"$65 Billion Wasted!" - How Twin Knowledge's AI Prevents Costly Construction Rework

Hey everyone and welcome to the Bricks, Bucks and Bites podcast, your weekly dive into funding, market and investor news in architecture, engineering, construction and supply chain. Before we get started though, if you're a startup founder or operator, responsible for innovation in your company, or you want a deeper, more insightful look into technology in AEC, you won't want to miss our newsletter. From deep dives into particular technologies such as AI, robotics, design, software and so on. Hot takes from investors betting big in AEC technology and advice on building tech companies including the stories and takeaways of the most successful AEC technology founders. Head over to www.brickshyphonbytes.com and sign up today to get this exclusive material straight to your inbox and now for the episode. Alright guys, seems like we're back to our usual morning, Friday morning slot, which is a nice, nice change from things as we have moved things up and JP, welcome back, the second week in a row of us. Yeah, thank you, it's, you're taking that one straight out of the VC playbook right where your podcast experience is more and cool and then the fun performance, is that correct? That's exactly right. Yeah, you know, I'm going to start managing my own brand. Cool. Alright, so we are going to speak about a company that JP, you invested in called Twin Knowledge today. I think this is a very timely subject as we are inundated with the, should I say, hype or news or you just can't avoid the AI stuff that's coming out these days. So, um, who wants to kick off, I guess, Martin, you probably could ask a couple of questions. Oh, I hate Patrick, by the way. Yes, good morning. Thank you. Hi everyone. So I think Twin Knowledge is a platform that allows folks search for information based on their document, probably documents, drawings, files, etc. Does this represent the right way of explaining to you knowledge? Yeah, that's right. The thesis behind Twin Knowledge's product is fairly simple is that there's a lot of information that exists in a whole bunch of different data silos. And so Twin Knowledge builds AI assistance to help you query that unstructured data where their starting is a topic that we're pretty excited about. It's actually helping real estate and construction stakeholders QA and QC building drawings. And what we learned as part of the due diligence in this company was before a building gets built, there are thousands and thousands of pages of building drawings that constantly need to be QA and QC for you actually take that set of building drawings into the field. And I still think about one of the most unique insights we learned from a group in this space. They basically said we have two people responsible for QA, QC of building drawings. Those two people are handed 10,000 pages and they just don't have the bandwidth to go in and sift through each page to ensure that the information is accurate. So in fact, what they do is they take a subset of 10% of those pages. They check to make sure those pages are correct. And then if they are correct, they make a wild extrapolation that the other 90% must also be accurate because of the sample that they chose from looked good. And I think before of us know that's usually not what happens in reality. So what Twin Knowledge is able to do is they use AI and computer vision to make sure building drawings remain accurate and consistent for various iterations and especially as those building drawings should pass between various stakeholders. If I may ask at this point what's the accuracy? So obviously the building projects have different types of information input. So some of them can be very well detailed. There's 3D models, architectural, structural, and mechanical. So you can put all these pieces together and perhaps figure things out. But there's lots of projects that would have only 2D information or very limited 3D information. So how useful is Twin Knowledge at that point where the information is not perfect? Yeah, it's actually a 4D design done on the project. No, it only requires actually a 2D input. So the accuracy of the AI was something that we spent a lot of time on because we felt that we did not necessarily have the in-house technical understanding of how does this accuracy contribute to producing a very high-quality AI agent or AI assistant. We worked with a few different technical consulting firms to really unpack the accuracy and the quality of the AI and the accuracy is in the I90%. I think there's always room for this technology to continue to improve and get to hopefully to perfect 100% accuracy. But at this point in time, it's quite a robust algorithm and we're very confident that it's capable to deliver some pretty complex products and products. As I said, what you put in is probably what you get out from as an outcome. So would that spot, if an architect and engineer and M&E person is doing a mistake on their 2D drawings, would that spot, there is a mistake as people don't see it or is it something that's not yet possible from a technical aspect? It's actually a question to all of you, what do you guys think? The technology can do that today and maybe if I could just take a quick step back here and contextualize it in the context of the market. So Patrick and I am sure we hear about this concept of rework and construction on a daily basis. If you can believe it, there is about $65 billion worth of rework that happened in the United States last year and the expectation in the market is that construction projects run over budgets and over the timeline that they're expected to be completed