Latest / Construction Technology Podcast / Groundbreak, Digital Coworkers & Nailing the Core: Procore Product Strategy (with Geoff Lewis)
Transcript
- Luigi La Corte: Hello everyone, welcome to the construction technology podcast. I'm Luigi from Provision. I'm joined my two co hosts, Mike from Build Vision and Andrew from Join. We have a very, very special guest. We have Jeff Lewis from Procore, SVP of Product. Jeff, how are you? Good morning. Geoff Lewis: Hey, doing well. Thank you guys for having me. Luigi La Corte: Yeah, thank you so much for coming on. Obviously you just wrapped up groundbreak. We know you have no shortage of things to do, so we appreciate you spending the time. Let's go. Huge hat. Amazing. Yes, I didn't Geoff Lewis: That's right. Luigi La Corte: actually stop by the hat booth, but I they was so busy. everyone was getting a a Procore hat. Geoff Lewis: yeah. The first year I went to Groundbreak, and I've been at Procord for seven years now, so this was a while ago. I was like, the hat bar. That that thing's probably not too popular. It it's probably the single most popular thing we do at Groundbreak. so I had to s I had to scrounge for an orange hat at the end. But yeah, it it's it it was a lot of fun. Luigi La Corte: Yeah, it looked really, really fun. So we we appreciate you joining. Maybe you can just give a quick intro on yourself for the audience that isn't familiar with yourself and maybe they've been living under a rock and don't know Procore, but maybe just give us the, thirty thousand feet view of you and the company. Geoff Lewis: Yeah, absolutely. So Procore, I would say market leader in construction management software. we have close to 75% of GCs in the US leveraging Procore today. And I'll I'll talk about this a bit more later, but we have a ton of growth opportunities in owners and subs. So really excited about what we're doing. I've spent a lot of time with at least Mike and Andrew, Luigi. Nice to nice to finally meet you. Luigi La Corte: It's meeting too. Geoff Lewis: I'm SVP of product at Procore, so I oversee our our product. Portfolio. I've been at Procore seven years, as I mentioned. and before that, I was the chief product officer at Honest Buildings, which was a smaller startup that Procore acquired. So there we went from zero to $10 million in ARR. And so, you know, basically had to wear every witch hat over those six years. And so anyway, love talking with guys like Mike and Andrew about kind of like the earlier startup experience, and then it's been a lot of fun at Procore. When I joined Procore, we were 400 million in revenue and And we're just about 1.6 billion now. So the numbers get pretty gargantuan, but yeah, that's that's that's me in a nutshell. That's right. That's right. Yes. Yeah, thank you. Thank you, Mike. Please I'll I'll Mike Powers: So so you might say you're responsible for Forex growth. Specifically you and your Yeah, cool. Andrew Zukoski: It's good. It's good. Yeah. Geoff Lewis: I'll be sure to bring you in right before my next performance review Andrew Zukoski: Yeah. Geoff Lewis: commentary. Luigi La Corte: So I just came back from groundbreak, like I was mentioning before we started recording and I was there for all the product keynotes and the AI keynotes. And I remember texting the the CTP texting group afterwards and said, like, hey, I'm like I'm bullish. I really see the view I really see the vision where we have ground and air from drone deploy and AI connectivity with data grid. I could see I could see the story. I think the procurement log stuff was very impressive and it's very nuanced and very hard to do and do right. Obviously, you're biased given you know we're presenting and you went through a lot of the highlights. What was your view of the event and how do you think the audience should digest what was presented? Geoff Lewis: Well, thank you for thank you for coming and the kind words. I I I tend to agree. I I think we we huddled up afterwards and we felt like the the feedback from folks was very positive, right? Whether investors or customers or our employees, partners. so that obviously makes us feel really good and validated. I I think a couple things are working. I think one, as you mentioned, the acquisition of drone deploy alongside Datagrid, alongside Procore, I think is working really well. And I think people really get the see, understand, act vision. And I think that's also something that we've been really excited about, right? I mean, like a lot of construction happens on site in the physical world. And so being able to partner up with drone deploy and think about reality capture and drones and, you know, what you could even do in robotics in the future. That's just super exciting. And I think, you know, we'll we'll get into this in a bit with AI. But I think that is a really, really important aspect of AI in construction is at the end of the day, all this becomes physical. And so understanding reality in the world I think is really important. So that was cool to see people really lean into see understand acts. I know the drone deploy team just was their booth was flooded with people. They got a ton of leads. There's a lot of excitement around what they're doing. So that was thing one. And then thing two As a product person, I basically do the keynote and then I I do about 40 meetings with different customers throughout that time. And you guys know construction customers. They're they're pretty good about telling you when they're not happy, right? Or if there's something they need you to go fix. And so, like I said, I've been to a bunch of ground breaks. And I would say over over past groundbreaks, you get a lot of feedback of like, hey, the cool new thing you announced, that's great. Can you please make sure your core product is awesome? Right. That's what we pay you for. We're also interested in these. ne products but make sure you nail the core. And I think in years past, we've We we've had a lot of different things going on, right? We've had a lot of surface area and it's been hard to put the appropriate amount of investment into all those different things. I think Luigi La Corte: Mm. Geoff Lewis: for a bunch of reasons, both because our strategy has been very much about getting the core right, but also because of we get to use AI to develop software much more rapidly, right? We I think did a really good job this year of both delivering on basics and doing things that I think sort of are beyond just what customers are asking for, right? So you know, I think you guys heard when I made the The joke in my keynote that we don't earn the right to talk about innovation like digital coworkers until we nail the basics and like submittal workflows. People were like standing up applauding when I told them we could bulk edit the submittal workflows, right? So I think that's that shows where I think people are at, right? They still Andrew Zukoski: Yeah. Geoff Lewis: really want the core of Procore to be amazing. We owe it to them to do that. And people aren't gonna buy our add-on products if our core products aren't amazing, right? And obviously, people's standards for software is continuing to go up. And so we have a real mandate to to go make that make that good. And so I feel like this year we did a good job of nailing both. one other thing I would add is James, who is the CPO at Drone Deploy, he texted me and said, when when I mentioned the submittal workflow bulk editing thing, a customer was like, F yeah, right. And then a few minutes later, that same customer when we showed the meta classes connected to the Pro Core AIO app was like, F yeah, right? So I think it's very cool to think that they're excited about both basic stuff that's happening inside the core app as well as some of the exciting new stuff we're doing, like you know, getting Procore onto certified. Andrew Zukoski: I feel like that's always been one of Procor's strengths is that embrace of the industry where it is, like the hat bar that people want to go to, like the core products they're gonna use in the field every day, but then the like commitment and vision to bring, you know, new things beyond that. But Geoff Lewis: Well, thank you. Yeah. I think the thing that I I was talking to somebody on an interview yesterday and they're like, well, hey, you you got acquired. People don't often stay at the company they got acquired by for seven years. Why have you