Proof of Work: AI Value Creation
00:07:03 · Joe Zein, Stuart Willson
And in all different stages of this pipeline, there's automation that can be that can be happening. Strings is doing a lot in terms of like the ingesting and the categorization, and then there's also some not even AI, but there's also some orchestration layer where something as simple as a task system, you know, could be drastically beneficial uh and and improve the process by a lot. [00:07:25] S0: In a in a best case, how long does that take today? And in a worst case, how long does that take? [00:07:33] S2: I think it's a function of how many portfolio companies you have, right? And how big your team is. As Yann said, the natural reaction and the natural solution has been throw more people in that portfolio monitoring team, so throw more orchestrators who will who will manage that process. But usually they try to, you know, the the those teams try to set up SLAs because they have to be reporting by a certain date, right? And I've seen processes, quarterly processes take longer, take you know, 45 days for it to be completed on a portfolio of 70 companies. So you can only imagine uh how that scales with bigger orgs. Yeah.
00:07:33 · Joe Zein, Stuart Willson
[00:07:33] S2: I think it's a function of how many portfolio companies you have, right? And how big your team is. As Yann said, the natural reaction and the natural solution has been throw more people in that portfolio monitoring team, so throw more orchestrators who will who will manage that process. But usually they try to, you know, the the those teams try to set up SLAs because they have to be reporting by a certain date, right? And I've seen processes, quarterly processes take longer, take you know, 45 days for it to be completed on a portfolio of 70 companies. So you can only imagine uh how that scales with bigger orgs. Yeah. [00:08:09] S0: And is the uh is the problem the cadence? Uh and I'm gonna guess the answer to this, or is it the right cadence with the wrong infrastructure? I mean, we when you I guess when you think about you know what this would look like in an optimal world, I guess it would be real-time reporting. But tell me a little bit about like what your thoughts are on both the cadence and the infrastructure today, and you know, where do you see it going over the next a few years given the technological developments that you both are deploying into these companies?
00:08:40 · Yann Magnan
[00:08:40] S1: I I I can give it a start. So I mean to me, uh the answer is fairly simple. The uh the problem is the uh infrastructure, it's not the cadence. Um and uh let me explain you why I'm I'm I'm putting that statement just here. Uh although I mean we all know that uh most of the portfolio companies would be reporting only on a quarterly basis. Uh here is the here is the thing is that the industry is growing really, really fast. And uh the um uh quite likely one of the uh next uh phases of growth uh will come from um addressing the uh needs and and the uh the demand from uh retail investors, 401k plans to basically invest into uh into private capital. If you want to get there, if the industry is going to get there anyway, that means that you have to report to your to these LPs faster and better. Um so with your clients, when you're a GP or the uh D LPs, you know, whoever they are, if your LPs are asking a monthly reporting, or actually we're we are working with uh with clients that are doing weekly and sometimes daily reporting and valuation, I mean that's what you have to do. It's as simple as that. So the question is what infrastructure, what technology do you put in place so that you can really deliver on that on that on that cadence? Um so that's one piece of the question. Of course, as I was saying as well, most often the popular companies will only report their uh uh only send their financial pack and border once uh in a quarter. But then there's also a lot of uh qualitative information that you get from their from your investment team. And that's another layer of information that you want to capture, and that can be done and accelerated with technology. Yeah.
00:10:21 · Joe Zein, Stuart Willson
[00:10:21] S2: Yeah, completely agree. I don't think it's a problem of cadence. The cadence at the end of the day is not something that you have a lot of say in, right? If your LPs want it monthly, you're gonna do it monthly. There's no choice. And the the big problem is the infrastructure, but then you also have the bottleneck, which is how much you know, it depends on on what kind of assets you're looking at, but how much is an asset like gonna change at the end of the day? If you're doing a daily report, a daily report that it's not changing, but it's only changing at the end of the month or at the end of the quarter, then gotta question the cadence, right? Is it a useful cadence, yes or no? [00:10:56] S0: Let's talk about signals. Investors, lenders want signals that are going to show up before EBITDA, as an example, moves. Yann, what are the signals that you see that show up that are distinguishable from noise? They're early warning signals of EBITDA changing in the future.
00:11:20 · Yann Magnan
[00:11:20] S1: Yeah, um, so I'll I'll maybe answer the uh question slightly uh differently here because there is the thing is that um private capital means a lot of things. You have like uh large cap, private equity, mid-cap, small cap, private equity, venture capital, infrastructure, all of that is private capital. And the signals for each and every of these uh strategies might be very different. If you drill down even within private equity, you know, whether you're targeting pharmaceuticals versus software, signals will be different. To me, what really matters at this point in time, where the uh industry is and where the technology is, is that the more data points you have, the more data points you have, the more data points you have collected, the more data points you have structured, standardized, consistently labeled, the uh better you are armed to be able to extract signals out of that. That's the key point at the end of the day, is being able to extract, to uh, to collect as many data points as you can. Joe, Joe was saying at 73 Things, we we have developed an ingestion technology. And the purpose of us uh when we uh developed that uh ingestion technology, which is AI-based, back back a few go a few years ago, the purpose or the reason why we did that was actually to feed in into uh into our valuation models, into our into our valuation platform. And our valuation platforms are fairly sophisticated. There's a lot of data points that go into that. But typically, uh, in the early days, we were collecting with this ingestion technology 20, 30 data points, a bit more when there was a business plan again to fit into our valuation platform. Um short moment after starting rolling out our technology, ingestion technology to clients, we realized that actually uh a lot of our clients were collecting each and every data points that they were receiving that that were sitting in the border that they were receiving from their portfolio companies, in the quarterly reports, in the compliance certificate.
00:13:18 · Yann Magnan
And the reason why they started to do that is actually because once you have again all these data points in one place, especially with the uh uh technology that's available out there uh today and the uh LLMs that that are out there, once everything is structured, then that's when you can overlay this uh data with some very uh interesting technology that will help you surface uh this type of signals. But there's no one size fit all. It's a lot, you know, it will vary a lot depending on the industry and the strategy and also how you can infer this type of signals from between one company that you're monitoring and the universe of company, whether it's in your portfolio or you know in in other portfolios.