Latest / Construction Technology Podcast / Construction Technology Podcast - Episode 006
Transcript
- Luigi La Corte: I think like the, like the pre-recording conversations are always very, very fun. We need construction technology podcast after hours or, or the pre pre construction technology podcast conversation. What's up? What's up everyone. Welcome to construction technology podcast. I'm joined by Andrew from join Mike from build vision and myself from provision. What's up gentlemen? How was the week? Andrew, what was the highlight of the week? Andrew Zukoski: Yeah. What was the highlight of the week? was the highlight of the week? We closed a two-year renewal with one of our largest customers. I'm not gonna lie. it like, yeah, it's good. like, I mean, we've working with them for years. Luigi La Corte: Boom, let's go. Two years. Why not five years? Why not five years? Andrew Zukoski: Actually, think there's an interesting thing there. We've never done a five-year deal. We've done three-year deals before. Some of it may be the commercial terms that we hold to for longer-term contracts. you have to pay for price certainty, and that disincentivizes the long-term thing. But I think it actually points to it's an interesting time in pre-continent design phase technology. I think a lot of people are like... Luigi La Corte: Hmm. Andrew Zukoski: picking up the rock of like their estimating system and how they do design management and what their role is and saying like we need to reconsider the systems we're using, the work breakdown structures we're using. And ⁓ so are looking to make two-year bets, not five-year bets right now. Luigi La Corte: I had a post, it was intended to be provocative. I think I'd said something like, anybody who's putting their customers in more than a one year length deals, it doesn't care about the customer, given how fast things are moving. What's your view on that? Well, this is the intent of the deal. the record, we have three year deals with our customers, right? And so obviously I care about our customers, but it was supposed to be provocative. Like things are moving so fast. Andrew Zukoski: ⁓ stabbed me in the heart, like. Yeah. Mike Powers: Hahaha Luigi La Corte: You know, where, let's say, Gen.AI and model company productivity software is going to be in one year, somewhat the benchmark now. And anyway, love to hear your thoughts. Andrew Zukoski: But yeah, so is that the, so one, like we love this customer very much. They've been a key partner to the company throughout our history. Like it's just like cool to continue the relationship. So we do care about them, I swear. Is it about, yeah, is it about the technology? Luigi La Corte: I have no doubt, the way, obviously. Andrew Zukoski: when you do these multi-year deals or is it about like the adoption and evaluation life cycle? Like the way I've heard it explained is like, yeah, we're gonna do, we do three-year deals when we go into price wide, because it takes a year to implement the thing, whatever the thing is, and then you use it and then you decide two years in whether or not you want to keep going and then you need a year to ramp down ⁓ don't like it. Like is that cycle, like that cycle is about like people and uptake of technology. Is that different in the Gen I world? Mike, what do you think? Mike Powers: So we've had more customers ask for exclusivity other contract terms because we're giving them a competitive advantage, which is a very interesting thing that we're working through of for customers on the buy side of our application. We go with a whoever raises their hand first on competitive projects because we know that we can buy better than anybody else for mechanical electrical equipment. And on the side, We have manufacturers that are like, Hey, we know if we roll your tool out, we're to be able to sell more. So how do we arbitrage this AI workflow so that we sell more equipment and get a head start? So having said that, then it's like locking in long-term, very expensive contracts because they're considering this a bet, like their AI bet, and don't want their competitor to have that same exact AI bet. which is the first time I've heard that because like in building connected world, were like the network in fact only helped everybody. didn't really, you didn't get really have a competitive advantage because you're using building connected. We have customers winning projects because they're using build vision. Luigi La Corte: And that makes sense if you're selling outcomes. Andrew Zukoski: winning on price, like they're able to come in at more competitive price because of the procurement advantages you offer. Mike Powers: It depends. It's like, oh, we need to release the switchgear March 20th, even though the project doesn't start until September 20th. I, the general contractor, need to be able to buy that equipment. Then we're able to say, can help you buy that equipment. It creates a profit center for the general contractor. They're able to make money off of it. So they're going to market saying, hey, we can do this at scale. We have this partnership with BuildVision to be able to do this at scale. Andrew Zukoski: and you're able to. sauce. Mike Powers: