The Data Problem That Kills PE Exit Multiples

An easy way to think about it is you have some sort of asset, right? So you might have a factory and you might make a car per year, one car, right? Now imagine you have the same asset and that asset can produce two cars per year. You have twice the turn, which means you're twice as capital efficient with the same asset.

So it's just the velocity of turning the asset and making that sort of work harder that's really kind of the notion behind the capital efficiency. So for argument's sake, imagine you're paid 100 days later than you actually sold the car. In that case, you're waiting for the cash to arrive. If you had that cash right away, you could have been doing something else with that cash that earns higher returns.

So all that accumulates and the opportunity cost of not getting paid faster basically has a couple of different impacts. One is that you're losing that opportunity to deploy that cash so that you could be investing either in your own operations, in your own business to grow your business. You could be investing in, I don't know, stocks, bonds, some other market asset that's gonna yield you higher returns. You could be letting it sleep in a bank deposit.

But having that cash allows you to have that flexibility to do something and earn returns on it. In this age of AI, if you have proprietary data, that becomes almost like a moat. But if you're just sitting, if it's just not, you're not creating a moat out of it. It's just like dirt sitting around.

You can't really assetize that. Again, to invent a verb, but you can't turn that into a valuable asset. So the question is, can you make the right investments, not just in the technology, because that's, as you said, relatively affordable now, but the whole governance, everything around it that makes it work as, as wheels, machinery in a company. Behind every value creation plan, there's a data problem nobody wants to talk about.

Fragmented systems, metrics nobody trusts. And decisions made on gut feel dressed up as analysis. Welcome to the PE Data Guy. Each week, host Graham Crawford talks to the operating partners, advisors and practitioners who are doing the work inside portfolio companies.

If you care about what actually drives returns in the market, then you're in the right place. The PE Data Guy starts now. Shota Ishii has spent two decades building systems that help companies see something really important, where their money actually goes. At Prosimo Tech, he works with PE backed manufacturers who often can't answer a simple question, what products are actually profitable today?

We're talking about why that gap exists and what it takes to close it before someone else closes it for you. Welcome to the Shoshoto. It's so great to have you on. Thank you so much, Graham.

Good to be here. It sounds like a fascinating role that you have and I know that you're operating primarily out of Japan at the moment. Describe to me the route, the journey that you've taken to get there. People always have many fascinating stops along the road and I'm intrigued to know what you as well.

Sure, yes, absolutely. So I studied applied physics and AI back at university and. But that was not really AI was not a career back then, unfortunately. So I had to sold away.

Ahead of your time, Shota. That'll give away my age. But basically, yeah, so I went into quantitative finance. So my first job was modeling corporate credit risk.

So I had to learn how to read balance sheets, but also look at it from sort of a market investor perspective and understand how to, how to understand capital structure, et cetera, and optimize it. And so that was sort of my first job, that company. Eventually it was a company called KMV and that was eventually sold to Moody's. And then after that I was one of the early members and I didn't know what to do after that.

So I went to Europe and hung out for a year getting an envelope. Always a good thing to do, right? So, so, and MBA is academically content free relative to AI and physics, but was fun. And then from there I actually acquired my first company.

So it was a very small company and I had to sort of run that for about five, six years. And that was sort of a disaster because I wasn't a mature entrepreneur at the time. But I did learn a lot about corporate finance as well. And then after that I went into a hedge fund that was eventually sold to Blackstone Credit and then I went to a US based bank where I basically was helping set up one of these innovation labs in San Francisco for AI machine learning in portfolio finance.

And that's when I kind of began to sort of strike out on my own because I saw a huge gap between the language of corporate finance and market finance. And CFOs didn't really know all the time how to communicate with some of the investors. And so that was when AI was maturing again or kind of coming up again. And that's when I began to think about possibly creating almost like a robo cfo.

And then that was how my journey started. I mean, everywhere you go, companies get bought. That seems to be the pattern to that seems to be the pattern to me. So certainly, you know, success.