in. That seems to be an expectation that construction companies and stakeholders and construction have sort of accepted as reality, but when we hear a number as large as 65 billion, not only is that a huge opportunity to go after, but it's also indicative of there's something institutional and structural going on in this industry where technology might make a difference. And so our thesis behind underwriting twin knowledge is very simple. It's Martin, exactly to your point, if you put bad data in, bad data is going to come out. What does that look like in practice? If you have inaccurate building drawings and you hand those drawings over to a general contractor and that general contractor circulates those with their specialty trades and that the drawings are incorrect, they're going to build against whatever you give them. It doesn't matter if there's a pipe running through a conference room or a toilet enclosed in a four wall space, they are given what you give them and they build against it. Our thesis is, let's cut that head of the snake off a farther upstream and start working with twin knowledge with the owner and developer group so that they have better visibility into the quality of their building drawings before it's actually bonded to the stakeholders and their counterparts in the field to build against. So where twin knowledge really excels is in that owner and developer and architect segment so that you're catching a lot of these mistakes early before the build is actually in the field and getting to be worked against. Um, JP, is it, is it looking, are they looking at the drawings from a build ability QA perspective like to say, oh no, that won't work or that won't meet regulation or you can't, you can't build a boat with a big hole in the middle of it because it will sink or is it more from the perspective of clashes between design or the potential that could be reworked like a pipe going directly through the center of a room or something like that. It's a bit of both actually so they use, they augment their data with external data sources at the same time. A good example of this is we have an investor that participated in this round alongside us a good friend of both Patrick's and mine and they saw a thesis for scaling this company within municipalities around the U.S. and globally where if you can augment the data that phonology ingests from its customer set, with things like building code and specifications in terms of how you build in New York versus Sacramento versus Berlin versus what have you, you're really creating an end-to-end solution that allows you to be informed of the quality and the accuracy of your bills regardless of what they're going on. So to your point, oh, and yes, the data they ingest from the customer is constantly augmented by external data sources, including things like building code. Interesting. Can they do a quantity take-offs as well? Hey listeners, I want to take a quick break to remind you about the Brickson Bytes newsletter. If you're interested in learning about the technology, shaping the future of construction, you won't want to miss the valuable insights we share each week. To get this exclusive content delivered straight to your inbox, head over to www.bricks-byts.com and sign up today. Link is in the show notes. Back to the show. Today, it's not a part of the feature set. I could see that being an additional piece of the technology at a later date, but right now they don't handle quantity take-offs. I think the quantity take-off is vertical within this market. I think it's something that we view as a bit of a commodity. I don't know if Patrick would agree with that. We're seeing a lot of carbon copies of each other within the quantity take-off sector, but in our mind, technology is really getting to a much larger issue, and that's helping to anticipate and prevent rework before it can rest field. Patrick, any thoughts? Yeah, I have a more, let's say, step back, conceptual question, JP. And to you guys as well, because you are also the practitioner, you will know how this works and the field much better than I do. So, you describe the process as sample testing today without a software like Twin Knowledge. And the problem with my wife used to work in a city group, in a large global bank, originally in sales. That's where she started and eventually graduated towards the internal anti-fraud department. That's what she did for many years. So basically trying to build systems to detect potentially fraudulent behavior and banks operate on sample testing all the time. That's why I mentioned this example. Now the thing with sample testing is, I'm not a statistician, but it works effectively when your outcomes follow some kind of distribution, doesn't have to be a bell curve, but some kind of distribution and the outcomes, the magnitude of the outcomes also follow some kind of predictable distribution. So sample testing on a power law, for example, doesn't make a whole lot of sense because you are extremely likely to miss the highest value outcomes. I'm not saying this is power law, but my point is the following when the general contractor industry or anyone trying to test for problems in their drawings, documents, et cetera, on a sample testing basis, the implicit assumption that you're making is that of a bank. You're saying, hey, I can rely on the underlying distribution and I can rely that the magnitude of the wave that I'm examining is also predictable. The problem with construction is, and that's my question to you guys, construction, when you get one thing wrong, it has a rat's tail of seconds, third, and fourth order consequences, and that makes it in my mind, again, I'm