Which is fine, but it's interesting, but the challenge, like that's like this interesting perspective that we're working through is like, we are on both sides of the trade. So we're like, we have to be very like separated in those positions of like having walls between teams and account managers because we're buying at the best price that we can get at the the time at the best compliance for the building. We're also selling and monetizing the selling motion of how we sell more for our customers. Luigi La Corte: It makes sense that they'd want exclusivity. We had this conversation with some large distributors in the market. The specific advantage was we would help them with their internal errors and emissions. I always him and haught about whether it was worth it because ⁓ ⁓ depends on who you're selling to. If you're selling to a distributor, the is that you distribute and you want as much service area as possible. How do you that decision when you're talking to customers, Mike? Mike Powers: So we go to the tier one manufacturer that we wanna work with. contracts are large, the contracts are seven figures, and we're gonna pick the winner that we think in each one of those markets. And that's who we wanna partner with in those exclusivity contracts. Because we we can sell and help ⁓ the rep better using BuildVision. Because in this world like a... ⁓ But even a, like a chiller, there's not a right answer for a chiller. You can't just like, Hey, I want this exact chiller. It's an engineered system. The numbers on the schedule are based on the air temperature in the spring in Boston. And if it's not cold, like there's all these things that like, so you're really relying on how do you sell the best value to the customer? So it makes it interesting. It makes it fun, but it's challenging. It's like, all right. So then that bet is that like, right, if we do have enterprise agreements with one of every type of equipment and rep in the country, does that give us enough capacity to have insights into projects, things like that? How does that kind of relate to other opportunities for us to provide value to our customers? Luigi La Corte: Andrew, have you ever been asked for this? Andrew Zukoski: Exclusivity? No, we Mike Powers: you Andrew Zukoski: of the ways that we ⁓ value to our general contractor partners is in the proposal and pursuit stage, where these progressive design build or even CM at risk jobs, the contractor is being selected to do more than deliver a well-defined scope at a well-defined price. And often, design is not advanced to the point that you can't even... You need some pretty big arrow bars on what this price is going to be, right? There's just not enough definition. And so the selection comes down less to who's got the best number and more to who do we think has the best understanding of the project and then who's going to be the best partner as we go through that. And people, our partners like to use join. You have an hour to demonstrate this basically and you need to convey a lot of information about the project. and you need to convey a lot of information about how you're going to work together. And so a lot of people do run that the way that they run their later OAC meetings where they have pulled together a few of the projects and joined. They're able to use join. Like, here's really quickly, here's how we see the project. Like, this is the major optionality we see. These are the risks that we see. This is where it's trending. These are benchmarks from other projects and why we think these numbers are here. And we have this implemented collaborative system where you'll have access to this and so forth. And so we don't, there is like a competitive edge I people tell us they win work like because of that. we don't ⁓ do We work really hard to not. Like, of sales teams can access customer data. It is the case that multiple parties have proposed on work, all of them join customers, all of them using join as part of that pursuit process. so especially in situations where it's like, let's say it's ⁓ the three finalists, two of them are joining customers and third one isn't, but realizes maybe they want to be, like it could be very tempting for us to really like go out a limb with a new customer and like provide them extra help or something like that. don't want to do that. Everything that we do, you can read on the website and we offer to our customers and they find value in it. But we're not part of the pursuit in the same way. We don't take on project scope in the way that Mike does. ⁓ It's a less fraught for us. Luigi La Corte: Yeah, that makes sense. know started a little bit late. Why don't we go to ⁓ our topic of the day? Which one do we want to talk about, CSI or RFI zero? Zero RFI. Mike Powers: I think the CSI fascinating. ⁓ Andrew Zukoski: See you soon. Luigi La Corte: Which one Mike? Andrew Zukoski: Which one, Mike? Luigi La Corte: All right, let's do it. Give me one second. So let's open up this post from David Stutzman, which you'll see here in one second. So David Stutzman, I'm saddened by construction specifications Institute. Yesterday I announced that I would decline a volunteer role on a task team to update uniform at. This