If you want a successful exit, make sure Shilter's part of your company. I'm sure it's no coincidence. Well, no, I think two, three points doesn't make a statistic, so we'll have to see. Well, you know, we can get onto statistically significant data later in the episode for sure.

But yeah, what an incredible journey and certainly more insightful insight the most into what makes successful exit, what makes a successful company. And I think you mentioned a disaster along the way. I think is always a critical part of any successful career is some sort of mess or disaster. And I've no shortage of my own that I could share.

So I like to, I celebrate them being worn as a badge of honor and offering incredibly informative educational experiences and instructive in a way that no MBA or manual or entrepreneur podcast can ever be is sometimes there are those unteachable lessons out there. Oh yeah, totally. That always land better, you know, based on. Based on experience.

So tell me more about the Robo CFO. Is it RoboCop style with four prime directives and uphold the law and all that sort of stuff? Like what's the idea for that and what sort of form has it taken now? So, yes, so the idea back then was actually I was.

So my client was actually an investment bank. And so, you know, the idea was that the first top 200 deals would happen by relationships, so wining and dining. So this is like M and A, but below the, you know, the top 100200 deals, they didn't have the coverage to actually look at all the companies. And so the idea was, could we use, can we scan publicly available data as well as social media data, other sorts of data called alternative data.

So sort of the data exhaust, if you will, of various companies to look for M and A opportunities, whether it's kind of a diversification play or concentration, vertical integration play, et cetera. And so this was kind of earlier on, in the days of AI, we didn't have kind of the LLMs right now that we have, so we had to actually, we were using versions of natural language processing, which are a little bit more primitive, but we were able to scan through and glean a lot of insight from the securities filings, the annual reports and things like that.

And so that was originally kind of the tool. And then so then the investment bank said, well, can you make a robo cfo? And so we said, yeah, probably. As someone, yeah, why not?

Let's try. Why not? Let's do. And then I realized that actually CFO is obviously, obviously do many, many things.

So I had to kind of refocus around what made most sense. And one of the opportunities was that sort of confluence of one private equity, private credit was kind of becoming, was ballooning at the time. And so there was an opportunity to look at port cos and possible port cos. So that was one macro story.

The other was that, well, out of all the things CFOs do, whether it's M and A or share buybacks or capex, the thing that we could really address in terms of the tooling and the toolkit we had and that was data intensive, was around working capital. So payment terms, inventory, things like that, because that gets into the operational data and it's quite low typically on the priority list of things that they have to do because it's kind of a difficult thing to do. It's faster to just buy back shares or do something like that.

But that's where our capabilities around data and analytics really could shine. So then we decided to just refocus and pivot more towards short term capital efficiency and maximizing that. Right. And we've talked, I mean you mentioned it just now and we've talked a couple of times previously about the need for clean data and a clean platform that holds it.

Without that, what problems do people run into? And with that, what sort of stuff does it enable? Right. So we ran into a couple roadblocks on our journey to try to get companies to improve how they leverage data to improve portfolio, sorry, company value.

And one was actually just oftentimes the financial kind of literacy of the CFOs. So where their priorities sit and how much they understand about cost of capital and the opportunity cost of capital and things like that around working capital. So some people are very, very sophisticated, but there's a huge variance around that. So getting sort of the entire organization, not just the cfo, but the people who are actually going to move the working capital, this is the salespeople, procurement, purchasing people, people in manufacturing, production.

Everyone has to sort of understand why that's an important set of metrics to improve for improving free cash flow. So that's one blocker. And the second was around again, what you mentioned, which is the data platform. So it's like, okay, I get it, in theory we should shorten the payment cycle from our customers or maybe delay payments to suppliers if we can, without getting them upset.

And the challenge there was, okay, well I understand that in theory, but my data is a Mess. I don't know which products, which SKUs are profitable, you know, which is slow payment, fast payment, it's kind of there. I have Excel sheets, but it's all over the place. Maybe, you know, I've acquired a couple companies so, you know, by different subsidiaries or different factories, may have different standards and different types of data.

