Latest / Construction Technology Podcast / Construction Technology Podcast - Episode 007
Transcript
- Luigi La Corte: What's up everyone. Welcome to construction technology podcast. I'm here with Mike from build vision and Andrew from join. What's up gentlemen. How's your Friday going? Mike Powers: Good morning, happy Friday. Luigi La Corte: Friday. All right, we're talking this morning about two things. We're talking about AI as usual, and we're talking about ontologies first. So ontologies are slung around a lot. I first heard about ontologies through Palantir. They really popularized the term when they were talking about Foundry and their custom ontology that they wanted to as the benchmark the baseline for every product and build on top. I think people in construction are hearing the word ontology. Andrew Zukoski: Good morning from the West Coast. Luigi La Corte: we don't really know what it means or not everyone has a fulsome understanding of what it means. talk about why it's even relevant. So in learning. You first need to understand the topics the kind of discrete areas of interest. ⁓ people call them taxonomies. You can also call them as ontologies. They're slightly different. but you need to understand how I put certain data in certain buckets. So if you're talking about like a, let's say a document ontology, you have different types of documents. You have specs, you have contracts, et cetera, et cetera. Gents. Well, first of all, ⁓ when talking to customers, are people asking you these questions? Are they using the word ontology? Do you find that they don't know what it is? Like, should we explain it? What do you guys think? Andrew Zukoski: I a lot of, I had a conversation yesterday with a large owner developer and it sounds like they're in a tech role, a lot gets funneled to them. but they're being bombarded left and right with people slinging ontologies and forward deployed engineers, which are two terms that no customer outside of that has ever said to me, but I think there's a lot of vendors making noise about this. So I think defining it would be good. Luigi La Corte: Mike, how are you? Are you seeing this with customers? Mike Powers: Yeah, we're definitely pushing it out to customers. We have forward deployed engineers that sit with and build out workflows. And, you know, our position and our thought process is like, like if you're sitting at a general contractor, they're using like ProCorporate PM, Primavera for scheduling. have CMIC or SAP for ERP, ⁓ versus Design, and they've got hundreds or maybe not hundreds, but dozens of subcontractor tools. And one like has its own idea of what a project is. ⁓ Um, so what Palantir's pitches or what ontology is, like ontology is the layer above this. Um, so it's like trying to define and share like what a project is, what an activity is, um, like what a crew is, what a PO is, really anything like thing that's connected between the two. Um, one, one, um, analogy that I heard that was interesting is like, um, the current stack is a bunch of instruments in an orchestra and then the, um, the sheet music. is ontology. So it knows how to like read and pull and like push everything together. So that's how like positioning it to customers. And all that means is like, hey, you've got thousands and ⁓ hundreds different silo data points. We will go figure out, normalize, bring all those pieces together and give you insights that you want. So if you use Outlook for email, we'll plug into your Outlook so we can read emails to get construction documents. You use Salesforce or Acumatica for your CRM. Well, we'll push any data you want to Salesforce or Acumatica. You get bids in building connected. All right, cool. We'll engage with building connected to get data. You use Procore for change orders and submittals. Cool. We'll connect the dots there so you don't have to see anything. And our entire pitch is like the user doesn't have to even engage with BuildVision as a platform. There's an agent that does every one of those steps and then they get their own agent that's branded to them. That's contextualized to them. That's harnessed to them. and just responds to their email, responds to their texts. And that's how we position ontology with that kind of idea of like the clue between all those different parties. Luigi La Corte: So if you're in construction, you're a general contractor and someone says, I want to give you my ontology, we'll give you products that are using our ontology, how would they separate fact from fiction? How do they know someone actually has proprietary data, a data set to have built an ontology versus, you know, it's all marketing fluff? Mike Powers: So that, think that's like the, I don't think that's the, that I would push back on initially, like, ontology is not creating that new data sources. Ontology is the glue that connects all the data sources that you historically have. And I think that's like the differentiator. if someone's like, Hey, we're creating this like ontology data lake, or we're creating these like new data sources by doing X, Y, and Z that ⁓ the point. Like there's connecting the nodes of your existing solutions. And then there's what These called orchestrations, which are like tools or workflows on top of ontology that can do actions. That's all ontology is. It's not, hey, we're going to go like take pictures and store project data in this specific instance ⁓ of It's connecting the glue between those tools. And then there's tools and orchestrations around that that provide value. Andrew Zukoski: How is this different from an SAP in implementation? Like I got to imagine SAP walks in, they say like, look, we think you can do everything on our platform, or Salesforce walks in. We think you can do everything on our platform, but in of fact, you're have a bunch of other data sources ⁓ and tools ⁓ out And you're want to be able, we're going to work together to model like what, what a deal is in Salesforce or what a customer is. then you build a build things that interact with it. Like, is there a difference between school ERP data modeling and implementation and the ontology way? Mike Powers: So is like the system of record. I say like ontology is the layer above the system of record. And I know that's kind of like, it's like such a fuzzy word, but like, again, you're not, again, I some customers are replacing SAP with Palantir tools, but ontology wise, ⁓ the