is the first time I declined a volunteer opportunity for CSI since joining the organization and not a decision I took lightly. have longstanding involvement with uniform at nearly since the subsession 92. I served as CSA liaison. What happened? CSI formed a for-profit organization that is solely responsible for new licensing requirements implemented through the construction standard. They're going to control uniform and licensing and its profits. This is not the CSI I joined 46 years ago. It's not the organization I donated countless hours to. So it looks like CSI has a for-profit arm. ⁓ CSI, for those who don't know, it's a not-for-profit that creates a standard. That standard is used for ⁓ effectively identifying individual components of a building or construction work. You have master format and then you have a uniform at uniform format typically used by consultants, cost consultants, master format or experience are used by general contractors and estimators. It's effectively ubiquitous. And you talk to any general contractor who's using master format. They typically have a riff of their master format. So instead of ⁓ three, three, zero, zero, zero for casting in place, maybe it'll be slightly different with some decimals. but it is the industry standard. So clearly they recognize that and now they want to enforce their IP. Fair, but a little bit of a bait and switch. What do we think? Michael, pass it over to you. Mike Powers: What I found interesting about this is when we started building connected, we did not use CSI codes for this exact reason. Because there was no, even though there was this like 80 % of people use CSI codes, they were very inconsistent across the space. So we use keywords instead. And we built out BuildVision in similar way. ⁓ I think it's a very old school company trying to grasp at straws to try to value. their IP because they've seen what book dealers have gotten with open AI and CURS and Claude. And they're going to try to create a defense mechanism so that they can sue the large language models to get cash out of it. Cause if you go into Claude or CURS or Claude or JadGBT right now and ask for the CSI code for a cast in place, it will give you a very detailed response around 03-3-0-0-0. ⁓ And I don't know how it's enforceable. I'm not a lawyer. I pretend to be one on TV sometimes. ⁓ But it's not the first principal problem of actually like the value add. It's just like this, hey, this is how we've named things historically. As more agents come into our workflows, they don't care how things are named. So this is like, it's an interesting compelling event right now, but I think as the market shifts, this will just be kind of a roadblock and people won't even use CSI codes anymore. Luigi La Corte: There was a case, sorry, Andrew, go ahead, and then I'll add my point. Obviously a case was brought before a judge concerning Anthropic. Anthropic famously both purchased and pirated a lot of books to be able to train its model and were subsequently sued by authors because they thought it was copyright infringement or RIP infringement one of the two. ⁓ Andrew Zukoski: Thank No, go ahead, go ahead Luigi. Luigi La Corte: So this case was decided in 2025 and Anthropic won and so did the authors. And let me give an overview of how they won. So Anthropic won because the judge effectively claimed that if Anthropic parrots out a response to a query that asks a question about a Harry Potter novel, it's effectively like talking to a human who did the same thing. So they have fair use for training. The judge... basically said that they can use books to train models because it is exceedingly transformative and constitutes fair use. The authors were still paid though, and they were paid because Anthropic illegally download and stored, I think it was millions of pirated books from websites. So I don't know if there's even precedence for them to say that, okay, I go to an LLM and I can ask them, you know, can you rebuild master format for me? I don't think they have a case because there's precedence suggests that they don't. I think what they're doing, is they're going to people who use CSI, let's say it was Building Connected. Let's say in Building Connected, ⁓ for a subcontractor to respond to an invitation to a project, they say, okay, you're responding to Miss Metals, 05-50-00. That's baked into Building Connected. I think they would go to Building Connected and they'd say, you're using master format illegally, that's RIP, you're using it because now it's a core piece of your product, or even ancillary, now you need to pay me a licensing fee. On their website, they have licensing fees, per type of user. They are charging contractors, architects, engineers, software providers, and the amount of money you pay them depends on how much money you're making. You can go look that up yourself. Andrew, what do you think? Andrew Zukoski: the, ⁓ I think CSI is an organization that's in trouble. think a lot of it comes back to just the evolving like pattern of product. know I'm a total broken record about project delivery, but like eight years ago, I went to a CSI conference in Seattle, Washington, and sat through like a fully attended, like a packed