So it's kind of a data mess that, you know, Graham, you're quite familiar with in dealing with. Yeah, very much so. Very much so. Back to.

You gave a relatively 101 explanation of it, but I'm someone who should never belong in a CFO role, so I would certainly be on the end of the CFO spectrum of CFO who doesn't fully understand the thing. So for anybody else in my position who's listening, what benefit does it bring to the enterprise value of the company and the return on invested capital that people are getting from the approach of pay a little more slowly, get money in a lot more quickly. And then the other thing again, that, you know, I've talked about is be more efficient with the.

With the inventory that you keep on stock. How does that flow through to return on invested capital for PE firms? Yeah. So, you know, we could get into sort of the typical MBA talk about, you know, return on invested capital, return on working capital, et cetera.

The way I like to think about it is, as a company, you need cash to operate, you need cash to pay your suppliers, you need cash to pay your employees, et cetera. Right. And there's a time gap between when you pay and when the actual payable accrues. And same thing on the sales side, when you actually book a revenue.

So you've sold a widget, a car, whatever it is, and you've sold it and you book that revenue. But you might be getting paid in 30 days or 100 days. So for argument's sake, imagine you're paid 100 days later than you actually sold the car. In that case, you're waiting for the cash to arrive.

If you had that cash right away, you could have been doing something else with that cash that earns higher returns. So all that accumulates. And the opportunity cost of not getting paid faster basically has a couple different impacts. One is that you're losing that opportunity to deploy that cash so that you could be investing either in your own operations, in your own business, to grow your business.

You could be investing in, I don't know, stocks, bonds, some other market asset that's going to yield you higher returns. You could be letting it sleep in a bank Deposit. But having that cash allows you to have that flexibility to do something and earn returns on it. So that's one thing.

And that ties directly into free cash flow. And of course that ties into kind of company valuation. And that's really the definition of capital efficiency. Because if you think about irr, the whole equation on the numerator side is do you take the value from 100 to 110?

So that's a 10% improvement. But in the denominator, does that happen in one second or one minute or one year? And so the shorter the time and the faster the turn, the higher the return on capital and the higher the capital efficiency. Right.

Or 10 years. This is happening. Yeah, we're 10 years. Right?

Yeah. In some cases. In some cases, yeah. That makes sense to me.

I think the analogy that popped in my head, I've never run an E commerce business, but is used a lot as examples in many of the books that I've read. It's if you can get paid immediately and you're in a position where you've got a positive return on ad spend, then that allows you to drive more money into ads to then keep the flywheel of the positive return on ad spend turning a lot more quickly than if you wait 30 days to get paid. That's right. Yeah, that makes sense.

And also just, I mean, an easy way to think about it is you have some sort of asset, right? So you might have a factory and you might make a car per year, one car, right? Now imagine you have the same asset and that that asset can produce two cars per year. You have twice the turn, which means you're, you're twice as capital efficient with the same asset.

So it's just the, the velocity of, of, of, of of turning the asset and making that sort of work harder. That's really kind of the, the notion behind the capital efficiency. Got it, got it. That makes, that makes a lot of sense.

And you know, I, I spend a lot of my time buried in, in, in data and people's data architecture. Typically what sort of data management or data architecture principles have to be in play in order for this to be a good platform for your work to be successful. Right. Okay.

So the problem is if you want to create cars or widgets or balloons or whatever, and you want to increase your kind of turn on the same assets, but you have 1,000 SKUs, so you have a red balloon, blue balloon, green balloon, yellow balloon, et cetera. Again, you might have bought another balloon factory over in the other on the west coast. Or the east coast or whatever. So typically you get into problems of master data management, of managing all the transaction data, all of this.