not the practitioner, much less likely to predict the magnitude of one error. Is that so two questions actually one? Am I thinking about this the right way? Is that how the practice actually works? Also is it more predictable distribution and two, if I'm right about this, what's then the thing that twin knowledge truly creates value for? How do I have to think of twin knowledge in that context? Well, yes, I would say I can't answer the second question, but definitely the first one, every error typically will have a knock on effect. If something is not detected, say early on, then in my impact, I don't know, say the electricians, but then I also have an impact on the plumbing and so on and so forth. That usually, like J.P. said, error results in costy rework. But to your second point, Patrick, I wouldn't know, that would be one for J.P. What was the second question, sorry? Then how do we evaluate the value of a solution like twin knowledge when you know the first question, basically, hey, is this going to be a predictable, is simple testing effective in construction, essentially? If it's not effective in construction, what's the value of twin knowledge that I can speak somewhat anecdotally to that and I'll just point to a few insights we gleaned from our due diligence. So we varied candidly during diligence, we tried to figure out where the highest value of this technology actually sat. And so we made a subset of introductions for twin knowledge to a couple of developers, a couple of AEC design firms, a couple of general contractors, and the intention there was to figure out where was the greatest poem. The poem that was most, I think, most aggressively felt actually pulling from the developer segment, they basically said they sit between the owner and the architect and are kind of constantly perceived as the bad guy. They have to go back to the owner, ask for more budget because they're over budget and over timeline. They have to go down and talk to the architect about why these building drawings are incorrect and why they're catching errors in this material. And so there's really no piece of the value chain where they win. They're constantly being berated from both sides of that value chain. So the COO, one particular developer told us, if you can help me stop being the bad guy for it with my partners, that would obviously save me a lot of headache. But more importantly, if you were able to save me even 100 Bips or 200 Bips in cost by the time we actually start building against this project, that's already meaningful. For me, I'm happy to share a percentage of that cost savings with Twin Knowledge directly. And so when you start thinking about, again, the market size of rework that occurred in the US and globally, that number gets very, very big, very, very quickly. But again, that's just a singular data point. In our minds, the value of Twin Knowledge, yesterday is QA and QC of building drawings. But their AI architectures incredibly flexible. They're able to tap into a number of different data sources. I think the thing that was very exciting about this technology was that the possibility seems fairly large, the possibility seemed very large. But we wanted to help the company on day one figure out where's the greatest pain, salt, and we feel like that pain is most apt and with the developer, kind of the architect. And JP, what it allows you to do is basically eliminate several testing, right? So basically saying, hey, I'm testing 100% and that's why I actually eliminate the problem that, you know, when you were confirming exists. And I don't know. We just had the toll brothers case study up on the screen here, if you, it's fairly light or fairly short, but one of the insights from the toll brothers case study was that they're able to do four or five X to work with half the team. So it's, it is a, there's some pretty significant efficiencies that are occurring at least with these early data points. Again, the company is young, the company is, you know, just raised their first round of institutional capital, but the logos are notable. The ROI seems apparent and, and we've got the network to help them continue to deliver. JP, you actually preempted a question I was going to ask, which was why, why start with drawings, but perhaps that was like the, do we still refer to this as a wedge or beechhead? It's kind of solution for now, because I'm assuming that the technology was not the commercial aspect of it is focused on the drawings, I guess the technology is capable of tackling other areas of being construction, but basically due diligence that you've done, finding the pain points for the developer seemed like the perfect initial market to go after. That's right. Yeah, but I would say wedge is probably the right term. I, again, it kind of comes back to where, where is the greatest amount of pain felt in Martin, you already kind of alluded to this with a 2D drawing. It's incredibly hard to query the information that comes out of that drawing. You can query information from a CSV or an Excel file or what have you like a little bit more easily, but with drawings, it's not as easy as saying, hey, let me use a control aspect and search for an MEP or HVAC or what have you. They are solving in our mind a very hard problem first. The technology is delivering on it quite effectively at this point, but in our mind, that is the right wedge. Let's go after where the greatest pain exists and deliver on it exceptionally well. Is this, so in the end of the day, this should be a search engine, kind of, which you put in all your data and you have particular query and it gives you the answers or whatever you want. I don't think so, man. No, I