room on the role of specifications and design build. And it was this, ⁓ large tech client like office build out and they were running through like how design and specification process was different in design build versus more traditional design build process. And like they ran through how they ran the project and how awesome it was and all this. And at the end someone asked like I think it was like, well how did the specification matter? And like the people on stage, one of whom was a specifier said it didn't like. that the spec is like a, you know, important tool for fighting in design, bid, build situations. It's also like where you spec like things like, you know, the temperature in Boston is going to determine the chiller, right? There's also a lot of like important technical information that goes into it, but it's like this, like, I mean, they're literally like, they're a thousand page club. ⁓ for like beating up other project stakeholders and they matter less and less and I think that means that CSI is a shrinking organization. I think they're seeing fewer sponsorships because the way that you sell building materials and the way that you sell chiller equipment no longer is you hunt down the specifier and you buy them a fancy dinner. It's you got to work with Build Vision in the case of mechanical equipment and ⁓ the result is that they're grasping at straws to try to find ways to do it. I don't think it'll work. think ⁓ the idea that there's standards is just so tenuous. Like we run into so many people who are still using not master format, but the 16 code CSI from 20 years ago. like people use uniform at level one. Once you get to uniform at level two, it starts to get pretty different company to company. And so I think it's so tenuous that they are actually the standard. And if you... ⁓ put more barriers to usage, it'll blow away. Like it just, it won't be used as much. Luigi La Corte: I also wonder what precedent says about this. Like you effectively have open source, open sourced your IP for decades and you have not enforced any IP rights that's proliferated. And now you're saying, I want to enforce that IP right. ⁓ Usually, at least from what I see in some contract law, sometimes you do lose your right to enforce a provision if you don't protect your rights or you reserve your rights. I'm not an IP lawyer. but I'm curious to see how this is gonna play out. You know, I wouldn't even be surprised. Yeah, go ahead. Andrew Zukoski: Well, like, AIA contract documents, like, there's way more intellectual property in, AIA contract documents, right? Like, the standards, there's just not that much there. It's just not that many pages of text. It's like some numbers and names and descriptions, right? It's just not that much. Mike Powers: The numbers, it's literally numbers. ⁓ And that gets augmented away in NEL and world because the numbers in systems actually matter because all the context around to solve it. Andrew Zukoski: I don't know that I agree with that. ⁓ Mike Powers: That's what LM is great at doing though. It's like, you don't need a database of like here's, we have to, the chiller specification is the 23, 68, 29. And you can go to 26, 68, 29. So you can find the, sorry, 23. You can find the chiller spec. It doesn't matter anymore if you know the context around what a chiller is. Andrew Zukoski: But what about, like the thing we hear about from our customers that's interesting is like many of our customers are investing really heavily in the work breakdown structures that they use internally. And the key thing that we hear about is trying to align between estimating and operations, right? Where traditionally estimating did master format and then operations had schedule value and different cost codes that they like executed the project against. And it meant that it was essentially impossible to round trip data because you're like, I got the chiller, but where was the chiller? How do I map from how I estimate to how I execute? ⁓ And people are looking to rationalize that. And it seems like you're not just going to have, not going to go do an ad hoc query on every project of, well, I'm estimating the chiller. What did the chiller cost on this last project? But maybe you think you will. Mike Powers: And elements were very, good at that. So having the context around it and like finding the trend. Andrew Zukoski: It's just like all the labels blow away in the wind. ⁓ Mike Powers: It won't matter. It won't matter. read your emails. It'll read your team's account. Luigi La Corte: I don't know, Mike. mean, I know you've looked into CSI. I know you've looked into the various branches. It's effectively a taxonomy, right? It's like a, it's a tree where every level you get more information. When you go down to the third level, which are the last two digits, and then you go into the dots, four levels down and the details matter there. And you know, was, well, you know who is, CSI. CSI is consistent. Mike Powers: Yeah. Yep. No one's consistent on them though. None of the engineers that are designing the buildings are not consistent. A great example, which we deal with this all the time is a VRF system, which is a refrigeration system that has a bunch of