So unless you have a data platform that allows you to capture the granular transactions in a timely way and also to monitor kind of what's going on at every single SKU level, you're just going to have very broad strokes view at the end of the financial year or maybe quarterly. And that gives you average values, but it doesn't get you into the kind of the transactional how much am I getting paid by whom, how late or how early per customer, per product, per region, per subsidiary.

And having that ability to rotate through the different dimensions of analysis is important. So it's really kind of what you're, what you're trying to do is improve the operations and improve the return on working capital by getting the insights into what I just mentioned across all these different dimensions of analysis. But to enable that, you do need that data platform, right? Yeah, and that makes a lot of sense to me.

And in fact, I love the balloon example because I've got a, I have a live one right now. So my, my oldest son turned 16 a couple of weeks back and I went to the, the, the dollar store to get some balloons. And yeah, high recommendation for dollar store balloons by the way. They last forever.

Like four weeks later, they're still sort of bouncing around the ceiling somewhere. Anyway, so I ended up having to get a square happy birthday balloon because in the number balloons they had all the numbers. And I'm not sure, Shota, how familiar you are with what teenagers are talking about these days. But they had no sixes and they had no sevens.

So if I'm a balloon manufacturer, then assuming I'm not tuned into the latest, the latest memes that all the middle schoolers and teens are talking about and yelling in five guys, then I would have no idea that was happening. Right. And I would have no idea that my sixes and sevens are suddenly flying off the shelf and are doing way more volume. I might know, I might observe, oh, the overall volume of balloon sales has gone up, but without knowledge by number, I'm not going to understand why.

Right. And you know, presumably if you, if you then say to someone, you know, you're a middle school intern, do you know why I've run out of 6, 6 and 7? They tell you very quickly, but if you're not able to do that, then, then you can't get there. And I talk a lot in my work Shilta about the difference between management grade reporting and investor grade reporting.

And I think you're talking about the latter here because you're having to analyze cash flow through the business through the eye and lens of an investor in the business rather than someone managing it. And it's perfectly possible that in the balloon example, the balloon business can continue to operate very well and serve its customers. But it's missing that critical layer of insight that might hold a key to something that could enhance profitability, that could make the business run more efficiently.

Because they don't have that SKU level data and they're not looking through the lens of how can I make this business more profitable? And instead sticking to how can I continue to give my customers what they're asking for. That's right. If you think about a company like Amazon, they're kind of digital native, they're digital first, so they by definition have all the data, which is why they're able to recommend the next thing to empty your wallet with.

Right. So very good at that, very good at that. And you know, so they have that SKU level data, they have it real time. You know, all these transactions are happening very, very quickly.

So they almost have a streaming mechanism to capture that data. And what, you know, the companies we work with a lot happen to be oftentimes manufacturing firms. So they don't have that Amazon like setup. So all we're talking about, and if you look at the multiples on Amazon, it is what it is and they're able to grow because they're able to have that data and sell more based on the data.

But if you think about the. And then invest the profits in data capabilities. That's right. And so it's not just a nice to have, it's a key foundation for the rapid response and growth to market conditions, to Covid to disruptions, whatever it is.

And our manufacturing clients, some of them, some of them are very good, but others, again it's not their fault, but they just were in an industry that didn't require this level of real time granularity. And so what we're trying to do is really Amazonify, if that's a verb, some of these legacy industries, right, and make it real time, make it data driven, make it granular. Because why not you now have the tools and AI and data platforms that allow you to do that easily. And I think when I referred to one of the roadblocks, one of the mental blocks is the CFOs or the management committee of these companies don't realize how easy it is to actually implement some of these.

Well, I'm not trivializing by saying easy, but. No, no, of course I understand. It's definitely possible with the tooling we have today that's available to actually get things moving on the data front such that they're closer to an Amazon type of operating mod. Another thing which falls down the priority list is this investment in technology in the mid market.

Often companies that have, especially companies that have recently ascended into this mid market space, they've made it right, they've got a big, profitable, successful company compared to where they started and there's a perceived penalty on change and they want to protect the revenue that they're currently getting. So they're not intentivized to like hey, let's see if we can spend some money on technology. But I find there's a bit of a misunderstanding about just how accessible and affordable some of this technology is.