done that? And one of the things I said is we just we we had a very shared vision, I think, between the way honest buildings work and the way ProCore works, which is very customer centric. Like we actually really like talking to our customers and hearing what they say to us. And look, we can't do everything, but we try to do our best to be responsive to that. And I think that makes it that makes it easy to to not screw it up, right? Because if you just listen to your customers consistently over and over, you're gonna be in a good spot because they're gonna guide you to to what they need. so that that's that's something I I feel like Tui in his core is just one of the most customer-centric people. And I think that's in Procore's DNA, and I think that's that's served us well over time. Andrew Zukoski: So that's that's interesting. Procor is a public company now, under pressure to grow, under pressure to enter new markets. it has been said that Procor has largely tapped out the G C market. Like, where the where where's the next billion dollars of revenue gonna come from if not listening to G Cs? Mike Powers: Yeah. Geoff Lewis: Yeah. I I've I've definitely heard it said. In fact, Mike, I was just listening to a pod from you the other week and I heard you say that exact exactly too two. Mike has also said it privately to me, so Mike is Mike is a very trick parent. That's Luigi La Corte: Mike, someone's calling you out on something that you said publicly. How Andrew Zukoski: Wow. it go goes around comes around. Yeah. Ne next you should ask ask ask Mike for public benchmarks on how tapped out general are. Mike Powers: J no n Yeah, I've I've I've consistently said this privately, didn't you? Hey, Procore pay Procore published their benchmarks, unlike some other companies that are in market. So kudos to Procore. Geoff Lewis: That's right. That's right. anyway, I here here here's what I would say. I think I I basically have two two responses to that question. First, I don't think we're tapped out on G C And I think there is another billion dollar opportunity in that market, which I'll describe. And then two, we have massive opportunities in both owners and subs and international. so maybe I'll maybe I'll start with those. you know, so I mentioned ProCore's now $1.6 billion in revenue. About a quarter of our revenue, so several hundred million dollars, is just from our owners business, right? So think data center builders, think retailers, schools, hospitals, utility companies, they're leveraging Procore in a different way than their GCs historically have been. And they're looking at portfolio-wide analytics of where am I trending on budget, where am I trending on schedule. Yes, what's happening with my GCs on my projects, but that's maybe more of an afterthought than the core. so earlier this week we launched a new capital planning product that is built for owners. so that that market has been amazing for us, and I feel like is it is easily a billion-dollar market. subs, same thing. it's a little bit smaller than our owner's business today, but gosh, there's a lot of opportunity in in trade contractors, as I think you guys all know. And then lastly, like, you know, I told you I'm going to Australia in a few weeks. I was just in Dublin a few weeks ago. There's a lot of demand for Procore outside the US. Now, there's also some tricky things about that because the local custom of construction is different in those markets, and we have to be thoughtful about how we actually position and you know make the product tailored to those regimes, but But fundamentally, there's there's a huge market in Europe. There's a ton of construction happening. It's huge market in the Middle East where there's also a ton of construction happening. So I feel like each of those vectors is another billion-dollar opportunity for us. What we've historically struggled with is how do you do that while also maintaining really good growth and customer orientation inside your GC business. And so I think a couple things have happened for us. One, again, AI helps us go much, much faster. So with the same RD team, we can now. Do you know two, three times more things than we could previously, which is very fun and exciting as a product person. But we've also actually organized product people around these markets. So for example, Eliza on my team is the head of product for owners. And so she wakes up every day thinking about owners. She spends all her time with hospitals and data centers and schools, understanding what they need from ProCore. And then she has resources to go build that, right? And some of that leverages what we've built in our platform. what we've built for GCs, but a lot of it is like let's just spin up something new for owners and working backward from what the owner wants. We do the same thing with subs. Brian on my team is our leader for the subs market. We're doing the same thing in Europe. That has been extremely helpful is is just making sure that not everything has to compete against the US general contractor because again that's that's our core business and Andrew Zukoski: resource separately. Geoff Lewis: we're very responsive when when Turner gives us a call, right? But being able to actually carve out resources and make sure we have smart people focused on these different markets has helped helped us a lot. So that's how I'd think about those new markets for us, right? I think we've I think we've been sort Andrew Zukoski: Yeah. Geoff Lewis: of accidentally successful over time and by focusing product people on it, we're able to be very purposeful and get much faster growth in those areas. Andrew Zukoski: I have a quick follow-up on that. as you think about going into owners, like the needs of a hospital system and the needs of a data center are total are are different. and especially in the trades, like the needs of I don't know, a painting contractor are very different from the needs of like someone who has a huge like a steel fabricator or something like that. Like, how do you whereas Geoff Lewis: Yeah. Andrew Zukoski: general contractors, it's a little more the case that like a project is a project is a project, at least in our experience. Like, how do you think about that further? level of verticalization or specialization? Is it to avoid it or manage it or further staff out different, teams and PNs to meet particular verticals needs or Geoff Lewis: Yeah, really good question. And that is definitely something we've noticed, right? what I would say is that it tends to largely not impact the product and very much impact the go-to-market. So what I mean by that is Actually, the way a hospital thinks about managing a set of projects and the way a data center think about managing a set of projects are not super different. and and those are actually good examples because hospital also has a very large amount of high-end installed engineered equipment that they really care about commissioning, right? So does a data center. Now Maybe a, coffee shop chain doesn't care as much about, detailed asset commissioning and test and balance and all the different things that these high-end asset owners would. And so there's you you can imagine like there's different jobs to be done that we need to support inside the owner's product. But I would say industry doesn't tend to be a major differentiator for peop what people need out of the core product, I would say within owners, right? Now obviously GC and owners have quite different product needs, but then within different owner segments, largely it's similar. But if you want to go target these people, People, well, if you want to go get in front of them, you have to go to the retail conference. The data center people are not at the retail conference. Now you go to the data center conference. Now you go to the energy and utilities conference. And so I think where we've seen it largely is you need to have different go-to-market motions around these things. And I think that's also instructive on subs. All of them care about managing labor, equipment, and materials. Different trades though have different intensities of where labor is important versus equipment versus material, right? There's certain trades where equipment is basically not relevant and it's all labor and materials. There's other trades where equipment is is massive. And so being able to sort of say, well, hey, our core for subs is