different compressors for it. And a lot of buildings are using them now. It kind of has a home, but then when an engineering firm creates a schedule on a drawing set, they might name it 17 different things. Andrew Zukoski: But nobody uses that stuff. Mike Powers: Um, so there's no, like, there's no delineage of like, um, why an engineer put this in the VRF package versus the split system package. There's like semi guidelines. So it kind of like creates this mood point where you going to 2365, 23.15 is supposed to be like an air scrolled, uh, water cool air, uh, scroll air cooled water chiller. Like it doesn't actually matter. Like that context because the. manufacturer that's producing that piece of equipment also has their own little spin on everything. Now they're starting to do like specific chillers that have like magnetic compressors. Well, where does that fit in this world? Like all these things like it's, it's an archaic taxonomy that actually doesn't have an outcome that's valuable. Like you're looking to provide a cooled air in the building. Andrew Zukoski: Yeah, okay. I guess coming... So I went, my son had his open house ⁓ at his kindergarten and we went by the school library and wondering around the school library I'm like, ⁓ yeah, the Dewey Decimal System. Like, I haven't thought about this in 20 years. Like, I spend much of my time thinking about books and finding books and like, I don't vaguely remember that there were more interesting things at higher numbers. Like, that's how Dewey Decimal System impacts my life now. So that's a fair... That's a compelling point, but like... Okay, so you're saying like all these taxonomy codes are going to zero. That's the... that like they... ⁓ Luigi La Corte: In construction technology podcasts parlance. Yes Andrew Zukoski: Yes. Mike Powers: So the reason you did it before was you have 3,000 pages specifications. The first thing I did on the first job site was at Turner. They're like, here's the spec book. Read one through 15, and then you read 23 through 27. It was 3,000 pages each. By the way, you read every single page. Great. A lot of knowledge in that. Nowadays, people copy and paste from every project. There's no consistency. specs don't match the drawing set. So what happens as you layer on these LLMs is one is going to automatically flag all the errors. If schedule an air-cooled chiller and the spec is a water-cooled chiller, AI automatically flags that. That's the first because it has the context of the project. It knows, hey, there is a cooling tower on the roof. and you spec'd an air-cooled chiller. That doesn't make sense. When you have a cooling tower, you have a water-cooled chiller. And that's a very simplistic answer, but I don't care what the numbers are behind that. It has the context of the building. Andrew Zukoski: Okay, yeah, so like if you're in the context of a project, if you're looking up information, you don't need the Dewey decimal system to find it anymore. It's great. Mike Powers: No. Luigi La Corte: Well, you do, but not CSIs. You don't need CSIs. You need the owners. Andrew Zukoski: not CSI. Mike Powers: You said package names here. You have a mechanical package, an electrical package. sure you're tracking some, I mean, right now you're tracking submittals on CSI codes. There's things around, need a CD and E around anytime. you know, are you complying, deviating or have an exception around the spec? But that's only because we've historically had, you did this on paper. So you'd have an Excel sheet that would say like, hey, did you get an approval for this spec section? Right, so that's a submittal log. Luigi La Corte: The reality is the owner is somewhat lightly basing their specification off of CSI or uniformat. They're modifying, it's different. Maybe it's historical to point. It's really just copied and pasted from the consultant who's working on the project and maybe makes modifications once in a while. But when the contractor's reviewing the document, let's say they're estimating, ⁓ see a chiller on the mechanical drawings, they are looking at the spec. They're doing a lookup. Like that will... that will go away. And I don't even think you need that organization. Like where I agree with Mike is that won't be mission critical anymore. The LLM for it to make a decision, for it to connect the chiller and the drawings to the chiller and the spec, it will not need a lookup. It will not need a table of contents. It just won't. But the owner is still deciding to do that. The consultant is deciding to do that. And that, I mean, they are using their IP. I think that's, I think that's true. Mike Powers: the The AI agent that is reviewing the submittal on behalf of the engineer doesn't care if you use Klingon or English when it's talking to the AI agent that submitted the submittal on behalf of the general contractor. Luigi La Corte: Yes. But the question is whether they're using their IP currently and whether CSI should have a right to charge for that IP. Because I agree with you, agents don't care and it doesn't even need to be, it could be in binary for all we care. Mike Powers: And we're the I Klingon, come on, it's obviously gonna be in don't... It's obviously gonna