And I've spoken to clients across different industries about AWS Cloud, Google Cloud and then the database technologies which might sit on top of them. Snowflake databricks, you know, choose your poison. And the initial reaction is oh, those are really expensive enterprise tools or like saved for the Fortune 500 and nobody else. But I've had multiple clients, three, four clients now, who have thought, oh, Snowflake's going to be really expensive because we have a lot of data.

But a lot of data is a very relative term and I don't even think the numbers that you would put next to Amazon's data would even make any sense. When a number's so big, it becomes nonsensical. But hundreds and hundreds of millions of rows, you can host and run that on any of those platforms for a couple of thousand dollars a year. It really is affordable.

There really are entry levels to the market and the technology has been democratized to a large extent, especially around storage and processing of data. So that's a place that I always find valuable to start. And you can get big returns for a really small investment I think in that particular area. And the cost of not investing is going to get higher and higher partially because, well, first of all it's so easy and those that will, will move faster and generate those returns.

And also there's other peripheral issues coming up now. Like you want to hold data as an asset. Data itself is a valuable asset, like a brand. Absolutely, for the company.

So in this age of AI, if you have proprietary data, that becomes almost like a moat. But if you're just sitting, if you're not creating a moat out of it, it's Just like dirt sitting around, you can't really assetize that again, to invent a verb, but you can't turn that into a valuable asset. So the question is, can you make the right investments not just in the technology because that's, as you said, relatively affordable now, but the whole governance, everything around it that makes it work as wheels, machinery in a company to turn that into a credible asset that's going to improve free cash flow and valuations.

Right? Yeah, technology. And another, my favorite phrase is technology and data are never a business case in their own right. They have to have something behind it.

They have to have some profitability or some objective in mind. And often, you know, clients make the mistake of me of starting with a tool based conversation. Right. They think the secret sauce that my team are going to bring to them is like, oh, here is the right combination of database tools and ETL tools to use.

And that's the thing that's going to unlock the value when in fact the tool choice is often completely unimportant at the level of, you know, complexity that we're dealing with. To start with. And it's much more about can we standardize and agree terms? Can we standardize and version control the business logic by which we are any number that's arrived at deterministically, like who controls that business logic?

Can we all agree it's done the same way? What are the key performance indicators of the company? How are those calculated and how do we contribute to them? And I think those things are often the first, the right first step before technology even comes into play.

Because without an understanding of how the tech or the data is going to make a difference, then, you know, you're often throwing, you're throwing darts in the, in the dark and hoping to hit a bullseye. Oh yeah, I mean I had a, I had a client who, so you know, they, they're a larger mid sized company, a couple hundreds of millions, right? Not, not huge. And so, and I think they hired a consultant and they were, you know, and the consultant said, data breaks or you know, snowflake or whatever.

And the CFO said, well, I don't know any of these, I'm a little bit afraid. And then we started talking and he basically said, well, you know, why don't we, I mean, I kind of coached him along in this thinking, but I said, okay, fine, you're afraid of all this new vocabulary and words like what the hell is snowflake? And so, you know, he said, why don't we just put on training wheels so for particular subset of the company, so it was a subsidiary, you know, let's just put in a postgres database and then like start capturing the data and then creating rules around that and find.

So this and this and then this got them to see the possibility of like, wow, okay, data does. Is doable and manageable and provides value. Right? Yeah.

The. Another great phrase that I've seen around the holes that I run through is if your data strategy consists of a list of tools, then it's a shopping list and not a data strategy. So I'm very deliberately, you know, not committed to any particular tool or platform as an answer because often it focuses attention away from the most important thing in the first place of like, what data should we collect and what are we going to do with it once we've got it and how is that going to.

How is that going to add value? What's the kind of. Often in these situations you can make a really small investment and see a light bulb go on or see a big change. What's the smallest investment that you've seen create disproportionate value for the private equity firm in the portfolio company that you're operating with?