solving for labor, equipment, and materials. Then any given customer can basically think about their intensity of usage, I think makes sense. And then similarly, you have to go to NECA you have to go to SMACNA, you have to go to these different trade shows. And you're not going to find a place where a bunch of electrical and mechanical engine, you know, subcontractors are necessarily congregating. So for us, I feel like it has been more go-to-market oriented than it has product-oriented. You know, so I think that's that's kind of where we're seeing it at least. Mike Powers: So one of the struggles we had at Autodesk with like going up and down the supply chain was like selling to an owner, an owner would provide a bit an instance of BIM 360, and then we would lose revenue with general contractors and subcontractors because those subcontractors and general contractors would just use the licenses for from the owner. how is Procore thinking about that of like conserving GC revenue, obviously selling to owners? Like what does that go to market motion look like? And how how are you positioning in a space where If Meta says, hey, I have a Procore instance and Turner's on a project, does Turner get heartburned by paying for Procore while they're using Meta's instance? Or well, how does that how does that like how do those map pieces work out or what are you seeing and how does that change as you're shifting up and downstream? Geoff Lewis: Yeah, really good question. And we think about this all all the time. so I'd say there's probably three factors there. One, it's largely a benefit because if you go talk to owners or subs and you say, why did you buy Procore? They're basically saying, Well, hey, all the GCs I work with already use it. Right. So it's actually it's a big flywheel for us to get access to owners and subs. And so that's that's the positive. What you then need to do, and and I think this is where we maybe didn't always have it right, but I think we've gotten really crisp over the past few years, is what independent value does each of those firms get from having the having ProCore on that project, right? From my perspective. It probably doesn't make sense for each of them to pay just to access the drawings, right? There's gonna be a central account, it's gonna be ACV based with unlimited users. And in that context, if all you need it, all you need is access to drawings, again, there's probably not enough independent value for all these different folks to purchase ProCore. So if you think about like the product announcements I was talking about for owners, capital planning, financials, budgeting, those are things that aren't shared with the GC and create independent value for the owner across all of their different products. projects, right? So now it's less about like, well, where are we storing the drawings on this one project? And more about, well, what's my system as an owner to run all of my capital projects? And then if you think about the sub, sure, they might get all the drawings from Turner's account, but where are they going to manage labor? Where are they going to manage equipment? Where are they going to manage materials, which by the way, tend not to be necessarily as project specific as they are company wide, right? Well, cool, I need to move labor from this project to that project. I need to move equipment from this project to that project. And same thing with safety. That's another area where, hey, no matter what I'm doing in Turner's account, if I'm using ProCore as a subcontractor to manage my safety program, that's hooked into my insurance, et cetera. Right. So I think. thinking through what creates independent value has been really important for us. And I think we sometimes struggled there, especially in a world where project management was really our only product. And then the last thing is Pro Core Connect is massive, right? So just being able to have a sub, create an RFI from their account, send it over to the GC's account without ever having to log into the GC's account, that is a huge time saver and then makes it significantly easier for people to think about my Procore versus their Procore versus this Pro Core. We want to get people out of the question of whose pro core should we use because ultimately I think that allows us to monetize across all the stakeholders. But again, you can't just be monetizing for the same thing without independent differentiated value. So that that's how that's how we've been trying to think about it. Mike Powers: So on on the connect piece, is that a band-aid until it is one infrastructure or is that something that will continuously Geoff Lewis: Ooh. Mike Powers: be used? Geoff Lewis: Okay, so I was originally a maybe shared your opinion, which is like Actually, what you need is one beautiful canonical version of the project and like that's what you need. And then everyone just needs their own permissions sliced to it. And then we started talking to customers and they were like, we don't want that at all. We hate that. And here here's a good example. We would we would talk to subs and be like, hey, well, you know, don't you just want when the with the GC marks the RFI officially responded to, don't you want to just have that update your account automatically? And they were like, No, absolutely not. Half the time we don't agree with the GC's closing of that response. We want to have our own separate records. And so I think the way we have it, which is you have an independent record for the sub, you have an independent record for the GC. There's connectivity between those, but each party controls their own view and sort of thinking the data and the metadata around the data. I we found that's actually the preferred model for the industry, and I think mirrors how people actually think in the world. Mike Powers: So in that example, do you see that there is a database with ABC towers for the GC and a database with ABC Towers, the files, the submittals that are always going to be separate or disparate, or will Procore merge those into one canonical record? Or do you like because I think like your position of like, hey, like I don't want the GC Geoff Lewis: Yeah. Mike Powers: being able to change a status makes complete sense, right? But that's like a workflow status. But like having a canonical set of documents, which I know we've talked Geoff Lewis: Yeah. Mike Powers: way too much about with Geoff Lewis: And and to be quite clear, for example, we let's say we store a document in an S3 bucket, right? We'll then have a link out to both such that both parties have it. The other thing you'll see is, hey, what happens if a connection gets severed, right? So hey, that subcontractor is no longer on your job. Well, if they had those drawings stored in their account, we then copy that and now it lives with the sub. No further changes to those documents pushed down to the sub's account. But it it's actually like one of the weirdest, hardest problems to solve because you go ask customer how it ought to work and actually this has never existed before right everyone has just been emailing things back and forth for you know I guess since you know since prehistoric times but that I think so so we've really had to figure out right in this world that we've created where now a sub and an owner and a GC can all participate and work out of their own account how exactly ought it work and it's way more complicated and messy than I I think I had envisioned in in the first instance. Luigi La Corte: Let's change gears for a second. so this Geoff Lewis: Yeah. Luigi La Corte: is this is the path forward, obviously. So the acquisitions help fuel some revenue expansion. Obviously, the product is rock solid. People who use the product say it's the Porsche of the Ferrari of project management systems or project management software. when I hear customer anecdotes about Procore pricing, it's usually I hate it. It's usually like a vehement reaction to. you know, a basis point of construction volume. And you know, that just may select for the the GCs and the the partners that you want to work with, and that's completely fine. but part of the promise I think that we've seen with AI is it should be deflationary, it should make project management software, maybe in a contrived example here, cheaper or less expensive. How do you see That potentially impacting the pro core business model. If