don't think that's what JP is describing. It was less of a certain right. It's more preemptive. So your drawings will be in there and it will say, oh, here are all your errors or issues or clashes or threats, risks, whatever you want to call it. That's right. The way they would refer to their product is as an assistant. You're almost creating an additional, like a digital FTE and other full time employee alongside you to help you query this information. Oh, so that's, however, yeah, my and in your defense, there is a picture on the website where it may suggest some migration would be even more direct probably. Is it not a very temporary situation in the light of AI and how things moving quickly? I guess one could argue that. I think things are moving so quickly with AI that it's possible that this solution is obsolete in for months. I think one topic we haven't really hit on just yet is who's at the helm of this company and Patrick, I'd actually like your input on this as well. Nice. Sorry. I was preparing a different segment now. Actually, I want to first bounce it back and then you'll hear my opinion. Okay. JP. How long are you living in Europe now? Three years. Three years. Okay. It's remarkable how well you combine the two continents, you know, the American vision, business savvy and the European engineers mindset and being fact based, you know, you're very toned down about this company, but come on, man, you love this company. You invested in it, et cetera. So the next segment is going to be called cheerlead for your company. Okay. Ted, remove all constraints, man, tell us what you loved about Ivan. Yeah. I have to say that's a some feedback I get regularly. I'm not very good at cheerleading for myself. No, it's great. I love it. It's like, I think this is the best way for the listeners to actually get into the substance. It's amazing. But the next segment you're allowed to cheerlead. Okay. Well, so the gentleman at the helm, he is a, he is an absolute product wizard. He's a gentleman named Ivan Padyshev, who actually founded a company back in, I believe it was 2007, called horizontal systems. If you follow the history of AEC for the last 30 years, horizontal systems was actually acquired by Autodesk in 2011 and became the backbone of one of their best selling products in 360. So horizontal systems is actually now been 360 blue. After that acquisition, what we were very excited about was Ivan didn't feel like that scratched the itch sufficiently enough. And so he then spent the next 20 years in AEC building and operating and thinking about what kind of the next two or three companies would be that he would want to lead. He ended up leading a company called Infra and AI and ML startup that was actually just recently bought a couple of years ago, he ended up becoming number two higher at Amazon Web Services to lead their AEC group. And all through the way, he's seem to just come with an absolutely sterling and stellar reputation. Everyone we've spoken to him about or about him has said, Ivan is just a top notch executive. He knows this industry extremely deeply, he's highly networks and he seems to lead a really positive impression on the folks that he interacts with. And I remember as part of diligence, actually, we were getting down to the final days of making a decision on this investment and I texted him and I said, my biggest concern here, Martin, to your point is and someone that is not in industry replicate something like twin knowledge. If all of a sudden, Palin's here wanted to come in and say, hey, we're going to build twin knowledge for the AEC vertical, is that an existential threat for this company? And Patrick very quickly quelled my concerns there and said, look, if Ivan has the industry expertise and the network that you say he does, that understanding of nuance in AEC is going to go a very long way in helping them succeed in our vertical, because there are certain again, nuances and specificities to AEC where having that background is going to provide a competitive advantage in the boardroom and selling this product. And I think that, I have to say Patrick for that insight because that was one of the kind of final pieces of data we needed to collect to get to the point to say, yes, let's put in a term sheet support this company. Ivan, I can't speak highly enough about Ivan, we believe he's the right person to lead this company and we're excited to kind of build the network around, and it sees me around him. I want to build on that because, okay, recently as in the last 24 months, maybe, we here at Founder Mental have been getting more and more excited when products get repped in white glove distribution motions. You mentioned Palin tear. It's certainly a darling of that motion or pioneer of that motion when it violated all kinds of narratives. The narratives was, as we say, as we're recurring, make it as low touch as possible, if it's not low touch, it won't scale. Big narrative violation from Palin tear when they actually, and also at the time, oh, you can't do business with governments, it's so hard, such a difficult motion, et cetera. Well, but why was it difficult, was difficult because governments need, they have very, very low risk tolerance, they have very high risk exposure. And so you needed to translate a being a technology company, being a new player on the block into track record, into effective relationship building. What is an effective way of doing that? White glove in your service and not immediately asking for a three year recurring commitment, but actually demonstrating the outcome before you make people commit to a long term contract with you. But that obviously snowballed into a