This was not a portfolio company of a private. So we work with companies that are port cos of pe, but also independent companies that are kind of targeted by activist funds as well. Right. So similar story, but in this case it was quite interesting because just as an example, they did not have a data platform and they're still implementing the data platform.

But we said let's in the first three months focus on finding pockets of cash that can be freed, so unlock cash asap. And the reason is that basically they were under some pressure to kind of release cash and reinvest that into something that was a growth area. And anyway, so we basically looked at per again, across all these dimensions, per customer, per invoice, per supplier, per invoice, look at all the inventory as well. And this is for kind of a metal company.

And what they do is they buy metal and they transform it in some way, bend it, stretch it, et cetera. And we basically were able to get really granular into the data per transaction, per site, per whatever dimension, identify which were the good customers and the bad customers. So we almost had a heat map of where we can focus on, on improving what's called the cash conversion cycle. So this is an important metric in working capital which tells you how quickly you're turning sort of payables or sorry, receivables to cash and how quickly you're turning inventory to cash, how slowly you're paying your suppliers, et cetera.

And we really kind of identified and zoomed in on that and we were able to identify again, this is for a $400 million company, about $80 million in savings along those dimensions just very, very quickly because we were able to stitch together the data and then run an analytic on top of it that looked at sort of relative to competitors, how quickly you should be paying, what's the negotiating power you have. And again, that involved bringing together the salespeople, the procurement people, production people, and then sort of getting, getting the data not only from the Systems and the ERPs, but also from the CRMs, understanding short term forecasts and seeing where they think they're going to land and then being able to calibrate their negotiations based on the projected demand and say, okay, well if you want this metal faster or delivered faster, you're going to have to do this and change the pricing.

So, you know, it becomes sort of a series of quick win negotiations that the company was able to capitalize on. But they had zero visibility into any of this. For example, just one concrete case, they were not charging for the storage fees of this chunks of metal in their inventory. So the client would say, okay, we'll come pick it up in whatever, 30 days and somehow it's forgotten.

So it's lying there for 90 days or 120 days or 200 days in some cases. So you have this long tail of inventory that's, that's not being picked up by clients and they were not charging their customers for it. But that visibility at the transaction level was really kind of difficult because again, per each transaction, our customer and their customer, they were just having phone calls and saying, yeah, we'll just keep it in storage for X more days, we'll come pick it up later. So that really wasn't being tracked, for example.

Yeah, and it's always a great place to start of with what, you know, what decisions are difficult right now? Difficult and also important. You recognize they're important, but you find them difficult to make because you don't have enough information. And I think you talked there about almost another layer to that of here are some decisions that you don't even have the data to consider that there might even be a decision to make.

But allow me to unlock a whole new bunch of decisions, each of which could be strategic levers to unlock, you know, additional working capital in the, in the business. That's right. And you know, I think, and once we go beyond the basics. So, you know, this was kind of a fire drill that, that happened in, you know, over three months.

But then beyond that, now what they're beginning to do is to create what they call sort of an intelligence package for their customer. So this goes back to kind of what I had in mind for the robo cfo. But what they do is they use AI, for example, to collect intelligence, news, financial reports, credit scores, etc. About their customers, their clients, and then, and then consolidate that into a data platform.

So then they're able to see what the strategic priorities are of their customers are. And then on that, on the back of that, kind of assess the mutual negotiating power and then look for win win transactions. So maybe it's a customer that wants to get, that is willing to pay faster, in which case your customer pays you faster, but you can give them a discount. And so this gets into all sorts of financial calculations about the cost of the relative cost of capital and how do you actually whether you can pay someone faster for discount, et cetera.

That gets quite interesting. But you need the data to be able to do that on an ongoing basis. Yeah, absolutely. And that's certainly something I see a lot of founders offering in the first couple of years of a company.