ontology would be like, hey, I can go grab your SAP data and connect it to your Procore data. That's ontology. ⁓ what Palantir sells, Palantir sells like, hey, we can rip out SAP and create a new ERP system. And that's different, that's orchestration, not ontology. ⁓ So I think that's kind of, that's where I would differentiate it to. Luigi La Corte: When I'm with you. Andrew Zukoski: So if I'm buying an ontology, I'm not buying a a substantial new database. Mike Powers: I don't think you can really buy, that's what's very hard, I think it's hard to like buy an ontology. Like you can't just be like, hey, I'm gonna go buy this and like it's gonna solve a problem. Andrew Zukoski: I'm going to organize my business and the various digital tools that serve the business and I'm going to be able to map like this business concept that we have a crew and productivity on it and what our holiday schedule is and I'm going be able to connect the dots between these things. Yeah, okay. Mike Powers: And that's what Kavanaugh, the contractor that's like kind of been the first kind of foray into construction, Palantir has talked about is like, hey, we like normalize all this data and now we have like an app on our phone that like lets us do T and ⁓ time and materials, right? And like that time and materials, that's not ontology, that's like an orchestration off of the data structure that's been created. Luigi La Corte: Hmm. So an ontology would be, hey, we have this ontology that we're gonna use to handle data to the correct place, right? Okay, so let's say within my ontology, I have RFIs as part of that ontology, right? So to recognize that it's an RFI, what I'm assuming Foundry does is they're saying, hey, we've seen enough RFIs, we have enough data there to be able to proactively... automatically identify this as an RFI so we route it to the correct place. Is that right? Mike Powers: No, because the, the bright position would be the RFI is created in Procore. Now we are taking that RFI data node and connecting it to my scheduling Primavera node, connecting that glue, that glue, that connection that between those two nodes is the ontology. Yeah. Luigi La Corte: I see. So it's not the identification. It's not the taxonomy. It's a layer above the taxonomy. Mike Powers: Cause you're gonna normalize it as part of this implementation process. You're gonna normalize it ⁓ around the textometry that you have, textometry you have set up. And that's where there's forward to play engineering that's where we're spending all this time and resources because it takes a lot of time to go, hey, let's go through your process. We have all these slots and here's how we think about the different pieces. Luigi La Corte: Okay, got it. So your ontology is really how you're routing individual taxonomy items to a location. But if I don't have a data source, let's say like I'm not building, I don't have anything in Procore. Okay, I just have a SharePoint bucket and I have RFIs. Can I benefit from your ontology? Or do you need data to be able to tell me, well, like this is an RFI and I'm going to route this to the appropriate place. Mike Powers: It didn't work. Andrew Zukoski: the ontology live inside the business? Sorry, Mike, I'm cutting you off, but just to check my understanding, like the ontology lives inside the customer, like, and it doesn't, it might, you build it for the customer and it might share patterns with other customers, but fundamentally you don't take it off the shelf in the same way that you don't take someone else's business off the shelf. Mike Powers: Corrected. Correct. It's like, it's like, it's not a data warehouse. It's not a BIM model. It's not like a BI tool. It's like a shared operating tool or model for the business. And then like, like one example I've heard consistently is like it turns nouns and verbs into governance software, governance software. So a project, and then there's a verb on that project. And it's like that governing process of those two. So if an RFI is created in Procore, then it has to go to Primavera. governing. It's like those two nouns that there's a project that needs to have a consistent node there. So that project now is in the ontology. It's project Luigi towers. So there is the project Luigi towers in pro core. There's a project Luigi tower in primavera ontology is saying here's the source of truth per project Luigi towers. And then what that allows you to do is that give you like that shared definition across all the different tools. Luigi La Corte: I Mike Powers: And then you can start doing actions on it or verbs on it. And that could be orchestration. So in that example, if you don't have an RFI tool, you don't have a Procore, you would create an orchestration or a tooling around how do I do RFIs. And you could use ontology to create a tool around an RFI approval because it has to go connect to these parties and knows where to go. right now, I'm guessing it would not be out of the box. One interesting model that they do build towards is their goal is, hey, We do, they call it gravel road to harden highways, which we've like kind of used as well. was like, they'll forward deployed engineer a workflow with a customer. It's a gravel road. So, Hey, we built this app for our customer that does TNM. If the foundational team on Foundry, for example, that sees value across the entire platform, they'll harden it. And now there'll be a TNM app that's available for anybody to use. don't have to rebuild it. And that's where it breaks away from the consulting model. that like an Accenture or somebody has historically leveraged where it's like, hey, we're gonna sit here and charge you millions of dollars, just build your super, super custom software and everything, something changes, you have to bring us back in and change it. It's between, we're a full custom dev shop, like Accenture to, we're an off the shelf building connector. Andrew Zukoski: Because Accenture doesn't own SAP. SAP no doubt looks at what people build and hardens things into the system. Kahua, I know their whole model is, they have this like programmable thing and then they harden application. They actually, they get paid to do it. Right. But it's if you have, if you have the same party doing the implementation that owns the like technology that's going to express the ontology, like you're able to harden repeated patterns. Mike Powers: Correct. Correct. Correct. Correct. Correct. Correct. Andrew