very, it's not secret, but into a defensible mode that Palin tear has built. And I'm seeing JP to your point, similar motions in AC and especially the C part of AC and the developer part of AC, being effective where customers in our project-based industries have very similar issues, it's idiosyncratic to many other sectors that way. And that's when wrapping a good product into a white glove motion, led by people with what we call in German barn smell, that's a positive thing. So not associated with something negative, it means you're credible. You speak the language. You can speak from experiences that the customer will share and relate to. That's the barn smell. And so Ivan clearly has that. So when you combine those things, and you give customers a slightly tailored and white glove experience to implement your software, I'm very enamored with these kinds of distribution motions. Is that what twin knowledge is doing, white gloveing it a little bit or low touching it? It's definitely more white glove than low touching this point. I think they're, you hit the nail on the head, I think, again, going back to my earlier point, finding the pain where it exists. I think Ivan is a bit, I describe him as a Swiss army knife. We actually ask city questions for those who don't know and me. What is white glowing and low touching in this point? I was going to ask the same. Sorry, that's the, that's the silly jargon. Okay, so white glove, white glove basically means that I'm not trying to give you a product and expect you to be able to use it out of the package. I'm not giving you a plug and play type experience where I'm trying to minimize my effort to help you implement it, help you learn how to use it, help you perhaps integrate it with other data sources that might be required to be effective. White glove approach means other people would call it consulting, but I think that's wrong, because consulting is not about outcomes. Consulting is about giving people an insight into something. A white glove approach in the context of software means, hey, I'm not leaving you alone with it. I help you get started. I help you pull the data that you need within all the other systems. The rate white glove company would be Infosys or TCA, a TCS or Accenture, SAP has lots of white glove consultants around itself, sales force. Did you guys know sales force is 280 billion market cap and has another 150 billion market size just for sales force consultants who help with implementation? That's white glove, Martin, right? As far as the old touch grows, here you go, don't bother me. Actually, it's a very interesting point, because my feeling only is that in construction, this white glowing exists because things are not that straightforward. You can't just get out, concentrate away. I think this is actually feedback from innovation managers we speak with, technologies that are pitched to them very often don't do what they are promised to do, just because it's so complicated, and not everyone desires the same outcome from technology, every large contractor is different to the green, so I think thanks for that insight, it's an interesting concept, actually, there's white glowing. There's an additional object, go for it, go for it, go for it, go for it. So there's an additional thought from the perspective of an investor and also the founder. When you play white glove the right way, you can actually make your CAC negative. So CAC as the customer acquisition costs, it's a very important, it's not a metric, it's a concept to monitor as a founder and then by extension as an investor because it does define how quickly and capital efficiently you can scale a company, it tells you other things too. A negative CAC means that basically the customer finances your distribution. You're making the customer give you cash in order to sell to him. That's what white glove is more effective to do than a traditional SaaS motion. You can see it in SaaS motions too, especially for example, you are able to close three year contracts, relatively high ACVs, front loaded on cash that also can work, but white glove is another very proven way to do that. It's not talked about enough and founders that are infected by more the last 20 years of how to build SaaS rather than the first principles that construction offers to them, they might forget this, but it's something that I absolutely love when I see white glove approaches. It's not the end or be all, it's not the solution to everything, so I'm not saying, hey, without thinking critically do white glove, not my point, my point is think about the first principles and a twin knowledge case, JP back to that company of yours, I think the white glove motion makes a ton of sense. What I was going to say earlier is Ivan is, I describe him as a Swiss army knife and the reason I do that is he's so effective at diagnosing where twin knowledge is tech, makes sense for his customers, so much so that he's done something that I think is several steps ahead of where we see a lot of founders before this company really hits it stride, he's started to quantify what is the value of working with us before it's really in mass in the market before the brand twin knowledge is widely known. I remember we introduced him to a few groups and we sat in the room with them as these introductions were occurring in 15 minutes, he asked a few introduction questions, he kind of figures out, you know, where, you know, I don't know if it's a question of body language or how people are communicating certain things about a problem, but he's able to pinpoint exactly where a twin knowledge should exist in a customer's architecture, and he runs very quickly towards it. I'm not