there are gonna be, let's say, a flood of competitive softwares that are now easier for third parties to build, do you think you're gonna have to meet competition where they are, potentially charging flat fees or reducing or changing the business the basis point model? Or do you think, you know what, that actually works for us really well? Our customers are actually okay with that. We don't to change anything. How do you react to that given that the market is probably gonna become more competitive? Geoff Lewis: Yeah, that is a really good question and obviously something we think about a lot. with respect to the the competition point, I I think what I would say is AI is deflationary for technology. I think all of technology though has always been deflationary, right? And so I think that when I'm saying, hey, we need to go make sure we're investing in the basics, we're not charging you more for that product. We're investing in that product. You're now getting more. And so from my perspective, the product, you know, you're paying the same price for the product today as you were five, six years ago. And the product is much, much, much better. Right. And I think that is that is the core of how technology works, right? Is that we as a SaaS company have to invest In our existing product, and you can't monetize all of that. Sometimes you're building net new products that people are also gonna take advantage of. But I feel like if a lot of competition comes out in the space, we then need to be very responsive and reactive to that. And part of the reason we have such a big RD investment is to make sure we can do that. And obviously, our just like a startup's RD investment can go much farther now because of AI, ours can too. But we're very reactive to that and we spend a lot of time working with our customers to say, hey, what other technology are you using? What point solutions are you using that sort of rip away at the Procore all-in-one platform value? And where does it behoove us to go shore that up? Where does it behoove us to maybe do an acquisition? Where should we just partner with somebody? But we spend a lot of time with customers trying to figure out where exactly we should be investing. so that that's kind of how I think about it. That's actually, I think, the beauty of SaaS from a customer's perspective is that it forces the software provider to continually improve the product just to get it running. renewal, let alone get additional economics from from from customers. And if you go back to what I was saying earlier, I think some of the places we aired a little bit in recent years was spending too much on add-ons and net new products and maybe not enough Mike Powers: Yeah. Geoff Lewis: in earning the renewal from our existing customers. And I think that, you know, for for a bunch of reasons we've been able to shore that up really nicely. Luigi La Corte: How would you respond to the general contractors who who think that as their construction volume increases, Procore is, you know, maybe in a justified sense or unjustified sense, is gets just getting more money as a result of them growing their business, but not by adding more value. Cause that is a like a very common refrain that we Geoff Lewis: Yeah. Luigi La Corte: hear when we talk to our customers. And now I understand it's like it's a very scalable business model for you and You are providing more incremental services because there's probably a relationship between amount of R fives or submittals or services, but that's still the perception. Are they thinking about it incorrectly? Do they just not understand the value? Like how how would you respond to that? Geoff Lewis: Yeah. I was listening to the to the pod a few weeks ago and I know you mentioned that it it sounded like basically the net net was a lot of customers were coming up to you saying, I love Procore, I'm literally at the Procore conference, and Procore has gotten fifty percent more expensive over the last five years, and that's that's very frustrating. So look, when you look at it, I would say that you you know you can't argue with the fact that customers are paying more for Procore in general than they were five years ago. But let's be really, really specific about what that means. most of that is that their businesses have grown quite A bit during that time, right? So there's a bunch of customers who've been with Pro Core for 10 years, and they'll come to you and they'll say, Well, this is very frustrating to me. When I signed on in 2016, I paid $100,000, but now I pay a million dollars. And they they're angry and sort of frustrated by that. And then you're like, Well, all right, I I hear you. Tell me about how much your business has grown during that time, right? And they're like, Well, you know, we had about 200 million. yeah, now we're at 1.5 billion in ACV going through Procore. and let's talk about your product set. yeah, well. We started with project management, but now we have quality and safety, we also have your BIM products, we also have financials, and we just rolled out pay. And it's so so you think about that and it's like, hey, is the customer paying a lot more? Absolutely, yes. Is the customer also getting tremendously more value from our perspective? I think also, yes, right. And I think the reality is We, as a software provider, leverage a bunch of softwares that we use, you know, Databricks, Claude, et cetera. I am not happy with the price I pay for any of those, but I'm also extremely happy with the value, and they're they're, I think, fundamentally very important to how we run our business. the other thing is we've explored a lot of different alternatives besides ACV, right? Because you get into these conversations, as you guys know, and customers say, ACV is just a terrible model. And then we'll we'll we'll actually in good faith talk to them about like, well, what alternative Alternatives could we use? All right, well, let's try users. And they're like, well, no, that's not right because we we really value the collaboration. Okay, well then let's Luigi La Corte: Mm. Geoff Lewis: do it outcome-based. We'll just do it based on number of RFIs and submittals managed. Luigi La Corte: Mm-hmm. Geoff Lewis: Well, no, that's not right because I don't control the number of RFIs. you know, that's based on the architect, right? Okay, well let's try. So it's so actually as you go around, you tend to come back to ACV, is sort of like the democracy pricing models where it is maybe not a perfect system. that it's the best of all the systems I think we've we've found to date. and and then the last thing I would just say is look, we we are a SaaS business. And so fundamentally customers have to renew on our product. We We every one, two, three years, however long their contract is, they have an opportunity to leave and go somewhere else. Obviously, if they do that, that hurts hurts our business. Like we are very invested in making sure customers renew with us, and that means we have to add enough value to the software. That means they have to actually be using the software in order to renew on it. We have to have the right pricing. We have to understand the competitive dynamics, and we have to have a good relationship with that customer. Obviously we've done a pretty good job of that, but that that's kind of how I'd think about the pricing model in in general. Luigi La Corte: Pass it on to Mike and Andrew who have some questions too. But one thing that I haven't seen discussed too much post groundbreak was the rework calculation announcement. And for Mike and Andrew, I don't know if I was texting this to you while Andrew Zukoski: Gracias. Luigi La Corte: it was being presented, but Procore presented a way to actually get a number for rework that's suspected on a project. and connected to rework, build libraries of lessons learned. Like if you've spoken to GCs, they always talk about these kind of, you know, kind of compendiums lessons learned that they just have this. That sit in Excel sheet. Wouldn't it be great if we could just bring that into your project and highlight when those lessons actually apply? Great for stickiness, it felt like, project to project. Because there's always been a question of, you know, can the GC just move to a different system at the next project? But if you have all these calculations of rework and actually put a number to it with the specific situation, it