. And here, the cost architecture that Crailick proposed was really about exploiting your buying power, consolidating squeezing suppliers, kind of in the history of procurement, you know, being a bit of a jerk in some cases, throwing around your strength. But in certain areas where you have lots of competition and a lot of leverage, it made sense. Moving on, bottleneck categories. This is really the bottom left of the quadrant. This is where there's high risk but low profit impact. And here, essentially, Crailick said, "Secure supply except cost premiums and find alternatives as you can go about your course of business." But essentially, you have to accept what's going on and you need to avoid the bottleneck if you can. And finally, non-critical spend. Here, this was really about simplifying, automating and reducing overhead. This quadrant led to what became known in many cases as transactional procurement or e-procurement. This was about capturing the transaction in lots of ways and automating as much as you could. Now, Crailick was really far ahead of this time. When I first became procurement consultant, roughly 17 years later, after this was published, well, 16, I guess, to be precise, it was still the Bible of procurement. Everybody was talking about it like it was just published. Now, roughly 40 years on, I would like to propose another thesis. And mine is called exact purchasing. It builds on Crailick. It leverages this great work. We can point to some charts I've come up with in this as we go through it. But essentially, it remakes Crailick's matrix for the age of AI, where we can consume tremendous amounts of data and we can look at spend much more holistically. So, for example, on a high level, I'm replacing the strategic matrix with Crailick for something I call the market risk matrix. So, this says that we need to look at this high-risk, high-profit impact spend from the standpoint, not just of diversifying our supply-based, securing contracts, but really looking at the broader exposure where we can bring finance in, we can look at it as aspects of treasury and how a CFO cares. It's where we have spend, such as metals, printed circuit boards, resin, energy, where we can look at hedging, we can look at indexing, we can model scenarios. It's also where the relationship with the supplier becomes a little bit less important. It's really the relationship with the market because the supplier is just reflecting the market. Now, I know probably what you're saying. The supplier also matters if the supplier for resin, for example, is in the golf right now. That may be a problem. But at the same time, at least in steady state markets, we can look at it from a more financialization perspective. I replace the other quadrants as well. I will go through this and talk about it. But the whole idea now is that we can really begin to update what Kralik built on for the modern era. So, let's go through market risk a bit more specifically. Market risk and the market risk quadrant applies to some of the categories I mentioned. Metals, resin, energy, agricultural commodities, food inputs. These are price-driven, largely by external forces, currency, trade policy, weather, opak, wars in the Middle East, new tariffs that Trump introduces. They're not as much influenced by the classic buyer-supplier relationship or negotiation. And really, the alpha for these categories comes from risk management versus negotiation. It's the art of structuring the deal. Are we going to index versus fix a price? Are we going to hedge? How will we time the market or model scenarios? Here, I would argue and I do that annual sourcing events are really borderline procurement malpractice. You can negotiate a great price, but you can get destroyed by a bag contract or a miss cycle. What really matters here is putting data at the core, external commodity and category information, real-time insights, predictive analysis, leveraging indexes, forecasts, futures curves, freight, currency, trade signals, much more. And really, the key AI input here are takeaways. AI runs this continuous intelligence desk. It watches indexes, it models scenarios. It flags when to buy. It connects input cost to bills of material and shelf prices. So the analyst, you couldn't afford for a category you spent five million bucks on. It could be agricultural commodities. It could be resin, plastics, metals. You name it. That's now an AI employee and they can do the work many times over. And it's where AI is going to play a huge role in managing these complex categories very significantly because it's not about the negotiation. It's about everything that goes into it and making sure you're looking at the market more broadly. Okay, the next bucket or next quadrant, if you will, I describe as cost architecture in the efficient purchasing framework. So packaging, private label food, contingent labor, fleet, product, design, development, all fit into this. The cost architecture category is where I believe the biggest lever is upstream. It's not in negotiating a price, but it's in looking at what gets bought and how it's specified really before the design window closes. Historically, procurement might be negotiating three to five percent on a component which is over specified by 30 percent at the design phase. So you're optimizing for essentially a premise which is not going to get you to where you need to go. You really need to look at how do you influence the whole process upstream? How do you get into it six months before a product is launched, not six weeks before? That's the big delta. So here, what becomes key is internal and external technical information, builds the material specifications, understanding material, substitution, matrices, and cost to serve models. AI, not surprisingly, also plays a huge