sure. And he's a small spoiler. Yeah, right. So we're very excited to work with him as a CEO and a founder, he's also has this incredible balance of pitching vision and kind of the quantifiable ROI of what twin knowledge can do for his customers. And I think that that combination is going to prove to be quite deadly in the low room. Yeah. So I think JP maybe while I'm reading into this is that your obviously cognizant of the fact that AI, there's a lot of hype around AI at the minute and perhaps anyone, almost anyone now probably code some form of AI tool. However, in this case, it was very much down to the founder who has that a good product with them, that you're more, more bullish on as opposed to the product itself in his current form. I think there are merits to both, but we are particularly excited about Yvonne. Yes. And again, Martin probably remembers this from the last time we spoke. Our strategy as a fund is we have 300 real estate and construction companies invested in our funds, those groups can feasibly become customers of the technology companies that we support with capital. And in our mind, that is a very productive combination because Yvonne can continue to exercise as brilliant product minds. He can continue to be that Swiss Army knife in the sales room and we can fill this top of funnel. So it's a really nice combination when we at Canberra could kind of serve up on a silver platter. The customer for you to consider, Yvonne can go in, fly some dice, diagnose, figure out where twin knowledge makes sense and then execute in that regard. So yes, again, I'm not a very good cheerleader for myself sometimes, but I'm pretty excited about this. You may be excited too now, so I think you've done a good job. I wanted to detach on this, so $3.7 million, so share whatever you can. So they raised $3.7 million. And how did you come up with the valuation and yeah, maybe let's start from that. I won't get too deep in the weeds on this one. The short answer is the amount that Yvonne raised is sufficient for the next 24 months. And there's a couple of reasons as to why we believe that's the case, but again, it goes back to we're going to be active partners to him and the company in filling his funnel with prospective customers. There's a lot of bandwidth in our mind that gets taken, you know, that is given back to Yvonne as a founder when we can kind of tee up the next introduction to a prospective customer rather than have him kind of pounding the pavement and going to market and finding the next customer himself. Okay. Not to say that he can't do that. Again, he's incredibly well-networked and actually as part of due diligence, during due diligence, he landed a couple big deals while he was fundraising at the same time. In our mind, what a founder can raise money and also do deals at the same time as due diligence is occurring. That's a really strong, positive signal. And he did that in two separate occasions while we were getting smart on his company. So that's a really cool signal. I hadn't thought about that. I like that. I might steal that from you. That's a pretty cool signal for the early years. So where is the product right now and to where this money is going to go over the next 24 months? Well, we'll need the skilled team significantly. It's a bit of a strike force team right now, which is great. And it's worked for the first, call it, 18 months of the company's existence. But now it's about figuring out where, from a head count perspective, we need to really staff up. There will certainly be some additions to the engineering side of the house, the BD side of the house, the sales side of the house. What we're taking a look at that, actually as we speak, and our upcoming board meeting, I think we'll have a, we should have a much further plan in terms of where we want to immediately hire. JP, not another trick question, because I can't remember what my colleague, when they looked at twin knowledge, if they were looking into that or whether, you know, the deal was already true advance and, you know, just for clarity, also for everyone, JP has an ownership target and his and my ownership target don't always work with each other. So, but that's fine, sometimes we also follow each other. So maybe that's going to be the case in twin knowledge, but that's also why we didn't progress to like the super advanced stage with twin knowledge, which is why I'm going to ask you the following question, because I don't know the answer, when, when you did your competitive land, land mapping, did you feel like twin knowledge had lots of peers or did you feel like this was something where they are extremely alone? It's a great question. So, candidly, when we built our competitive landscape, I think it was something like a hundred and eight cell lines long, but the competitive landscape wasn't just focused on the AEC and real estate vertical, it was more a broad view of who among the, you know, the VC back generalist portfolio companies might, you know, might be able to take a right hand turn and move into the AEC industries. So the map, the map itself was pretty extensive, but as you kind of drill down on AEC and real estate, we felt that they were fairly unique, not only in terms of the technology that they built in terms of being able to QA and QC building drawings, being again, where a lot of the pain exists, but also the customer that they were selling to, we felt that they had really figured out a pretty intelligent insight around the customer we want to go after, the one with the greatest propensity to pay for something like this is going to be farther upstream than it is going to be down with the contractor ecosystem. In our