could be pretty powerful. So I thought that was I thought that was really cool. Mike, Andrew, what what's top of mind for Jeff? Mike Powers: I think one of the we cut it out a little bit earlier is just and tie into the AI piece is the the digital cowork offering that y'all are offering now. instead of kind of individual agents, would love to kind of hear like, you know, what was the thought process? I'm sure there was thousands of conversations internally about how to position the product. how should customers think about that and like what they're paying for and kind of the value around this idea of a pack as opposed to, hey, here's an MCP or here's an individual agent to do work on. Geoff Lewis: I think we had thousands of internal conversations and thousands of customer conversations about this. I I would also say that we're still in in flux, right? I think we're in flux with every other provider of AI in in the world, including the the Frontier Labs on how it should work over time. So I definitely do not think we have the perfect answer, but I can share how we've been thinking about it. In general, I would say that the push from customers is for their users who use Procore today to use AI through Procore tomorrow. That's typically what our customers are asking us for, right? And I think Andrew Zukoski: Why? Geoff Lewis: Well, I think that that comes back to the structure of the firm, right? Which is, hey, there's executives at that firm, there's IT people at that firm, there's analyst people at that firm. Th those people might be Claude or or you know Codex may be the appropriate interface for for them, right? But there's also a rank and file of folks think I was talking to Brassfield and Gory, and they say they hire several hundred recent grads from Auburn every single year, right? And I think what we've heard from them is we don't want you to show up And then have everything be probabilistically, generatively decided for the project engineer. We want it to be much more deterministically decided. What process? Right? Exactly. Andrew Zukoski: That that determinism is already where ProCore shines and you already have all the permissions and it's not one more thing people have to log into. Okay, that makes a lot of sense. Geoff Lewis: think think about it it's a fundamentally an SOP driven business, right? And so that I think that idea is at the core of why people want again and we can look at different users within the the the firm, but that's why they want at least their mainline users who are using Procore today to can to actually access AI through Procore, right? Now Andrew Zukoski: Yeah. Geoff Lewis: what I would say is if you think about the submittal workflow from twenty twenty five, that's Procore is basically moving information back and forth and then ultimately All the knowledge work of evaluating whether that submittal submission is compliant with the spec is left to the project engineer to actually go open that document up, do that work, and then respond in Procore, right? So what the customer is most interested in is, well, hey, can you take that knowledge work that is happening at that point in the workflow? And instead of the project engineer having to spend their valuable time doing that, can you have your agent embedded at that time, wake up as soon as a submit a submittal is submitted? Check it against the spec, check it against the drawings, and draft a response for the project engineer to reveal, right? Well, that still is deterministic from an overall workflow perspective, even though that particular step is probabilistic in in in terms of it doing the knowledge work for the PE to reveal, right? And so our view is like that's what people want fundamentally, that's what customers want fundamentally, is they still want deterministic workflows for some of these core processes, right? and you know, I think as I've heard you guys debate this endlessly on recent pods. I feel like that's probably aligned with the Andrew and Luigi view, which is like UIs still matter and maybe maybe not not Mike Powers: Ha ha ha. Geoff Lewis: exactly with with Mike's view of Claude, Claude will rule the world. But but now let me talk about MCP because I do think there still is a place for that as well. There's a ton of executives and, you know, sometimes senior safety managers, financial analyst folks, CFOs, who, even if their teams are in Procore, they are not in Procore. And so in that context, I think a MCP is hugely helpful because they can now get access to information coming out of Procore that allows them to make better decisions without having to go in Procore. And we're of course designing Procore to be more executive friendly, but I think we also need to face the reality that most of the work done in Procore is done by the rank and file and the executives don't spend a ton of time in there. So we, you know, I think you guys heard that we did an Announce a Claude MCP at groundbreak. And the purpose of that is actually to allow for folks to get data and reporting and analytics out of Procore in order to solve for that use case. So I think we're we're not, we don't, we don't really have any blanket views on it. We're just trying to be thoughtful and use case specific about how to think about it. So Netnet, I would say, for existing kind of users of Procore, largely what we're hearing from customers is instrument AI into their workflow effectively in an invisible way. gives them more power and makes them more effective at their jobs. But for executives, it's people who wouldn't otherwise access Proport. Let's make sure we're really smart about how we get that data out. Mike Powers: So why not why why a hundred names and why not just one agent? Geoff Lewis: w one, I think technically under the hood, you'd probably I think I think you said recently, hey, this why is why is it not just one agent with a bunch of skills it accesses? I I you know, fundamentally that's that's how it actually works under under the hood. Mike Powers: just I so as Andrew Zukoski: It's 'cause you it's 'cause you permission it differently. Mike Powers: someone from building connect but all right but as I I I I as a building connected old school person we had this Geoff Lewis: Yeah. Mike Powers: promise that bid leveling was going to be like solved by leveling out bid forms in the application. So you guys have a bid leveling agent Geoff Lewis: Yep. Mike Powers: and I'm in Procore Pre-con, like and I'm in the bid forms, why would I even need to contextually say I need to go to bid leveling agent when I would just use the agent and already have the context of bid leveling in that example. Geoff Lewis: I think what you just Mike Powers: is my th is my thought and I'm curious, 'cause I saw in the bo I just looked at the video and it's like, Hey, this is the bid leveling agent, and an Geoff Lewis: Yeah. So two two two two two things. A r really good point. I think on the first instance, there is still the empty prompt box problem that people have with AI. And that is true both for users. That's also true for buyers of AI. It's like, well, what what does it do for me? Right. And so one of the reasons we've packaged up the agents in that way, you've got the bid leveling agent, you've got the budget analysis agent, you've got the daily log agent, is is actually just to help people understand the different jobs to be done that AI can do. Now, all right. What what is that actually under the hood? It's sort of this like you know reasoning engine that's calling different skills and different MCP actions in Procore. I don't think you need to package it up as an agent from a product, like actually how the product works perspective, but it is helpful to go communicate to customers what they're trying to do. Luigi La Corte: Sure. Mark from a marketing perspective. Yeah. You're informing them Geoff Lewis: A marketing and an internal adoption perspective at their company, right? So if they if you give everyone Claude, it's like all right, well what do I do with this thing, right? I is is both the reality and some of the concern. When a customer is like, I'm using ProQuare AI for two things. I'm using it to check submittals against the spec, and I'm using it to check new RFIs to see if they've already been answered. Cool. Now I can go tell all the rank and file about that. And it's un it's intelligible, it's understandable. And so like it it