be in Klingon. ⁓ Andrew Zukoski: gonna be in Klingon. Come on, we're the Klingon contact, like, consortia, right? Like, we gotta go ship that mode in the software. Luigi La Corte: It's just whether or not you think that ⁓ a world where they are entitled to charge for their IP. Mike Powers: is going down the whole question around IP and construction. Like what is actual IP and construction? There's not even data ownership in construction. A designer designs buildings, the owner takes that information. Then they're sharing it with a ⁓ general contractor, it with a subcontractor. Every one of those levels, someone using some sort of tool to extract that data. And when a contractor is like, ⁓ yeah, what are you doing with our data? Well, what's actually your data? in this example. Luigi La Corte: In this example, well, it's the owner's drawings, the owner specs, the owner owns everything. They're ultimately paying. So they hire a consultant. say they own means and methods they own, well, they own the process. ⁓ Andrew Zukoski: I production rate. Mike Powers: So what is the general contractor on that example? What does that mean? From a software perspective, from a, have this this bucket. Luigi La Corte: Um, I'm just trying to, trying to connect this to the CSI question here. So what do they own? Well, they own the process for building it. It, depending on what the IP clause in the contract is, like the contract could say, we own everything. We own all IP generated from this, this project. What's the IP? I don't know. I'm not an IP lawyer. How we build the thing, the process. Mike Powers: I think that my pushback is like, this is an interesting first push into construction IP law. Cause there's going to be pushback from general. They're going to go after people. They're going to have to go after contractors and suppliers and software engineers or software companies that are not paying them. And then it opens up this massive can of worms of like, where is ownership and what can actually be owned? Cause ownership legally is a very, very like strict term. Um, there's five types of our ownership. Andrew Zukoski: Yeah. Mike Powers: And a lot of the things that are happening in construction aren't really ownership. So it's going to be interesting. think it's open to spin it back to the CSI question. I think it opens up this can of worms where anyone saying they're like training their own models. Well, what are you training your models on? You're training your models on the construction drawings? Well, who owns the construction drawings? Well, we worked with the general contractor. Well, you don't know the general contractors and own any those construction drawings. your, your tees and sees right now is, but your tees and sees right now is like, you're the, uploads the documents are taking the risk on that. They have the rights to share this information. There's going to be a huge backlash on this. There's a reason why Turner signed an open AI enterprise agreement. Luigi La Corte: It kind of feels irrelevant though, right Mike? Andrew Zukoski: ⁓ is there a little bit into it? Mike Powers: Because in that example, they're not training on any of their data because OpenAI cannot train on the data in that enterprise agreement because Turner doesn't own any of the data. Luigi La Corte: You're saying here, you're saying that the one of reasons for signing with open AI is that they're not trained on the data. But that's usually a very common MSA or terms of service carve out and ⁓ with a company like Turner, like usually like the large companies that we work with, the number one red line you get is you can't use my data or how are using my data. ⁓ Mike Powers: their data with their data them, that example. Luigi La Corte: Well, without going up the chain, what they give us, because we're not auditing where this data originally comes from, right? In the exchange with two parties where I'm one, let's say Colas, Colas is saying, I'm gonna upload my contracts. In that case, that's their data, right? I'm not gonna litigate that and say, well, is it actually your data? The point is, it's not material for me. But tie it back, bring it back to the CSI. how do you think contractors should be thinking about this? If I'm a general contractor and now I'm being told, you're using my internal codes in your estimating should be paying CSI for this? Because it seems like they want them to. Andrew Zukoski: just ignore it. mean, like somebody's gonna. Mike Powers: Why don't we release our own open source code? We'll do our own open source one that'll be free and everyone can use it and we'll never charge a penny for it. And we'll go get everyone more comfortable using the code. Correct. Andrew Zukoski: Yeah. Luigi La Corte: Introducing Klingon based equipment... Andrew Zukoski: Yeah, Klingon. Yeah. The Klingon work breakdown structure, the future of construction technology. All right, gents, on that note, I gotta go. Mike Powers: video. Luigi La Corte: Going to zero Good senior Andrew. All right. We'll cut it. We'll cut it short here. Good senior Andrew talk next week See ya see ya Mike. Bye Mike Powers: Shit. Andrew Zukoski: Thanks guys. Mike Powers: See you guys. Andrew Zukoski: Bye.