role here. It can build and maintain substitution matrixes. It can change. challenge specs against competitive benchmarks, it can run shook cost models continuously, it can flag over extensions around specifications before a bill of material gets locked and the product goes to market. It's really the engineering layer that never existed at scale because no human team could maintain it across thousands or tens of thousands or hundreds of thousands of skews. So cost architecture again really is about looking upstream and impacting it with AI. My third quadrant is transaction capture and you're probably thinking this sounds a lot like Crilik and you'd be right, but there's some nuances and differences. So transaction capture really applies to categories like maintenance repair and operations, MRO, office products, commodity IT hardware, energy, billing and the like. So here in these categories, typically the negotiation gets you a rate card or a skew based rate card, but whether you actually capture what you've negotiated is the big question. And typically the answer is not really, it can be death by a thousand cuts. Off-contract purchases your teammates, suppliers swapping out skews which kind of look similar, but they're not entirely mad or sprained invoicing errors. And oh, by the way, some of those invoicing errors may be deliberate, you're never gonna know that for sure, but that's what happens. And then phantom surge charges that keep showing up. The challenge here is historically, if you follow Crilik, strategic sourcing to Claire's victory at contract signing and pops the CAVA. As I talked about in my essay, I say CAVA because procurement doesn't pop champagne too expensive. But for these categories, contract signing is really where the value starts to accrue and where you need to focus. You need to focus after the fact not before. It requires lots of information, high-value trade document data, invoices, POs, shipping notifications, goods receipts, the list goes on, internal data, transactional data, every PO, every invoice, every skew, every price paid versus contracted price, every SSOrial charge, every search charge, every phantom charge that gets added in. We need to go back and monitor religiously. From an AI perspective, this is a huge AI use case as well. With AI, you can audit every invoice, every time, every location. You can flag every deviation from standard in real time. So this is the compliance engine that humans can't run. And they can't staff across dozens or hundreds of facilities and thousands or tens of thousands of suppliers. This is source to PO execution and enforcement in the AI age and it's category specific. The final category in my efficient purchasing matrix is relationship governance. So governance really plays a part in complex services categories, but also ones where services form a major component. So obviously in areas like IT and SAS, dealing with a hyper-scalers, consulting, how you manage Accenture, McKinsey and others, legal spend, BPO spend, and distributed facility services spend like janitorial services, maintenance on facilities, pest control, waste management. Here, these categories historically are defined by relationship asymmetry or information asymmetry. So the supplier in other words knows a heck of a lot more than you as the buyer know. And there's really, you know, essentially two scales taking place here once. There's the enterprise scale with rate cards, SLAs, SW scope creep and consumption based pricing. And there's distributed information outside of centralized procurement or that centralized supply relationship. If you're managing pest control across 600 locations, you need to know if the guy with the truck actually showed up. Why he's charging you all the time. My favorite story here in pest control is actually tied to a goose. Yes, there was a pest control supplier in an organization. I work with that had essentially goose pest control. Now the problem with geese is that yes, they can be pesky and they can be a pest. But they typically only show up at your doorstep during migratory season, at least in the case where this facility was. But the goose pest control supplier was charging every single month throughout the whole year. And when the contract came in to be looked at, it wasn't that the price for pest control was overly high on a one time basis, as that it was being charged all the time. And you had to realize that goose is for migratory animals. The supplier changed their tune pretty quickly on that one. But again, data drives what's necessary. Here, the vendor's margin across these categories typically depends on the buyer not paying attention. At enterprise scale, complexity is really the cover. But at distributed scale, it's about value and invisibility. It's about that pest control supplier who's charging you every month, but only needs to be working two or three months. Data is huge for relationship governance. We need contract terms, rate cards, SLAs, SOWs, reports on utilization, change orders, service verification logs, the list goes on. So here, AI does close the information asymmetry gap for buyers. Why? It can read every single SOW, every change order, every rate card. It can cross reference invoice frequencies against actual delivery records at the location level. It doesn't get tired reading page six of an invoice. By the time you're on, probably the 200th line item. It doesn't mind monitoring what's going on across 600 locations in your retail operation. It really is the governance intelligence layer from contract analytics tied down to the broader source to pay continuum. And man, there's a lot of information necessary to make it work. So my primary argument here, over-arching argument, if you will, is that exact purchasing was always the right answer. It