that is another value piece of it. Now Luigi La Corte: Sure. Geoff Lewis: is that limiting? Yes, but you're sort of constraining the infinite choices that are available in order to sort of increase clarity and and clarity thought. And then to your point, Mike. We're in the process of embedding everything deeply into the workflow because you're exactly right. you shouldn't have to be on bid leveling, click an agent, and be like, hey, level this for me. You should just show up. It should have pre-leveled all the bids. And so that's that's what we're working on to do. And so like we're farthest along on things like submittals and RFIs, which are kind of the you know most leveraged capabilities in Procore, and we're working to bring that to everything. But the the vision state would be exactly what you described. I go to Procore, the bids have been submitted, I show up as a Mike Powers: Yeah. Geoff Lewis: you know, pre-con manager, that is already analyzed for me. And now I can have a multi-term conversation with AI about like, all right, is this scope truly leveled out on you know the third bidder? Like help help me walk through that. So that I think what you described is exactly the vision. And it just takes takes a little bit of time for us to get there there fully. Luigi La Corte: And it's it's coworker packs, right? So you have these like five agents that you can purchase and you'd purchase more agents if you wanted to expand that pack. Is that right? Geoff Lewis: Yeah. So the way it works is there's basically three tiers. There's a starter pack, which is those five agents you were you're talking about. And basically it is AI powered intelligent search, which is basically, hey, across all the documents you have in Procore, pull up answers that are extremely detailed. It's daily logs, submittals, RFIs, and contract review. Right. So it's like, hey, we've taken all the different things you could do, we've bucketed down into the five most important things for you. And that one is sold as a unlimited price basic basically right so you pay a price there's no token caps and that was really important because there's a lot of friction from customers around figuring out tokens, figuring out credits. And so Andrew Zukoski: They can't predict it. Geoff Lewis: by just saying, hey, just pay a price, Luigi La Corte: I wonder. Geoff Lewis: you could use it as much as you want. You know, we're not, we're not concerned about that. That I think has been really helpful for people to get into it and and understand how to how to basically give it give it a shot without figuring out how to budget for it. Luigi La Corte: I wonder about that though. This might be a stated versus revealed preference thing though, because if you go look at Anthropics S1, 80% of the revenue is not coming from subscriptions, it's coming from consumption. And they're working already with some of the largest companies, all Fortune 500 companies, a good percentage of them. So I mean, kind of my critique, and I wonder if this is a valid critique. If you are purchasing packs, right? Agents one through five, but I am going to the single text box and I need to go use agent six because I have not purchased the next pack. My experience is being impacted. Geoff Lewis: Yeah. Luigi La Corte: Well I'm sure you I'm sure you you thought about the trade-off between, okay, let's just go flat fee and then consumption on top, which clearly people are doing. we do that within provision. I'm sure you do that within Procore too. How why did you justify the decision to cut off the value from the sixth agent? Given that you didn't think they would want to go to consumption. how did you make that trade off? 'Cause now in that situation they're like, damn Geoff Lewis: Yeah, yeah, yeah. Luigi La Corte: it, I don't have that six agent, I don't have the let's call it, the the work breakdown structure agent. Crap, where do I go now? Now I need to go find another provider. Geoff Lewis: That is a very sharp question. and I I basically what I would say is we looked at a couple things. One, where could we make sure that every agent we delivered to you was gonna be amazing, right? Because that that's the other thing too. It's like that sixth agent, that twelfth agent, that you know, eighty-fifth agent. If those aren't good, is that actually helping your experience? Like one thing is we wanted to just constrain the number so we could make sure the quality was high. And obviously we're we're focused on improving the the quality of all these things. Two, there was this feedback Because we would show customers both the enterprise version and the starter pack version. And a lot of times they'd be like, I really just want to start with these two things. and again, is that stated versus revealed press preference? You would definitely have customers, Luigi, who would then be in there. An end user who maybe wasn't a part of the buying process was like, Well, I want it for this sixth use case and this seventh use case and this eighth use case. So we're definitely considering, okay, should we actually give you more in the starter pack as a way to you know get more adoption, get more usage? So it's it's a live sort of trade. trade-off conversation, but those were some of the things we we thought about was, hey, we can narrow it down, make it really digestible. We think that's gonna be the best thing, but we we may look at at expanding it out over time. Mike Powers: Awesome. And just as we think about without sharing formal pricing, I don't think you guys have posted. Is this the expectation for customers? Is it like, hey, we're starting to use PAX? Is my broker contract gonna 10x or is my broker contract gonna increase 10 at a 10% if I'm like leveraging PACs across the portfolio? Like how are how are you positioning that to customers? it's like selfishly, like our Geoff Lewis: Yeah. Mike Powers: anthropic bill continues to ramp up and they now switch me back to enterprise and I wish I was on user because I'm spending too much money. I'm curious how you're thinking about that in a in a world of very more of more constrained costs than a startup. Geoff Lewis: Yeah. So starter pack is is much more, you know, in the ten percent versus the ten X, I would say. I think this also bridges back to the point you were making, Luigi, about eighty percent of the revenue of anthropic being in the enterprise. you know, Procore is a major driver of that revenue for for them. And I think that we wouldn't have drew if you had asked me or Steve Davis, you know, who runs product and tech, hey, do guys want to pay, you know, millions of dollars to anthropic via tokens, we would have said absolutely not, right? That that doesn't make any sense. However, there were so many subscriptions and end users who had $20 or $100 a month plans who were all like Claude is amazing, that it then made sense to say, okay, let's get a API license to Claude. And then of course that grows through usage. So I think the what I would say is like starter pack for us is probably our version of their subscription, right? Because I think subscription is the gateway for a lot of people at Anthropic to get started and then get onto enterprise where you're now in sort of uncapped consumption world, right? But I think if you go ask customers to start on uncapped consumption world, it's a struggle to to get them there. Right. So that that's one of the things I think we've been using starter pack for. And then I think the Mike Powers: Are you just just on that I was just gonna say on that? Do you expect are you guys in the position now where if full if someone's like ripping it and using on every single workflow, is it cost beneficial for you all yet? Or is it the expectation that it'll be cost beneficial in in a year? 