If you can pay up front, then we'll knock 10% off because that gives it access to critical working capital that they wouldn't otherwise have, especially if they're bootstrapping. I know we touched on it briefly at the start, but you mentioned AI there and I think we probably set a 2026 record for a podcast that we're 30 minutes in and AI hasn't come up in conversation. But since you mentioned it and you know you were involved in AI twice before it was cool. Like not just a little bit before it was cool, but a long way before it was cool in some of its earlier forms.

It's a big thing that's front of mind for a lot of people. My guest last week talked actually about how particularly in the mid market, people are kind of a little bit frozen with AI. And there are certain elements of being a mid market company that make it hard, harder to adopt. Like you haven't, you can't just bring in McKinsey and Bain like the big enterprises are going to do.

And I think on the other end of the spectrum you have, well, folks like us who are in the founding stages of companies who desperately need the operating efficiency and can't afford not to use AI. And then in the middle you've got Mid market who, who seem to be kind of lagging a little bit. And my guest last week actually talked about, he does AI advisory for portfolio companies. And even still as soon ago as a couple of weeks ago, he was talking about how in a room of 100 CEOs, only eight of them said they were regularly using AI and that's on a weekly basis.

So there's certainly a gap in the mid market. So what sort of things do you think mid market companies should be doing to prepare for the inevitability of having to adopt AI as the tools get more and more democratized, affordable, well understood, more secured, like the wave is undoubtedly coming. Like what can companies do now to prepare? That's, that's a hard question because the usage of AI.

So you know, if you, if you think naively about the usage of AI and how consumers, like people, just normal people use it, they think of, let's say chatgpt and you know, they just write whatever question and they think of it as their pal. Of course it gets quite expensive quite quickly as you're, as you're passing tokens through the API. So companies need to be judicious about what type of data they're passing through, both from a privacy context as well as the cost context. Right.

So it's not as if I'm just going to plug into some AI chatbot and that's going to kind of provide me insights. What I think of is again data as an asset. So this data is actually critical asset, unique asset that, that you can leverage to train possibly AI, possibly just more mundane statistical models, machine learning models, but that will improve your things like your pricing, things like your negotiation with the supply chain, things like your internal operations. So those are critical, kind of mission critical core things that some AI models can do, but other times you just need simple regression models, sometimes you just need other approaches.

But having that data available to actually kind of make use of it and to figure out, okay, can I figure out dynamic pricing for all these SKUs that I have and do better demand prediction so I can smooth out my production so that becomes a critical asset to have and how to actually do the things I said in terms of the analysis I just mentioned and that skill set is going to be critical, but that doesn't mean just throwing all the data into some sort of AI API and saying, okay, what should I do?

That's one thing. And the other, having said that, there are other things where around legal or sometimes sales where an AI becomes a helpful coach and that's probably the more common use of AI that people think about. So I think there's kind of at least two different ways in which mid market companies can actually leverage AI. But they need to be very careful about the governance there.

Yeah. And governance is critical. And I think we had a brief phase where the technology was in its kind of earlier days where prompt engineering became a thing and there were even companies hiring prompt engineers. And now as AI has got better at kind of prompting itself and understanding actually context engineering is where it's at.

Is your data well ordered, well governed in a way that it can be understood? Because an AI is not going to understand a bunch of numbers and rows and columns in a table. You're the one who has to provide the context of like when you say revenue, what do you mean? Right?

Because through any sort of cycle of cash there could be revenue in different status, gross revenue, revenue with various fees come off net revenue. Except I mean you're the finance guy. But you know, revenue takes many forms on its, on its way through. And in fact at my, we had a great example of this during my, my time in big banking, right?

Active customers was, was always a number that people were interested in and we actually managed many, many different definitions of active, active customers because it depended, it depended who was asking. Like certain regulators might have different definitions that they use of what they consider to be an active company, an active customer. If we're, you know, pitching for someone who's looking to advertise with us, then, you know, surely we want that number to be a little higher or for some reasons we want that to be a little lower.