just wasn't economically feasible. You couldn't afford an analyst for a $5 million category like metals, auditing every single emerald invoice line item in other cases, or a contract specialist reading an SOW across every specific detail, or another analyst in purchasing calling store number 247 to check in on what was delivered or not delivered. The human model simply can't staff it. So essentially, under Crailink, we subbed for strategic sourcing. And that was essentially a compromise that calcified into orthodoxy. AI removes the economic constraint. And now with exact purchasing, we can do the work that was always necessary, but we simply couldn't afford to do it with humans at scale. But now we can do it for every category. And that really is the punchline. So now's a great spot to call on my favorite sidekick, Doris Raley, my favorite redhead, who hopefully will have some intelligent things to say on the topic. Anyway, I task him with building this stuff for real, so we'll see if he can make it a reality. Let's see if I can ring him. [MUSIC PLAYING] I've been thinking a lot about data recently. And I don't have to bore you on all the details. I wrote this long paper called "Exact Purchasing." And the thesis of exact purchasing essentially goes back and critiques a gentleman who wrote a paper in Harvard Business Review 40 years ago called "Purchasing Must Be Come Supply Management." I was actually a very good paper. It'd form the basis of what I did for the first 15 years of my career in procurement, taxing, consulting. But I wrote this paper blowing up the notion of what this gentleman Peter Kraylick wrote. And the reason I want to blow it up is not just to create havoc in chaos in general, but I think AI changes the equation in all of procurement and supply chain regarding how we consume data. So we can ingest vast quantities of information all the time, clean it up on the fly, push it out to other systems, and we can do things we couldn't do before. So I'll turn it to you, and I would love your thoughts on as you are building digital employees. How does the frame of data change? Like if we were building what we're doing for humans versus for these always on AI employees. So that's an interesting topic, I think. We are building AI employees, right? But the next hop is that they are working in tandem with humans. So I also need to keep that in mind. I'll give you an example. The AI employees, as I said, they were 24/7. They don't mind being pinged every second or every minute. But I wouldn't want the user, the real human, that manages one of those AI employees, Ben, and not only and so on, to be pinged every minute. So I do have also that human constraint or human experience aspect in mind. I think the job was once in the responsibility of job definition or the ability to make an impact changes now for the humans using the new technology. So in the past, humans and Pokemon, and so on. Head so many dicks, dicks. to be done that they spend most of their time being very responsive and needing to do a lot of the tactical work themselves which means that's time on relationship building and strategy and analyzing and reading analysis reports and so on or going to conventions and enriching their companies with knowledge and insights now suddenly the game changes and so I think the definition of what is a human task to be done changes as well so I'm gonna put a pin on it they employees they you know they walk as you said 24/7 they're reading Ices and spend matters they're searching online and always you know checking every new PO record the European all of the time in many different divisions of the companies trying to find anomalies and opportunities to maybe combine or separate orders and so on so it's a very different way of doing business than what used to be done obviously there was a lot of other than the LLM's and very fascinating NLP 2026 capabilities of agents there's also classical machine learning and classical way of doing procurement analysis and classical rules being embedded in the project as well so the life of an AI employees very very unique I think I'll give you another example so if the companies that we're working with they want to leverage this new technology and the concept of AI employee to enable everybody even from outside of the procurement division to communicate with the procurement AI employee you know so it will enable to it will enable every employee to have a direct line of sight with the procurement code which is you know very empowering I think to the employees and the procurement group another example is now that they have those AI employees they can force and structure what of their processes what I see from working with these guys their earpiece are always filled with very dirty data and tons of errors and duplicated rules and so on so it's very different for as I said for the humans the reality changes as well and I think that what every procurement officer and buyer they need to aspire to also needs to change I hope this answers the question somehow it is from my perspective it does it does I mean as you think about you know the limits of human capability so take somebody with let's be generous twice your intelligence let's assume you're slightly below average so you know quite quite intelligent and they're phenomenal at consuming vast quantities of information right so they're they're good at spotting flaws mistakes they can analyze it they can report on it on just maybe a log scale basis it's probably the way to think about it and the log scale is not a linear scale you know log scale goes for counting in numbers you know one ten a hundred a thousand ten thousand for those what is the log scale difference that AI employees can theoretically consume versus humans and also realizing that as you say we're