Zukoski: the system. ⁓ Mike Powers: So when they say they're going into construction, they're saying, hey, we've got 10 orchestrations that are construction specific, so we can go in out of the box and it's not as much as a custom built. Andrew Zukoski: Okay. Luigi La Corte: So if you're approached by a software vendor who's not palantir, okay, they're slinging ontologies, like my pushback again is if I were a GC and you're telling me what an ontology is, it's such an abstract concept. Andrew Zukoski: Go. Luigi La Corte: You actually don't want to talk in terms of ontologies. You want to talk in terms of outcomes and workflows. This is what the GC understands. But I if sitting in where I currently am as a ⁓ builder, if someone's going to say, we have an ontology, we have some understanding of how we orchestrate, let's say, taxonomy items into final workflows and products, would want to say like, well, I to understand that you have a track record from a taxonomy perspective. You need to have data. You need to know, let's say what an RFI is from an ontology perspective. I'm assuming you just need to have a track record. Is that the right way that a GC should be thinking about this? How do how do you give them a bullshit detector so that the word is just not being slung around inappropriately? Mike Powers: I think if someone tells you they're selling ontology and construction that's not Palantir, it's bullshit. Andrew Zukoski: I don't know. I actually, want to hear like going through this, right? I've been learning. I'm learning here. I want to push back on what I, I want to counter what I said earlier about how I'm not hearing this from customers. one thing we hear a lot about from customers, they don't use the word ontology, but like, the project record is centralized in the project record, just having like a place there's a record that is definitive. And it's not just like, well, there's a project number and we, we pray that it gets typed in consistently across every application, right? Like they have. Mike Powers: Yeah. Andrew Zukoski: They've pulled that out and they put it somewhere and they've mapped how it maps out to all the other different applications and systems that do it like they're building that piece of the ontology and then they're starting to like define that would happen consistently on top of it and that might be expressed in code So I feel like actually they are like a of people are are going this direction and so customers are doing it and then Isn't it seems like there's a layer of integrators would be compellingly like that could go do this, right? Why is it just Palantir that can do this? Like why, why can't build vision show up and say, Hey, we're gonna, we're gonna, we're gonna work with you to define your ontology, like map it out and execute on this. It's just that like, it's so complicated and sprawling that only companies of Palantir scale can do it. Mike Powers: From my understanding, think structurally a lot of companies are not built this and their infrastructure is set up. Like I know there's been a lot of struggles with Procore and how does Procore talk about a project between a subcontractor and general contractor around the same project. I don't think Procore has officially solved that problem still. So there's pain points If you look at like a Kahua, I mean, who is this kind of a workflow tool, despite what they say, like on a system of record? I know it's tough. It's like, very, it's, very like tough. think like your bet is that you're hiring the engineers that go work at Palantir are going to be very, very intelligent. And the foundational piece of Foundry that they built, they've made very structural positions about how to build a software that I don't think is easily replaced. And it goes way above my knowledge of how software is built, that's been, people that are smarter than me have told me that position of how Palantir has built ontology as a structural piece, as a foundry or whatever system it is for whichever industry is very, hard to replicate. Andrew Zukoski: Yeah, I feel like the, if I were Oracle, I say, that's bullshit. Like, come on, we go in and we, we, we grab our arms. Yeah. Mike Powers: I'm so Oracle's different, like right. Yeah. Oracle's different, but like Oracle hasn't been able to execute. Like they they buy, they buy a construction software company, like, it hasn't even been fully integrated into their entire Oracle platform. Like with Primavera, like that's, that's like my pushback is like Palantir would make the bet that like there would never be a standalone, they wouldn't buy a construction tech software and have it still operate as a construction tech software. Andrew Zukoski: Yeah, sorry. I don't mean Oracle applications. I meant the Oracle ERP line where they have been successful. It would be interesting to know whether it's ERP implementations or application layer payments and Primavera and Aconex and so forth, which drives more revenue to Oracle from the construction sector. the spend on these ERP tools is... Luigi La Corte: I would bet it's Primavera. Like Aconex, I don't remember the revenue numbers when they were purchased, Primavera seemed to be like the de facto scheduling software for the industry up until like the early 2000s. Now it seems like, I don't know, like Microsoft projects, a of other tools, but like I still see a Primavera and P6. Andrew Zukoski: chunky. I bet Textura. We gotta figure out how to answer this question. ⁓ Luigi La Corte: We need someone from Oracle to come on and tell us. Andrew Zukoski: Yes. And the reason this is a side note, but Textura's pricing model is a brilliant way to tax the entire in a way that nobody knows. Well, in the end of the day, the owner pays for everything, right? And like Textura gets written in. ⁓ Mike Powers: subcontractors. So it does. So I'll push back on that. So my dad works at, my dad works for Kone, a massive elevator contractor. They have to absorb all the textura fees and for him to get approval on a $500 textura fee is a pain in the ass. And he goes, I hate using textura. I'll just talk to the contractor and pay the contractor directly as a company that has billions of dollars of revenue and has completely cashflow for him to get approval on a $500 expense for textura is so much extra friction that's unnecessary. Andrew Zukoski: Yeah. Mike Powers: and he's not billing back the owner, they're