mind, that was a very, a very good insight to execute against. Look, I think again, we are cognizant of the fact that AI is going to continue to proliferate very, very rapidly, but if I go back to my earlier point, we think we have the right person at the helm to really go and execute against this mandate effectively. You said that earlier that from the customer's point of view, they would love like one, two percent of savings kind of to know upfront what they could save. So my question is, I don't know if it's a good answer or not, but how can we figure out that this is going to be one or two percent upfront and how much are they saving really? This is a good question to ask or not really? Yeah, no, I mean, I think it's the right question to ask. We did a few back-to-the-Napkin calculations, both from the seat of the developer and also the seat of the architect and the engineer before we invested in this company. And maybe I'll just walk you through what that looked like. In the case of the developer we spoke to, the COO that mentioned he's getting berated from both sides of the value chain, they have a development in national tenancy, pretty big mixed use development. And I think it's a combination of multifamily apartments, retail and some office. And what we did is, all right, we took a look at how much it cost to develop per square foot or per square meter in Nashville, a building of that built. And then we said, all right, if it's anticipated to be this big, you're building it this much per square foot and you're doing about a 10 to 15 percent cost overrun, you're missing your budget by 10 to 15 percent. What is that value translating to and what is the feasible revenue potential that that can move for knowledge if this COO says or delivers on the fact that he would be willing to share those cost savings with twin knowledge. And again, that number becomes very large very quickly. Can they save one to 2 percent on the bottom line? Maybe. We'll see. I think the company is too young to really say definitively that they can save 100 to 100 depths per for their users, but the use cases that they've published here on their website, I think, are pretty strong. In fact, there is value to be had, determining the scale of that value is going to be something will work on the team regularly. Okay. I want to piggyback off of that Mark Martin and all of you. So JP, you've cited the 65 billion rework problem. I think you were referring to the US market, correct? Correct. I mean, it sounds massive, right? And so the sources of rework can be varied, right? So it can be whatever change orders that all of a sudden the customer thinks they need something else. Okay. I mean, that's then in the control of the customer, if they really want to change, can be badly planned. It can be badly executed. Maybe it can also be that your subcontractor has priced in a way that they actually speculate to get to or to create certain change orders because they actually make the margin with the change orders and so on and so forth. But I want to focus on this practice of making money with change orders. And that's sort of part of the 65 billion of rework I'm assuming now. Not about twin knowledge necessarily, but solutions like twin knowledge. What are you guys? What's your intuition? This practice of making money with change orders. What will solutions like twin knowledge do to that practice? Could it create sort of a completely new business dynamic? It's a very abstract question, but perhaps you guys will be able to refer to. I like the question a lot, actually. So I can marketplace that you could based on the differences, no? So I actually don't know, so what Patrick is asking is because you are eliminating issues within drawings, does that reduce the opportunity for subcontractors to come back to you and say, I need more money because this wasn't in a design document lately. And then might they offer differently in the first place? Might they price their own quads differently in the first place? Yeah. It's kind of an interesting second order dynamic to see. But let's, we can, everyone here, I think we are so ingrained in the construction of how it works. But let's go back. Like, if we buy something outside of construction, we are usually buying the product act, which works, and we know how much we're paying. We usually know what the value of it is, right? We kind of think that there is this change order, or we get another 5, 10% of this contract. Why this solution should not be implemented as a default in, and this is, I'm speaking about my ideal world as a designer. Why the solution should not be implemented in a software package that someone is designing on, and figure it out once they are designing it, not once it goes to pricing. Because that's like, why would you even, why do you even sell it elastic and they probably sometimes, like, that's how it should be from the design. I will. I'm sorry, we'll one day, but we're not with tackling was the issues in front of us. Yeah. But why would you tackle problem, which, which is maybe not the problem, because the problem is the designers giving you crap, and you have to work with the crap. That's why you need software like that. That's a problem. Yeah. But maybe the incentive for the subcontractor is to win the job and make the margin after they won the job. Yeah. I mean, they're sort of a different dynamic here at PlayMart, isn't it? But that's not how it should be, right? Yeah, but it is mine, but it doesn't mean that it's correct, and it doesn't mean it's going to last forever. Mine is. This is inefficiency, right? And technology is there to tackle. And that's my question, Martin, what is the trend of