not going to surface that information to humans all the time because we we as semi intelligent beings are only keep the look processing so much I speak for you there by the way but what is the log scale difference that we're seeing already in the systems being able to to process and I say you know systems as AI employees but what are we dealing with on that scale so first of all I wish we had more time to talk in an i3 send because I've become a little bit trusty I wanted to say that the intelligent joke coming back to your question I think there are some things where the AI employee is just faster you know just ten times one million times faster than in than a human employee would have done that for example reaching out to suppliers right my AI employee can write 100 drafts in a minute out of them following the policy document then they optimize yeah they are there so this is where we can be just faster and obviously you can translate that quickness into money saved time saved and so on and and you can think of it as you know log exponential cooks but I think there's another group of capabilities that are I think of those as zero one you know no one would have done that it's not that someone would have done a bit of it and if you were at multiple they would have done more than that a bit no procurement officer will honor email models and neural networks analysis and so on for every purchase we just don't do that we don't work that way we don't compare every you know ten thousand dollars a request we don't reach out to the requesters and distill their needs and tell them listen you're trying to buy I don't know headphones you didn't request for certification but it's really important the procurement officers they just don't have the domain expertise in all of the things that are being requested so these are just a few examples where it's not a matter of you know dialed the dial it a bit more it's something that would not have been done to is being done I'll give you an analogy from the R&D world we can now use agents to do security tests for our products it wasn't done in the past if you had you know any certification process that you would have need to do some penetration testing but now we can do it every developer can do it it's not something that it used to be something only domain experts could have done you know and now everybody that develops websites or applications or I don't know anything of anything can do it used to be something saved only for the experts of the experts so I think this is a sort of a change that we are seeing right now so if I were to summarize I was thinking more along the lines of kind of one axis which is agents AI employees can consume vast quantities of information on a log scale compared to humans and distillate presented take action but that's one axis the other axis is by nature of what they do they're consuming different information and they're gathering different information not just the same old sets and it's the intersection of those together which really makes the difference yeah I'll give you another example one of my agents in the in the procreteration for frame agreement research one of the suppliers I gave it access to to like 15 minutes of deep research of analysis and it found out there is some relation to some Russian entity meaning there is some risk associated with that supplier risk of sanctions maybe TR risk and this is something that not everybody in the procurement can do or will do for one of their right so so I think it's not it's not always just making things a bit faster or 10 times more okay final question there's a lot of legacy tech out there I know you worked you know in it in your career and I'm not talking about the fun forward stuff like quantum but you know in my world it's it's all the old client server procurement apps your PMRP it's it's single tenets ass multi tenets ass give me the one minute in the door world view on the next three to five years on what happens to all this legacy it's a question I get a lot is this just sunk cost or what happens and I realize this is a can of worms but you know what's going to happen to all this stuff which is out there that's that's an interesting question obviously I cannot tell what what would be in the future but what I think the direction everybody's going now is being very intent focused so I want to buy you know a soccer ball I don't care about Amazon website and I don't care about the competitors website I just want I just want to have I don't know buy tickets for a movie for me and my wife so I think agents and obviously those the big tech they have an interest it will be more of sporadic you know purchases and so on but I think agents will take most of the internet and software as we know it instead of sass like dashboards into a more proactive intent driven outcome driven kind of operations so for example it instead of talking in the air I will talk about us our agents will our area employees reach out to the buyers and say listen there's an opportunity now I let now in metal miner that this goes down this goes up this says we should really negotiate something. You know, it's been very proactive. And if you are someone working for you that is so capable of working all of the time in 100x speed, there's no little need to go into some dashboard. And even if you add some kind of a dashboard from what I experience is that our users, they want to consume data or to interact with the system a bit differently, obviously between different languages, but maybe you want to ask about the value of some commodity and I want to ask about the direction. You know, so a dashboard is very static. I as the developer or the product engineer decided that this is the dashboard. It has a graph and good luck with that. But the cool things now