AI is not going to generically understand that you need to provide all of the, all of the, the context and the information on the way in. And I think that's the, I think that's the clicking point for me. I think when people learn how to understand NGO context and I've done some, I can show you offline shows because I know we talk regularly. Some amazing things in the past couple of weeks with Claude code and I even opened, I opened Pandora's box of openclaw as well and span up a virtual private server.

And I've got an openclaw friend that I talk to regularly as well. Yeah, no, I mean cloud code, anti gravity, all these things are amazing, I think but for me the focus is even though I love the possibilities of AI and everyone's talking about it, I characterize it as the mandate or the need for mid market or any company really is around strengthening their knowledge foundation so when data becomes knowledge and intelligence and can you create the context? So you know, I use this term ontology, right?

Which is, which is to say, okay, how do all these things hang together? So for example, I was, I mean I'm actually doing a project now where it's like this client has 10,000 product sort of like, sorry, customer names, but the names are all very similar and it's the same company but they mistyped it in Salesforce or whatever. So I have to kind of resolution. Right.

Very common, right. And oftentimes it says, well, okay, it's the loading, you know, it's the whatever, it's XX company, but the loading bay A or B, but it's actually the same company. So I need to map that to one company. So in this case and it could be like Company X Germany or Company X Toronto.

And I need to create a knowledge base or an ontology where I can define, okay, this company belongs, the headquarters is in Germany, but the subsidiary is in Canada and it's based in Toronto. And so I have kind of a map or knowledge kind of graph of all the things that are relate it. And this goes for everything from like, you know, customer information or product information, supplier information, financial understanding as well. So what do we mean by revenue?

How do we count revenue? You know, is it, I mean, if I recall correctly, didn't we work have like community adjusted EBITDA or something like that? Oh yes, that went well, didn't it? Yeah, that's right.

So, so, you know, so I think in terms of management accounting principles, you kind of need to sort of have your own understanding of how you understand your own business. And so all that together I call sort of, you know, the knowledge of the company. So once you have the knowledge organized, you can then use that to, you know, do AI. You can just create dashboards and understand what's going on.

You can, you know, make management decisions. You could create agents that are going to do the negotiation for you. So but you need that knowledge base to actually make sense of the data from the business. But to create the knowledge base you need the underlying data in the first place.

So step one is create that knowledge platform, then create the knowledge base and intelligence layer on top of which then you could create, whether it's AI agents or, you know, whatever it is, just dashboards, right? Yeah. The importance of context, investing in data as an asset. Two things I know that were both that we're both super, super aligned on.

Thanks so much for coming in today, Shota. I think that the stuff you're doing is fantastic, way ahead of its time. Using your early start and AI to create an offering that's really cool in this field of increasing that working capital. If folks are listening and want to find you want to find out more about what you offer, where would be a good place to go?

So yeah, just my email is shotarossimotech.com so that's one way you can just reach me. I'll put it in the notes. Yeah, yeah.

And yeah, I mean I'm around on LinkedIn, there's a website, you can put it in the notes. But I'm very happy to hear anything about what people are doing in this space, whether it's other practitioners like you, Graham, or whether it's companies. Because I think we need to think about this together. It's kind of a new frontier.

And again, a lot of these tooling I'm thinking is quite new. So I think we need to collectively establish the best practices here. So very happy to be online with you, Graham, and hopefully we can collaborate some more. Likewise, I look forward to continuing our partnership.

And you're right, there are a lot of pieces in the puzzle that make up the sunny future that's possible with all this amazing new tech that we've got. But if it's installed and thought about in the right way. So yeah, thank you so much for coming on. We'll keep talking.

I hope everyone else enjoyed this. And yeah, take care, Shouta. Awesome. Thanks, Graham.

Okay, bye. Cheers. Bye. Thanks for listening to the PE Data guy.

The place where private equity meets data. Please forward this episode to your favorite private equity friend. Thanks for listening. See you next time.