just like writing it off. Andrew Zukoski: ⁓ but they, alright. The way I've heard it, and you can do it, that you know who the texture of shops are, and so you pay for it out of your bid. Like, you don't bill it back to the owner directly, you say, like, okay, like, Luigi GC is gonna hire, like, he's gonna hire Andrew Zukoski's shitty subcontractor, like, to... do doors and windows, the first project where I'm like, wait, why do I have to pay? This is like the key thing that Textura did. Luigi might pay a subscription to use Textura to administer it, but if I want to get paid, I have to pay 500 bucks a payment or whatever it is, like crazy money. The time I do a job with him, I'm... ready to throw bricks at him because I have to spend all this money on it. The second time, like, ⁓ yeah, he's using Textura. Like, ⁓ it's in the contract. Like, just turn the knob up a little bit on that bid. Because I also know that everybody I'm going to compete with on the job is going to do the same thing. Right? Like, won't impact, like, baking it in won't impact my competitiveness, like, for the job. And so Luigi is able to go to the owner and say like, look, these are the three competitive, these are the three bids, here's the range, let's pick the one. And so Luigi might not understand exactly how much money texture is pulling out of the system. The owner never sees it. Like nobody act, nobody except for Oracle is able to attribute how much money they're pulling out of every project. And I think it's a lot. Mike Powers: Yeah. I hate, just like the rent seeking model is just like so, so poor. just like gives a bad taste. Andrew Zukoski: ⁓ it's, it's, it's, I, I Luigi La Corte: I'm not even super familiar with what they specifically do. So it's payment software. Andrew Zukoski: Yeah, it's peanuts. Mike Powers: So TechSure is payment software, it's advanced pay. So hey, like we're gonna pay you net 10 as opposed to net 30. And then you're willing to like take less of your payment to do that. So, and then there's a fee structure around all those pieces. I've heard of companies that are trying to like do it the other direction, which I thought was interesting. I talked to one of my friends that mentioned about of a company that like is creating a... not a liquidity market, but like a clearing house around payments of like, Hey, here's my outstanding payments. Here's my tears, how much on the dollar discounts I'm going to absorb to like clear today, as opposed to clear next week. Um, think it's an interesting model to push back on a texture up because it's, it's the opposite direction. It's not taking the fee that goes to Oracle. It's just taking the, Hey, I have a dollar outstanding of, um, you know, accounts receivable. I'll take it right today for 96 cents. And then they just take a small transaction fee in the middle of it, which is interesting. It just going a little bit the other direction. Luigi La Corte: So if we bring it back to ontologies you're a general contractor and you're to implement software, do you you're gonna look off the shelf? And when do you think you're gonna say, actually, I wanna work with a forward deploy engineer. I want someone like a Palantir to come build something custom. Mike Powers: Unless your revenue is above a billion dollars, buy something off the shelf or use Claude Co work to do it yourself. I would, I would. Luigi La Corte: Okay, let's say it's not Palantir. Let's say it's not Palantir. I would actually, okay, would push back massively on the Claude code. I don't think any contractor now should be building with Claude code. Mike Powers: Arp. If you're doing a singular point solution that does not engage other parties, think Claude Code, Claude Cowork can get you very, very close to what you need to do as an individual user. That is like, so that's like kind of, everyone should go try, before, that position to anyone listening to this podcast outside of our moms, if you have a pain point that you're trying to solve internally, go spend $20 a month, download Claude Cowork onto your desktop and Luigi La Corte: I agree. Mike Powers: Ask Claude how to solve it and it will tell you, know, give you a good pass. And if it doesn't work, talk to Luigi because Luigi is able to solve it because he can supercharge and do complicated things that Claude's co-work cannot do. That's why. ⁓ Luigi La Corte: We, yeah, I agree with that. I agree. Like you should know what it's capable of doing. You should. But I don't, I wouldn't recommend someone vibe code an internal tool that's going to be used across many people where it's mission critical. Mike Powers: Correct. Correct. Luigi La Corte: You have to maintain that software. You still need to deal with error handling. God forbid it fails silently or when you really need it. There still is a chasm between something you can vibe code and something that's production ready, that's reliable. I think maybe that will change. Maybe that will change. But still, let's say it's not Palantir. Let's say I want to engage Build Vision. Okay, I want you to build something custom for me. Mike Powers: Correct. Correct. Correct. Luigi La Corte: should contractors be thinking about this, Mike? Let's say like if you were giving your mom advice on how to let's use the general contractor, when should she look off the shelf and when should she, let's say, engage with you? ⁓ we just lost your audio. Mike Powers: I muted myself, sorry. I think it's a size position where you think it's a competitive advantage for yourself. If you're just trying to manage RFIs, submittals and change orders, and it again doesn't tie to a competitive advantage, I think you should go off the shelf. If you're trying to do something that's like, I'm trying to get ahead of my customers, or not my customers, my competitors, I think that's where it makes sense to leverage, because the cost is high. For Build Vision, Build Vision for an enterprise customer is the lowest contract we have is high six figures. And the reason for that is we tie to tens to hundreds of millions of dollars of increased revenue, increased gross profit or reduction in cost of sale. That's what makes custom valuable. If that can be proven out. If it's not, again, I go back to this world of like, if it's a workflow tool, you know, try to do this in Claude Co work then grow from there. If it's multi-party, I would look off the shelf for tools. But you're like, hey, you