with you know, chat GPT like experiences that I can ask whatever I want. I can ask to see it graph and then I can ask, you know what, show me the same graph, but not different scale show me this graph is opposed to some other commodity and so on. And dashboard usually don't have this kind of flexibility. So, so currently this is the direction everybody is going at. Also, I think European and in order to enter by software is really aimed for the lowest denominator. Meaning they will just a fancy database. Store procedures with story. Yeah, system of record. And I think this is great, but this is not a job humans should do anymore. Like to be the operator that operates this very black box system that stores records. This is not a job for you, but it's a job for a machine. A human should do you know, relationship building strategy decision making. Reverse responsibility taking decisions and not those you know inputs and outputs of data. This is not for us to do. Well, as usual doorhead some informative things to say on the topic. The British might call it interesting. When Americans say interesting, they they mean it when the British say it. They mean something else, but no door had something interesting to say. I'll leave it to you to determine whether I'm speaking as a Brit or as an American. Okay, let's move on. My next topic I want to get to building on on the essentially post-craylick approach to procurement is. How does this AI outlook change for legacy SaaS providers in procurement technology. So we're talking about a huge market here. So, I'm going to be talking about the market. I'm going to be talking about the market. The market is a lot of data. The real challenge becomes what happens to the software companies where the entire product is designed for humans versus outcomes. That's a question I want to get to on the individual vendors in a minute. But there still is life in this market. Forcurement intake and orchestration. Really systems of engagement versus systems of record continue to grow. We see the zips, the oro labs, the level paths, the Tonkeens, the focal points, lots of others. I apologize for those I may have missed in that list. But those vendors which focus on passing the human aspect of procurement and then making sure that all the systems are connected to derive an outcome. Agenteic extensions to legacy suites are also valuable as well. This is really a short term lifeline. It's not a long term answer. The idea here is that the human is not just in the loop. The human is still responsible. But the human has, you know, a bionic suit, if you will, to get some things done leveraging AI. Cooper among others are coming out with their capabilities here in the quarters to come. It will be quite interesting to see if they can pull them off on schedule and the like. Obviously there are a whole host of new AI focus vendors where AI is the primary selling point as well. This is happening in different areas like contract management. It's happening in sourcing. But we're absolutely seeing progress here. I am biased in this regard again, where we're trying to build towards autonomy. Where the product is the outcome. Or the output. It's not enabling the human. But there are other approaches here, which can be fantastic as well. And I want to reference them. As worth consideration to. And then there are data plays, which are doing quite well. Many data companies in procurement across categories. There's lots of providers in IT spend, for example, green cabbage, who I must must disclose. I work with as an advisor and PI. Also a very strong provider. And then lots of others there as well. There are providers for direct materials procurement categories like metal miner, which I'm a co founder of. Others in the bulk chemical space and the transportation space. The list goes on. And then there are those in very specialized categories. Like some I mentioned for the kind of the distributed services management, like waste management, pest control, like fine tune. I also need to disclose I'm an advisor there, but doing great work on the data side. So there are lots of providers again, who can still continue to drive value in this world before autonomy. And it takes shape and takes place. But I think the fundamental challenge and we'll get into it now in a minute as we talk through the individual larger source to pay vendors. Is these organizations who not only set up their product. For a pre AI world. Their evaluations, their balance sheet. The entire structure of the organizations are predicated in many cases on a world that happened before this AI shift. So really the punch line here is that legacy procurement. SAS was built on Peter Creelix, missed up or mistake that there's a universal process for everything you buy. The tech stack encoded that mistake in code. Put a subscription fee around it and call it a platform. Now, AI doesn't just threaten this market. It threatens the entire premise software was built on. The question isn't whether these companies can add AI features. It's whether the underlying model event driven one size fits all stage gate process centric survives contact. When we have continuous category specific processes with with AI native alternatives that can do the work of purchasing managers purchasing associate's analyst. At a fraction of the cost of humans with infinitely better data. Now there's a big question on how quickly we get there. But I do think it's one worth posing. Okay. Now we're closing in on the last section. A lot to talk about here. I want to get into specifics around SAS vendors. I'm going to talk about Cooper, Jagger, SAP, Arriba and Evalua. Again, I have no bias in my reporting. I have no reason not to be objective. I'm not a shareholder. I'm not long. I'm not sure I have friends at nearly all