twin knowledge? Imagine twin knowledge was rolled out across every project in the United States, 65 billion of rework, but we know we have this dynamic. What's going to happen to that dynamic? I think there are other issues as well, Patrick, perhaps as to the why subcontractors go after their main general GC's contractors, where everyone would call it, which are not necessarily it, which are not just related to drawings. Yes, this will eliminate part of that, but there are other reasons like, I don't know, someone didn't move their ladder in time. So, therefore, I need to put in a four month extension to the contract. Whether? Yeah. Yeah. Someone put a huge tariff on us, and now we can't do this effectively. So, yes, it will eliminate, but no, sorry, it will reduce, but not eliminate. What I'm trying, what I was trying to say with my five pens earlier is that you, while right now, we are deferring the moment of the truth, if the design is handed over badly or it's not fully compliant or put together, we are deferring the moment of the truth. The moment of the truth should be before the design is handed over, and then everyone is on the clear page, everyone knows exactly that they are buying iPhone 16, and that's the product. That's how much we pay for it. Yes. We also live in a world where there's no wars and peace, and everyone is in utopia, and we're having such a great time. I mean, this is, no, that's how only how construction works, I think, that you kind of keep having to keep having. I think you'll be surprised, actually. You are a QS. You are a QS. That's why you are, you are in the business of adding money to the projects. No, I, well, yeah, whatever. So, you're talking your own book going, no, no, no, yes, you are. I'm sure there's. It's not a show about us, it's not a show about us. I mean, I, I let the fire here, that's great, but the metaphor, okay, so I actually have a, so I was first thinking of the analogy of drugs coming into your country. Let's not use that. Let's, let's use a metaphor. It's all pointing at the American right now, it's left, no, we all have to. That's why I do share that problem, but, but let's use the metaphor, okay, so you have a bucket, and the bucket has holes, right? And so you're kind of thinking, it sucks that, you know, I calculated with this volume of the bucket to hold, well, sorry, hold this volume of water, but it really doesn't because there's, you know, this hole, and then you're plugging that hole and you realize, oh, actually, now the water is seeking, there's another like small tear in the bucket, and now the water is slowly seeking through that other tear and it's creating a completely new hole. So until I give you a perfectly intact bucket, the water is, is going to find a way. So I wonder whether that metaphor is going to apply here is that, hey, until I fix the incentive for this contractor, not to win the bed with the lowest bit, they're going to find other reasons in order to create change orders. That's kind, sorry, I'm not making that prediction, I'm not saying I know, I'm saying that's the question in my mind. Yeah, I think I would make you right, I would think that people would start putting much more ladders on site so it gets in the way of their work so they can use that as an excuse. So, yeah, we can't use the drawings anymore, but the only way they can save things, obviously, I'm talking, we talking about, between knowledge is to check the technical details drawings. Right? Currently. So we, we're not talking about ladders, we're not talking about the weather, yeah. So at least let's get the design aspect done correctly, not a bit. And there's not a bit. Not a bit, right? I'm not saying this shouldn't be implemented, I'm not saying that at all, I'm only thinking about the second order consequence of this, of this weird dynamic and practice that we have where a source of rework also is that I underbid in order to make my margin with the change orders. Exactly. That's an ethical, that's not good business practice and people should go positive, they do this, honestly. That's brutal truth. I don't know what else is to say about that. If this issue fixes in your lifetime mind, then I will be very surprised. I don't know, I'm not there to fix it, but I think we'll put things as they are. You know that Martin is actually an active subscriber to three different longevity sites and their vitamins. He plans to live very long on. Yeah. I got your advice to stay eight hours a day, so I'm trying to, I'm trying to, I'll remember this episode. And when you are at the age of 149, I'll send it back to you and I'll see if you've been there. Okay. Move it on. Move it on. I think we're nearly at the end, actually, of our time, so. Any closing thoughts on the space of twin knowledge, guys? Yeah. Any thoughts about A.I. If you want back to founder, you love, I think JPU founder. Yeah. I feel confident in that too, and it's, I tried my hardest to work on this one with Patrick and the fundamental team, so maybe at the, at the next round, we'll, we'll finally get that chance. I think we have a company like that in their portfolio, or, or, no, no, not directly. Not not directly. So there are companies and I will not mention them because this is not, this is about twin knowledge. I've companies in my portfolio that could become partners, that could become solutions that, that's not in bad, but rapid together between both companies where they can distribute each other, etc, but not exactly what Ivan and, and twin knowledge are doing, not currently my portfolio. Okay. All right, guys, JP, if you want to join us for a Fred Week in a row, then we, I love to have you. Let's do it. Happy to.