know, I want to, I want to create a new revenue stream for my business. So a great example is talking about my favorite my favorite company, Turner, that just released or just started first equipment company. The business model around that is fascinating. Turner owns the insurance policy on a project. So they have CSIP and SDI with that their safety policies. You cannot use a six foot ladder on your job site, on our job site. So what do need if you don't have a six foot ladder? You need a scissor lift. Who actually rents out scissor lifts now? Well, now Turner does with First Equipment Company. So thinking about, hey, I want to go build a software to help First Equipment Company manage their processes. That is a huge opportunity to build a custom workflow software for Turner. because they have SAP and they have Procore and I can create an ontology layer, an orchestration layer to make sure that every project that comes through Turner's pipeline, the first pass on any type of subcontractor that any kind of PO that goes out says, hey, first equipment, you need to go after this company to rent them equipment. I think that is a great, great opportunity. It's directly tied to Turner's bottom line and it increases their revenue. That's where custom. ontology, Palantir world, that's where it makes sense. If you're like, I am on a project and I want to manage submittals better, I would lean towards Office Self Software as best as possible. Andrew Zukoski: Okay, so I feel like it's starting to click more for me, like where the ontology fits. if I'm, I, customers that we talk to have increasingly complex operations with an increasing number of revenue lines. Like more and more is getting self-performed, I don't just mean work on site. More and more folks are setting up centralized equipment purchasing or rental like stuff. And it's like, if I'm running that like, yeah, you probably want off the shelf fleet management. Like. Mike Powers: Yep. Yep. Andrew Zukoski: You don't go build fleet management, especially if it's a new business line before you take it off the shelf. The trick is you have this increasingly complex operational beast and you need to be able to connect the dots between like this thing that exists over here, which is how the equipment rental company thinks about a project and this thing over here, which is how your CRM thinks about projects, like so that you can line them up and get them going quickly. And so one would be Mike Powers: Yep, correct. Andrew Zukoski: If your business is expanding its number of revenue lines and offering a more complex offering that likely will drive a higher margin, you're going to spend a lot more money on IT. Mike Powers: So I think, if I was Palantir, my account was Turner, this is exactly how I would have positioned this workflow. Hey, you use Procore, SAP and some field management software. I am going to create a tool, an orchestration that every time a PO has created an SAP, I'm gonna use an LLM to scrub the submitted purchase orders from all the subcontractors. I'm gonna use an LLM to figure out how many scissor lifts do they need? And I'm going to create a report that shows up in the head of first equipment company's email every morning that says, hey, these seven projects were just awarded. You need to talk to them about scissor lifts. And then the next step would be, ⁓ let's even get into the purchase. Let's get into the before the bids are done. So now I'm going to connect to your building connected account and say, hey, every one of these bids, all of the bid forms need to say first equipment company is, is renting out scissor lifts. What's your deduct for that? And then that deduct is it calculated summarize. And then my pitch from Palantir would be, I can go do this for you. I'm gonna be able to increase the revenue of First Equipment from a billion dollars to $2 billion this year by doing this. I want 20 % of the gross margin that you make on this. And if I don't do any of that, if I don't prove any of out and it doesn't grow revenue, you don't have pay Luigi La Corte: my push But my pushback on that would be, so you're saying if you're making over a billion dollars in strategic advantage, you should go hire Palantir. But there are a lot of other people who build custom software besides Palantir who are probably capable of doing this. And so say any general contractor or subcontractor, ⁓ think at this point has room for improvement. There definitely is. From some process improvement, some way to cut costs, reduce overhead, that definitely exists. But you don't think that they should engage with someone. to help them optimize their processes unless they think it's a competitive advantage themselves. Cause see now a thriving ecosystem, at least what looks like a thriving ecosystem of consultants and custom software implement implementers targeting construction who are using their relationships. So they a construction industry. say, you know what? Now I'm going to be ⁓ a consultant. And what going to do is I'm going to go. Andrew Zukoski: Yeah. Luigi La Corte: Basically get kickbacks from software companies so I can promote my product and I'm going to go use my relationships with subcontractors and I'm going to sell them those products and take a sales commission. I mean, it's a little bit. It's channel sales. Yeah. mean, it's, it's. course. Andrew Zukoski: mean, that's channel sale. That's like, that's like Autodesk. Autodesk did this around their BIM ecosystem, right? Like where they, they're still, all the resellers are still out there. Cause it turns out like off the shelf rabbit doesn't get you very far. Like you need this other stuff that just lives inside of it. It's a different set of problems than integrating between things. It's just like, I need more fucking buttons inside this tool. But they help, they, they're able to bundle. Mike Powers: Yeah. Do you still do it? You Andrew Zukoski: They're not a forward deployed engineer because they're not actually a part of the vendor. You can't take the lessons there and bake it all the way back into the underlying tool. It's not integrated that way, but they come in and they provide training and implementation consulting and then customization and integration so that you can actually realize the value of like whatever the tool is. Like that's. Yeah. Yeah. I think that's a good model too. Yeah. I think that's, I think you need last mile on