these companies. And I hope they remain my friends afterwards. I'm just calling calling balls and strikes that some people would say. So here's the scenario that's keeping the investors of these companies awake at night. AI agents are already starting to reduce head counts of companies. I think that will evolve to AI employees. And the challenge is that fewer employees as Jason Lempkin has pointed out. And I'm going to quote a few other points from a mere fewer employees means fewer software seats. Historically, a lot of these companies were valued very highly around 2021. Many were valued as highly as 20 or 30 times. revenue. And more recently, a lot of these companies were acquired at 20 to 40 times Epida. That may be too rich for the AI world. A lot are loaded with debt, especially if they were taken private by private equity firms or leverage bio firms. And many of these organizations, you know, are subject, again, based on the debt to private credit. And we see some of that private credit with with with Carlisle and others exchanging hands now at a fraction of the price that it used to reference some material here in a minute also. So what does this mean for the individual companies? Cooper. So when Tomapravo acquired Cooper, the deal closed in February of 2023, it was roughly an eight billion dollar enterprise value transaction, including 2.6 billion in leverage debt, which was comprised of a 2.44 billion dollar seven-year term loan, plus a hundred and fifteen million dollar revolver. It was led by Oak Tree Apollo and Blackstone six-threate with HPS, according to pitch book, on the debt side. And that resulted in a debt to equity ratio at the time of the transaction of roughly 65 to 35. At the time of the deal, Cooper was just under a billion dollars in revenue, roughly 8.4x times EV to 2NTM. And by 2024, Cooper had hit a billion dollars in revenue and continued to grow. So that's the good part. Here's the bad part. The organization to roughly 2.6 billion dollars in debt, one to 1.2 billion in revenue just after the deal closed, or about a 2.2 to 2.6 debt to revenue ratio. If we do the math on this, and we look at at EBIDA, EBIDA was not disclosed, and obviously there was some cleanup in terms of staffing and everything else that was roughly by my back of the napkin calculations, a seven to ten times debt to EBIDA ratio, which is heavy, and it's heavy for software PE. They may have bought some of this back with efficiencies more recently. And again, this is my speculation, but it is high. It's so high that Jason Lemkin and his SASTRA newsletter listed, Cooper, as one of the potential organizations at risk in terms of this SAST by outworld based on what's changing with AI out there today. So if we look at the software, debt bomb, as Jason Lemkin calls it, Cooper is kind of positioned right in the middle of it. It's not the worst. There are companies with a huge debt load like Citrix and Anna Plan, Warsaw, and Cooper. But Cooper is still very much in there compared to others, like Avallara, Phinextra. And vendors from across the enterprise software landscape who didn't take on quite as much debt from a ratio perspective. But again, quite a large deal. And not referencing Cooper specifically, but Carlisle and Blackrock, according to Jason Lemkin, have begun buying discounted software debt. So we see a secondary market for the debt put on these deals, which essentially means you can discount the valuation as well. Now, Cooper's continued to grow. Yes, there have been some stumbles in terms of pipeline from what I've heard at the end of last year. And we've seen, and I've written quite a bit about this on LinkedIn, some changes in customer set. You can even see customers commenting on my post as well, referencing that too. But again, a leverage deal done before the AI era took off. And certainly a lot of potential to keep growing, but also risk from a shareholder perspective as well. So watch Cooper closely. I was a huge fan up until the time of the deal. And then really as long as Rob Bernstein remained at the organization, but there have been a lot of changes since he left. And I think the cold rob was very real in that organization. And without Rob, it certainly has faced some headwinds, which it didn't face with him. Okay, let's move on to Jagger. Jagger is in a Malgum. I remember Jagger going back to when the company started Bravo solution and cyclist. And they added others like pull for tool. It's traded a couple of different times in the market already. Most recently, Vista equity acquired it from Syndvin at a target valuation of roughly $3 billion, including debt, which is roughly double the 1.5 billion Syndvin paid in 2019. Kind of if you look out there in the ether, roughly 300 million in revenue, I think I suspect it might be slightly more than that. But at least if you go online, that's what you find with significant EBITDA, a very profitable organization, likely well over $100 million in trailing EBITDA at the time of the last sale. Now what's interesting if you look at the data from that last sale online is that $3 billion included debt and that language is key. So that's enterprise value versus equity value. The deal structure wasn't disclosed by Vista, but at 3 billion enterprise value on 300 million in revenue, that's 10 times revenue. But if Vista used typical leverage on the deal, say 40 to 60% debt on EV, we're looking at roughly 1.2 to 1.8 billion in debt on 300 million in revenue, which buys back the valuation down to about 4 to 6x debt to revenue. So in EBITDA, roughly 125 million with 1.2 to 1.8 billion in debt, that's a 10 to 14 x debt to EBITDA ratio, which is aggressive. Now again, since they have a lot of public sector and university clients and hopefully some very solid manufacturers, especially in Europe as