'Cause like you have customers that are spending a hundred thousand dollars on, you know, tokens because they're using every agent possible and you're only charging them ten, or is it more balanced? Geoff Lewis: Well, at least with starter packs, it's it's it's balanced, right? And and sometimes you're honestly trying to get them to drive more adoption, right? Where it's like, yeah, hey, the margin on this is amazing, but we actually need them to drive more adoptions so that they upgrade to the you know, the the the long lasting pro version and get out of starter. And then with the way we do pricing for pro and enterprise, it is fundamentally credit-based. And so we're always we're always matched on our cost, which is what the customer is using. So then that goes back to your point of like, well, hey, could Pro Core AI be 10x your Procore bill. Well, it depends on what you do with it, right? If you have, you know, let's let's think about the maximalist version of that. If you have a drone fly over your site every hour, right? And you have a thousand, you know, a robot dog walk around and take every photo and you have a site camera and you want every frame of that analyzed every second, it that fundamentally could be 10x your Procore bill. I would also question I would also question whether there's real value in all of that today, right? But hey, that that daily drone flight. Andrew Zukoski: That's like a that's like a staffing. Yeah. That'd be like staffing it with a hundred project engineers. yeah. Geoff Lewis: I exactly. So I think that that is where I would think about it. The nice thing though is that it allows you to get some key workflows solved for. Definitely on a workflow by workflow basis, you can say, well, hey, this project engineer used to take a couple hours to review this submittal, now that now it takes a few minutes on average. Great. That is very clear time savings. Let's you know, let's look at the token spend of that. It's a dollar, great. Like we're in a massively ROI positive position. but Luigi La Corte: Mm. Geoff Lewis: because it is uncashed. And you know, customers are only gonna continue to pay and use what they find value in. And so if they did run all those drone flights and site cameras things, they'd have to get some Mike Powers: Ha ha ha. Geoff Lewis: extreme value out of it to keep doing it, I would say. Luigi La Corte: D do you think that it like capping the like I know you said for your digital coworkers it's gonna be a hard number and it gives you X mini users or X mini agents, right? Do you think that creates an incentive for Procore to somewhat limit usage as opposed to expand as much usage as possible given that you have the upside of a of a would-be consumption model. for example, okay, you're gonna spend $100,000 for five coworkers, right? you have a 90% margin on the token spend, assuming some kind of average spend. but what AI is really good at is like augmenting the workflow and in some cases supplementing, you know, bodies if they do not have a project engineer or three project engineers per site. So, you know, they really hammer the API calls and now you have a, you know, a a 60% or a 50% margin doesn't look good on the books. So I mean I don't I wonder how you would respond to that. and I'll Geoff Lewis: Mm. Luigi La Corte: I'll give you an option to to respond to it obviously. But the obvious the alternative being, you know what, let them run free and we're just gonna retroactively bill it. How how how do you think about the long-term incentives of Pro Core if you if you cap the cap the number? Geoff Lewis: Yeah, and just to be clear, the way we have it set up is the starter pack is the one with the caps, right? The Luigi La Corte: Mm. Geoff Lewis: pro and the enterprise are the ones that are then on on credit model. The the Luigi La Corte: I see. Okay. Geoff Lewis: starter though is a nice introduction to it, and we're taking any cogs risk on that because it just makes the the conversation easy for the customer. Hey, do I want to pay 10, 20 grand to try this thing? Yes. Okay, great. And then we manage the cogs risk. Could we be upside down for a period of time? Possibly. If we're upside down, probably they're getting a lot of value out of it. And then we'll convert to the the pro package and that's where they have the that that's where it starts to be consumption based. So just to be clear, we we agree with you. Starter is really a thing we are doing in order to get people over the hurdle of having to think about tokens in in the first place, right? Basically what we were discovering was people are pretty interested in this stuff. Then you inject this token conversation and they become very concerned Mike Powers: Mm-hmm. Geoff Lewis: and on the sidelines. And so our view was: well, okay, let's take that off the table. Let's get people started on this thing, prove value, and then you could bring that token model in. But at that point, you're you're already sold on the value, and now it's just a commercial conversation versus bringing this weird, Mike Powers: Yeah. Geoff Lewis: bizarre new framework into play before they've even gotten any value. Luigi La Corte: Got it. Andrew Zukoski: All right, I got a random question. If you had twenty million Geoff Lewis: Hit me. Luigi La Corte: Got it. Andrew Zukoski: dollars to start a construction technology company tomorrow and it couldn't compete with Procore, what would you build? Geoff Lewis: twenty million dollars is enough. But Mike Powers: Mm. Luigi La Corte: Is he self financing? Mike Powers: No one here is raised up Andrew Zukoski: You can use all of that for marketing for future fundraising. That isn't acceptable. Geoff Lewis: Yeah. That's right. Mike Powers: None of none of us have raised over twenty in this out of these three, right? And review? Yeah, all right. All right. All right, never mind. Geoff Lewis: So I I would say I really like the physical world and I think that's where a lot of the unlocks for construction will be. So think, we we obviously talked about drum deploy earlier and wanting to be more and more embedded in physical reality. I think robots, the you know, it will be an important technology and construction. I I'd be very hard pressed to tell you what time frame that will be, right? I feel like it's it it reminds me a little bit of BIM, which is next year is gonna be the year of BIM and and Andrew Zukoski: Don't make that comparison. Don't make that comparison. Yeah. Geoff Lewis: It it always grows a half percent. but I think you know, I think eventually there will be a, you know, chat G PT style moment for robots, right? And it will become much more reasonable to use them on site. Now, when is that gonna happen? There's real physical world safety concerns about these things. I I don't know. But to me that's that's the space that I would wanna go play in and and figure figure something out. And some of these world AI, physical AI model companies raise massively more than twenty million just because of some of the the extremely hard tech problems. So I think that would be the place I would want to go go focus on. Luigi La Corte: What company that's done something similar right now are you looking at personally that you think is very interesting? There's there's a whole slew of these physical AI companies for construction, but w wha which one specifically has caught your eye and why? Geoff Lewis: I I wouldn't say any particular one has has stood out to me. I think they're all really interesting and I think they're all taking different approaches to it. I think the one thing that is really hard about physical AI and robotics in particular is that you have to get really good at a specific use case. And then any one of those use cases is actually not that big in aggregate, right? So think like Earth moving. Well, that that's a pretty big use case, right? And and I can imagine that being, you know, entirely automated over some period of time, but it's gonna be really hard to then re-leverage that for you know drywall. finishes or something, right? you just have all of these different tasks that are coming together, physical tasks. And so I think it's the thing that I'm interested in is how do you get generalizability across that? Because that is that would be one of the biggest markets ever, right? But when you start to have to go very specific, because that's just the nature of where you know physical robotics is today, it then constrains the market quite quite a bit. But these are still extremely expensive problems to solve. So so one of the things I'm I'm just thinking about is how do you get the economics to to work for some of these physical economics problems, or is there some sort of, like I said, ChatGPT moment that sort of makes the cost of getting some of these robotics off the ground and you can then Luigi La Corte: Mm-hmm. Geoff Lewis: solve all these use cases much more cost effectively. That's that that's what I'm sort of watching and I'm interested in obviously we haven't cracked the code on that yet. Andrew