there might be a little bit of code there. Luigi La Corte: Yeah, I'm not pushing back on that. I still think that's a good model. It's a good lost model. Mike Powers: I think. Yeah, and the value there is that you're helping those customers implement a cloud or an open AI internally and building out the tooling for them to do that without them having to do it. that's going to get commoditized away because, these models will get better and better. I don't think it's going to zero, but everyone has access to the same tools. There is going to be an arbitrage because everyone thinks there's an arbitrage of construction because people think construction people aren't tech savvy. Luigi La Corte: saying that's going to zero. Andrew Zukoski: Going to zero. Mike Powers: I will still push back that building a data center or hospital is the most complex, most technical thing, maybe outside of building spaceships that go to the moon. And I think that's where people are going to get burnt, where they're going to go spend $20,000, $30,000 to sign up for someone that's like, hey, I can help you build a bid leveling tool in Claude Cowork right now. They'll spit it up and it'll break as soon as they start doing multi-party. Luigi La Corte: Yeah, it's not really about the product. I think it's about the design. It's about the integration into their system. So I think that the lines between somebody who can say I'm going to implement software for you using third parties and now I'm actually going to do some custom code myself. I think those barriers to entry are reducing. And so they're not hiring a Palantir, but they may say to a third party, hey, you're pretty good. You understand the market. Who should I buy for contract review? But also help me it into my system by building custom stuff. Now it's not. You know, it's not so different. It's not so different than let's say custom software. And I think, I think that's probably a good thing. I think there needs to be more than just implementing off the shelf software. It's obviously more, it's more, it's human behavior change, but now it's bringing it into the systems we use. And so now like if you squint, it's kind of like a, like a makeshift Palantir from Wish, you know, it's a little bit different. Mike Powers: ⁓ So I think the better positioning for that is what Sierra just released. Sierra is Brett Taylor's customer customer success company. What they released, so they built an agent for your customer support success world. So, hey, I'm on ⁓ DoorDash and I a because they delivered Taco Bell, they forgot my Chalupa. Sierra would be the first party of engagement and they would... if appease the customer and close the case, Sierra gets compensated for that. If it doesn't, it goes up to the next level of support, Sierra doesn't get compensated. So they built this really interesting framework around agents. And what they've done now is they've just released what's called Ghost. Ghost is a custom agent builder in their application that lets customers build whatever agents they want. So if I was Procore, Procore acquires Data Grid. Grid gives them the ability to build agents. How do you put scaffolding around Helix, around new Data Grid acquisition to let customers auto build their workflows, leveraging LLMs to build out these workflows that are even better for these agentic workflows? I think that is a very, very value creative process. But I the challenge for that is it's tied to a top line revenue number for the general contractor, or it doesn't reduce the cost of goods sold on the bottom line. So for them to charge, you know, three, four X for an agentic outcome on a submittal review doesn't mean they can charge, can't charge three to four X for a submittal review because it doesn't actually move the needle on those two metrics of does it increase revenue or reduce the cost of goods sold. Luigi La Corte: I mean, why wouldn't it reduce the cost of goods sold? If I need to staff a project with two PMs, but I can bring that down to 0.5 PM and my PMs on multiple projects. Sure, they're on site, they're going to site trailers, they're talking to the site super, but you know, there's a lot of time spent reviewing information, trying to find information, trying to review submittals. Like why would they not be able to do that, let's say with, with, with. Mike Powers: Yeah. I don't think that's been proven. So yeah, let's go to the example of, building a hospital. I've got 10 employees on the job site. 20 % of the time, those 10 employees is doing manual processes like submit reviews, change orders. 80 % is the field. So even if you make those 10 employees five times more efficient, none of those 10 employees can go to another job site. They still have to be on the job site from six to six. So that reducing cost, like could you do with eight employees instead of 10? I think that's a challenge. That's a pushback because those 10 employees, those 10 are probably each managing a different trade. And that 80 % of time is with a trade in the field, not with me reviewing submittals. So you have an MEP super and an MEP project manager. Do you you can combine those? Possibly maybe, but do you want to? I don't know if you want to as a general contractor, combine those into one super, super MEP on a hospital project. And you can go down the list and that's how they staff those projects. Sure, you could replace an admin. You're not replacing a PX. That is on seven projects now because you have an AI tool that does middle reviews for you or bid leveling. I think that's a very, very challenging position that hasn't been proven out yet. Luigi La Corte: There's always ancillary staff behind the scenes though. like you can look at the coordinator role, right? So like I always hear people saying like, I want a coordinator. need like an AI coordinator. Coordinator consists of like individual workflows. One may be submittal review. Okay, so now if I have one coordinator who used to work on three jobs, now they can work on six jobs. And my project director, who's managing many PMs, well, maybe they're doing four PMs they didn't have the bandwidth to understand the issues across more than four projects, but now they do eight. So I think if you extend those two things, that shows maybe the constraints of the PM level. I don't, to it seems like will go down. The more efficient, the more projects can be shared with individual resources, those