well, if customers keep signing and renewing, that's great. But if you have that level of debt, even with such a strong cashflow to service it, there's potential risk. Now, Vista's playbook is well known, operational efficiency, margin expansion, bold on acquisitions. I don't know one startup that's been around for more than a few years, which has not been called by Jagger's corporate development team, which is smart. And they know they need to grow into the debt, which means either accelerating revenue, margin expansion, or both. And M&A is likely to be a part of it. In Jagger's defense as well, they've announced the JAI, Agentec push, and public sector is a major key. I tend to believe public sector revenues are more durable than private sector revenues with AI. But there you have it, Kooplin, Jagger, both strong legacy SaaS companies, but now strong legacy SaaS companies competing in an AI world, both with significant debt on the balance sheet done at a time before the AI revolution. So those are certainly two interesting companies to benchmark and look at. I want to go to SAP Arriba, which if we look at the growth of some of these other companies, SAP Arriba or SAP Arriba, as some call it, is probably one of the more controversial ones. Historically, kind of post SAP deal, Arriba had a negative net promoter score. You can look online and find this. And a whole bunch of reasons for that revenue models with supplier fees, limited innovation, and the like. But, and this is a giant, but SAP has been very, very busy rewriting Arriba from the ground up on SAP BTP, their business technology platform. The first release has already happened this year, with updates throughout the year as planned. This is not a refresh. It's a complete rebuild on the 25-year-old platform. There's AI agents, what SAP calls their dual agents embedded. There's a bit analysis agent, which looks at total cost. There's AI supplier response summaries and contracting capabilities. At the time of my notes for this podcast, there were roughly 34 for active, dual agents or prompts available. And there's more coming. Now, the negative. Roughly only 3% of SAP customers run SAP Business AI and production today. So we're looking at essentially a solution for something that is going to get sold on many levels. And procurement may be the guinea pig for it. SAP has got a big advantage. You know, it is not levered up as an organization like CUPA and Jagger. This rebuild is about the farm move, but the cash loads have been really incredibly durable for Arriba under SAP for a long period of time. And this really kind of ground floor build-up, especially around direct spend, could be interesting. I'm certainly keeping my eyes peeled and looking at it. Next, I want to turn to really, again, I don't want to play favorites. I have no reason to play favorites, but let's call them my favorite unique source to pay vendor who I think could also survive a transition in the AI world. And that's a value of, I've known a value of since it was essentially a 50 or 100 person company, came out of France, founders moved to Silicon Valley, and had their own belief structure and what to build. It was not just a multi-tenant SaaS. It was single-tenant in certain cases, and essentially a lot of custom configuration work for specific needs. A value is still growing organically. They have not levered up the balance sheet. It's founder-led unlike CUPA and Jagger today, and certainly like Arriba talking about founders who have been in the business for 20 years. It's a recent event drew roughly 1,500 attendees, and they are moving to a genetic in AI as well. So they're pushing humans plus agents plus data on one data model like Arriba. Reba, they programmed over 30 different use cases so far on the AI agent side with Azure claiming no code customization. And they've got a lot of autonomous things on the roadmap. Now, whether they get there or not, we'll see. But again, of really the legacy for based on what they have today, they're likely the best positioned. Not having debt on the balance sheet certainly helps, but having an engaged founding team still deeply involved also matters as well. Now, you might say some of their stuff again is legacy. It is quite old from an architectural standpoint, but again, they are making the pivot to AI. And I think they're going to be quite interesting to watch. I also think Reba under SAP is going to be very interesting to watch on this ground floor build up, but there's simply a lot we don't know yet. And we're still waiting for the case studies and for those organizations who want to go first. So I know we covered a lot today on these different vendors. And again, I want to state for the record, I am not advising any of them. I'm not working with their competitors. I'm just observing as I used to it, spend matters as an analyst. But again, I don't even have the data of spend matters anymore. So these are my opinions. So discount them as you may. But I do think we are in for a huge transition when it comes to the source to pay ecosystem. And it's not just a product transition because of AI. It's a financial transition. And we need to consider the balance sheet of these vendors. We need to consider the debt leverage and the debt ratios and whether they can keep growing despite the changes and valuations of SaaS companies. And that debt trading at a discount, which may be representative of how investors are looking at legacy SaaS overall. Thank you for tuning in to another episode of Algo Ego. I'd like to thank Gain as always for underwriting this podcast. Algo Ego is a pulled-who production. It's produced by Yo-Haim Atel and Alma Benzav with executive producer Daniel Zeus. The original theme music is produced by the amazing Gal Lev. Until next time, thank you for tuning in.