ancillary resources. Mike Powers: So in that example, I would make. Andrew Zukoski: That's like pretty tough. That's like pretty... like shaving... shaving two degrees away from the hypothetical like RFI thing. Like it's like... like maybe, but I pay a lot of money for that. Like I'm not going accept a ⁓ 4x Procore bill one day in the I might have slightly fewer project directors. ⁓ Luigi La Corte: Yeah, okay, I agree, but do they need to forex their bill? I mean, that's probably something. Okay, that's a Procore problem. Mike Powers: They needed 10X. They needed double revenue. Andrew Zukoski: That's what they want to do. Yeah. Mike Powers: They need to. So think one of like, and I think this, I make the bet that the companies and the contractors that invest in their people to train their people to understand what it's like, understand why a smittle is approved or declined, to why an RFI is approved or declined, why a change order is approved or declined, not with an AI agent, but like as an individual will perform better in the longterm. So the companies that are like, hey, ⁓ We're going to AI agent away, submittal review, change order review, RFI review will not, be worse off in five to 10 years than the companies that do not do that. I like, know I like, I've, we've hired another person on our team from construction he always talks about his Turner experience. I know I like am a fan boy of Turner. Turner does a great, great job of talk ⁓ teaching and educating their early employees. Like we went away for Turner camp for a week and we all sat together and learned about construction. Andrew Zukoski: No. Yeah. Mike Powers: That is their moat, that is their value prop. And I've met of other people that have like shifted away from construction into tech. And everyone that's from Turner says that they're from Turner. Everybody else doesn't really talk about it. this point, Victor, Victor, so Vic, so this point Victor was Turner and Barton Malo. He talks more about his Turner experience than his Barton Malo experience. Luigi La Corte: Andrew, you disagree. Well, obviously he says what he says, but what's your perspective on that? Andrew Zukoski: Yeah, Sorry. I was gonna say, don't... Turner's... There, Mike. know Turner's very special. Like, Turner's very special. ⁓ Mike Powers: I'm just telling you, they do a great job of teaching their, but that's whole position is like that. Yeah. Andrew Zukoski: Yeah, yeah. I'm just here to say others do too. There's a like where we see, if we look at like our larger customers, the ones who like, it's not scientific, but if you're like the ones where I'm like, ⁓ I like to own equity in this company. I would like to have a piece of this construction company just of how I see them performing. Like one of the patterns is there's heavy investment in that onboarding and teaching. And that a lot of the AI applications that they're hungriest to apply are around like accessing that corpus of the Turner way or the Clark way or whatever, right? the, it's not a, ⁓ that they're looking for is not in like in auto approving RFIs, it's in can we accelerate that time for that PE to look like, can they look like someone would look like in five year, used to look like in five years experience with only two years of experience? Like that's what we want. Mike Powers: Yeah. Yeah. Luigi La Corte: But obviously you're not pushing back against people using AI to speed up the workflows. would be kind of hypocritical, right? if you're to vibe code a tool within Build Vision, you'd have to imagine that someone construction is going to try and automate a submittal review. And you can't say that, hey, you're not going to understand construction as a result. There's many ways to solve that. They may be human in the loop. imagine, Mike, GC's coming up to you and say, hey, there's many things I want to do within my organization. One of them is I want to reduce the time spent reviewing submittals, right? So that I can approve or reject or submit with comments and revise a little bit faster. What would your recommendation be? Mike Powers: Yeah. So the positioning that I'm taking is like AI can help to supercharge those individuals to do that workflow, but it's not replacing that workflow. So when you're charging for those workflows, it's just a nice to have time savings. And when you extrapolate that out, that time saving then doesn't replace that PM. It just supercharges that PM. So when you're trying to tie to ROI, that's where I think it's a, I don't, I, if so, I don't know the answer to that problem. Luigi La Corte: Okay, I agree with you. Mike Powers: Like I think that's just a challenge in the industry. It's like, I think it's great that now I can review an RFI quicker and get a response quicker. It's great that I can review a submittal quicker and get a response quicker, but it's not just automating in the background and now it's just solved. So you're not replacing that time. Like it took me one minute of an hour to release an RFI. I'm not going home an hour early because of that as a super on a job site. Luigi La Corte: You may, may go home an hour early. Mike Powers: You can't, you can't. have to the job. As a general contractor, you're the first one on the job site and the last one to leave. Luigi La Corte: Well, if you're staying late because you have a backlog of submittals to review, you're probably going home a bit early. Mike Powers: Okay, that's outside of the six to six time. So that's six to eight. Cool, you can press six to eight. You're still there six to six. Andrew Zukoski: Thanks. Yeah. Compressing 6 to 8 seems like a good thing for families everywhere. All our moms would appreciate that. Luigi La Corte: Yes, yes, that's not changing. Yeah, I agree. Mike Powers: So you're billable hours, right? It's a great thing, but I'm not saying it's a bad thing. a great, great, great thing. But that's to six, which you're really billing eight to four at to the owner. already overextending that hourly rate. You're not compressing that six to six. You're not getting from eight to four to eight to two. So ⁓ there's monetary billable hour cost savings because of Andrew Zukoski: Yeah, it might be like you get like, it's slightly easier to hire because you're known for getting your PMs off the job site a little bit earlier. But that's like, that's this like, and that's a good thing, but that's a second order thing. That's going to take it like, yeah. ⁓ It's going to be hard to. Yeah. And it's to be hard to build. Mike Powers: Sure. Luigi La Corte: Yeah, you're not saying it's not valuable, you're just saying the value is marginal and it's not a needle mover. And I can see that. Mike Powers: Correct. Correct. If you're charging six, seven figures for trying to tie to revenue or reducing cost of goods sold, that's the whole position. Luigi La Corte: Sure. Andrew Zukoski: Yeah. Yeah. Yeah, but 20 bucks a month an employee for a chat GPT or a cloud subscription. Fuck yeah. Luigi La Corte: Yeah, and that's part of the argument. Mike Powers: Let it rip or okay, it's a hundred bucks a month into the plan grid. It's the plant plan grid model. Great. Let it rip. Plan a thousand bucks a seat. Cause that's the shift, right? Like, Hey, we brought internet to the job site with Procore. We brought documents into the field with plan grid. What is the next step? You bring an AI into the field. Cool. It's a workflow tool. It makes it. It's great. It's great. I'm not saying it's poor, but it's not, it's not like this. Hey, now construction sites are going to, you know, open in 12 weeks as opposed to. three years. Luigi La Corte: So what needs to be broken to make this work? Where would you invest your time then to move the needle on revenue? Because if you're saying the sub is really constrained by, site super at the GC is really constrained by interacting with subs and that's not going away because it's still a very human task. So that's what the choke point is. Where do you need to invest if you're building software to get that 2X on revenue? Mike Powers: tied to revenue driving forces. And shout out to my friend Justin at Shepherd for just raising $42 million to drive revenue in construction around insurance. Again, this is outside of I'm on a job site doing automated smittle workflows. They're allowing brokers, insurance providers, contractors to get better rates on insurance using technology. That is a direct bottom line revenue driver. Hey. Luigi La Corte: That's it. Mike Powers: It cost me a million dollars last year to get my GL policy for this project Shepard. got it for 850 or whatever the math is. That a direct bottom line driver. ⁓ that's in the tech solution, tech world where how are you tying to that outcome? Yeah. Andrew Zukoski: Yeah. Mike Powers: My biggest regret is that I did not invest, if I could have invested. Luigi La Corte: Maybe you can get the the Series C. Mike Powers: I wish. Andrew Zukoski: Yeah, worse than worse than worse fuckership. Luigi La Corte: be diluted less. We'll be diluted less. ⁓ many rounds. ⁓ Andrew Zukoski: Yes, that's what saying. Mike Powers: But yeah, but go, but go through the tech, like you can go through the tech, like the tech If I was a venture capitalist looking at investing in construction tech, where's the value props of like, how do you move revenue needles? How do you reduce the cost of goods sold? And I believe both of you guys are doing this as well. Very like that's, that's the pitch. But specific workflows on our SOR, it's It's tough. Luigi La Corte: Personally at Provision, we don't really touch operations and there's a lot of reasons for that, but I do like the idea to your point of revenue drivers and pre-construction is just like a revenue team. ⁓ like a sales team. They don't get the credit of a sales team, but they really are sales team. Mike Powers: Mm-hmm. Andrew Zukoski: Most... Selling a building is probably the most comp- Like, you think enterprise tech sales has it hard? Try selling a fucking building, like... the, like, I was thinking- Like, building a hospital is like the most complex thing that humans do. I'm trying to, like... ⁓ ⁓ Mike Powers: Spaceships, I would put... Luigi La Corte: We need to give every estimating team a gong. We need to give every estimating team a sales gong and they need to be as hyped up as when an enterprise salesperson makes an enterprise. Like enterprise sales, like I have this skew, let me sell this skew. is like no one knows what the hell they want and I have to make this thing work and I have no budget. Andrew Zukoski: Yeah! Yeah. Mike Powers: You guys understand the next data centers that are coming out, they're liquid cooling the GPUs by dipping the GPUs in liquid. Luigi La Corte: No, tell us more. Andrew Zukoski: Is it oil or are doing... Yeah. Mike Powers: Like that's the technology that's coming out. So now chiller manufacturers are stressed out because now you might not actually need a full liquid chiller, water-cooled chiller, which is currently used to cool these data center racks. But now it's a self-sustained rack of GPUs that has a self-sustained loop that it's like the engineering behind that is insane. I'd ⁓ say GPUs, ⁓ building building construction are the three in that order are the most technical. Andrew Zukoski: I think complex construction is more complex than spaceships. ⁓ Mike Powers: The G, spaceships, the zero gravity, the, like to be able to go from a, change, my pushback would be, cause I'm a mechanical engineer, would be the, ⁓ so a civil engineer, it's hardest class is statics. ⁓ is like things like bridges. That's the easiest class for mechanical engineering. And that would be my pushback. That would be my pushback. ⁓ Luigi La Corte: Whoa, I must have left it. This is pure shade. This is pure shade. We take dynamics too, okay? Andrew Zukoski: Whoa. That's fair, but like, like, yeah, the rockets on fire, like buildings you plan around fire, like buildings change shape. It's somewhat less, but you have thermal compression and expansion. Like buildings are taller on hot days. Like, ⁓ buildings get wet. Like vacuum to air. Like that's easy. Who gives a shit? Salt water, salt water. Mike Powers: ⁓ as an optional. Luigi La Corte: If you want to go into structural. Yeah, these buildings move, Mike, okay? Mike Powers: Yes, you need a saltwater spray. Make sure your engineer in the specification in Boston, put saltwater spray on your rooftop units for your Florida project. That is is ⁓ a, that is a spec, a scope gap that happens all the time, but Luigi, Luigi picks it up. Shots. ⁓ Luigi La Corte: That's great. Yeah, non-corrosive rebar. That's right. This provision, shut other provision and I'll make it. All right. We're going to, we're going to leave it there. Thanks for joining everyone and we'll talk next week. Andrew Zukoski: next is. Luigi La Corte: Massive shit.