Zukoski: Mm. Luigi La Corte: Would you dabble into AI native general contractor or subcontractor? Obviously, people w outside of construction are buying these services businesses and really infusing AI, just a whole industry now or cottage industry around roll-ups and making these service businesses more efficient. Would you go into that space? Why or why not? Geoff Lewis: I think you can make a lot of money from doing these rollups. A number of our customers, especially in subcontractor worlds, are part of these private equity-backed rollups. I think there's a lot of value to those kinds of things. And I think there's one of the problems that has always existed for construction is that it's been such a local business that construction has not therefore had the sort of requisite scale to invest meaningfully in technology. Right. And so I think as contractors Get bigger and as trades get bigger, I think that's actually a boon because customers like Turner can invest a lot more in technology because they have a lot more scale to it. Someone like Quanta, who's a major MEP contractor, can invest a lot more in technology than a small local version of that. So I I'm pretty bullish on what that could do for construction economy generally, because more vertically integrated operating entities, I think, can ultimately score. Squeeze out more efficiencies in construction. So I I I like that fundamentally, and I think the cheaper, more efficient construction gets, obviously, we're all beneficiaries of that. A lot of what Procore is doing is trying to support those kinds of entities as as they Luigi La Corte: Mm. Geoff Lewis: grow. So I'm I'm pretty bullish on that. Now, the AI native aspect of that, I think that You can't in in five years, you will not be able to be a good general contractor or specialty contractor if you're not leveraging AI. And so I think you could maybe use it as a way to get a jump on others, right? Hey, I'm gonna go out and and become really good at AI. I think that could give you a multi-year lead on your competitors. but you know, at the end of the day, if I'm a AI native general contractor, I I think it's gonna be hard to outcompete Turner over the next few years. And I think they're gonna become the AI native contractor. Faster than you can become Turner is what I would say. So I'm I'm not like bearish on it. And I think as a way to use it as a kind of near-term growth lever, I I think it makes some sense. but fundamentally I I think that it it it reminds me of well, hey, I'm gonna be an internet driven accounting firm. It's like, well, yeah, like I think that that's gonna make you more efficient than using, you know, not the internet, but I also think that over time every accounting firm is gonna use, you know, email and and and spreadsheets and all these things. So it it Mike Powers: You'd be surprised. I don't know. They might not. They might still fax that today, even a day even today. Luigi La Corte: Ha ha ha. Andrew Zukoski: Our customers still pay us in paper checks. I don't know about you guys, Geoff Lewis: Exactly. And I and I think that goes to well, hey, how do you generally outcompete people? You you you run your business better. And so it's for for me, being on AI is critical. Like if if we if if we as a software company were not using gargantuan amounts of AI to power our software development stack, we would be not good as a company and you should short pro core. Thankfully that is not us at all. but that I think is just really important. And I think that there's an opportunity right now to be out in front of everyone else and use that to your advantage. I just don't know that that's like a lasting sustainable competitive advantage in what is fundamentally, you know, effectively a a services business. Luigi La Corte: Maybe last technical question before Mike does rapid fire if you're open to it. the magpie model. Obviously, like a a Jay is you know, marketing Procore now and and actually delivering on it, becoming an AI native company. So, you know, part of that you have models, you have researchers. Magpie three wasn't was unveiled at groundbreak. based on the benchmark that you showed, it beats I think it was GPT six on internal benchmark that you had made. Can you share any public details? Was this is this a homegrown model? Is this like a system of action model? yeah was this like a fine-tuned quen parameter size and like is there gonna be a kind of a a more scrutinizable model being released to the public so people can do benchmarks with other models? Geoff Lewis: Yeah. Absolutely. I heard you guys on the episode with Josh saying, what a shocker Procore's benchmark suggests that Procore has this model. so I thought that was that was funny and and well discerned. okay, so magpie is so this it's called Magpie three. We have we're currently on Magpie two point five. There was Magpie two point before that, and then Tiago and Datagrid's original reasoning engine was called Magpie. So that's the the series of magpies. We didn't just show up. With Magpie 3 from the start. It was it was an evolution. but what it what it basically is is a reasoning engine for how to make underlying calls to I I basically I think of it as like a harness, right? It's basically like a custom Luigi La Corte: Okay. Geoff Lewis: construction harness that powers Procore. Underneath it, there are some sort of fine-tuned models that do specific things, and then there's also calls out to all the MCP actions in Procore and some general models that that you can leverage for like core-based intelligence. So then what is the benchmark? The benchmark is basically just a golden set of evals that we have, right? And so think about all right, you feed it a submittal, you feed it a spec, and there's construction pro has figured out what the right response to that submittal is. Is it compliant with the spec, right? Well, cool. When you feed our that data to magpie running on you know whatever whatever model we're running. How does that perform? Does it get the answer right? Right. And does it present the right facts? Does it answer the question correctly? And then basically, I think we have 250 different questions that all of our construction pros came up with. That's basically what we're saying. It's like, hey, if we're whatever, 80%, I forget what the exact numbers are. That just means that in general, we're quite well tuned to answering those specific questions well and effectively. And that's like our core eval set that we run nightly against our product. We obviously have tons of additional evals for all of our other agents, but that's like the core of it. So that that's how I would sort of think about it. Luigi La Corte: it, very cool. Mike, take us away. Mike Powers: Awesome. All right, cool. Let's let's do rapid fire. you do have the ability to not answer, but you have to buy Geoff Lewis: Okay. Mike Powers: each martinis the next time we're all in New York for every answer that you don't or every question to answer. So first one, Mary friend, kill autodesk, service tightener trim and trimble. Geoff Lewis: I'm gonna go with martinis. Mike Powers: Fair, fair. if you be a G if you could be the CEO of a G C tomorrow, which G C would it be? Geoff Lewis: Ooh. I well you and you and I Mike have talked a lot about Turner and just what an amazing business that is. so Mike Powers: Put together a fund. Come on, you and me, let's go, baby. Geoff Lewis: So I j I just think that and here here's what I would say. Not only are they very, very good at the core thing they do, but they're really thoughtful about all the different business opportunities they bring together, whether it's insurance or procurement or self-perform. so that that's one of the things I think is really clever and creative. And I think that's where the industry is gonna increasingly go. Mike Powers: Yeah. Andrew Zukoski: Luigi, pick cocktieth and then you can be their boss. Mike Powers: yeah. Yeah, yeah. what do you expect to spend on AI as a percent of your team's salary next year? What are you budgeting to spend on AI internally? Geoff Lewis: Ooh. I don't I don't know exactly, but it's meaningful, right? And let me put it that way. we've had to make decisions of hey, should we hire more engineers or should we pay more for for tokens? Mike Powers: Mm-hmm. Geoff Lewis: Right. And we've made both decisions. It's not like you know, universally one or the other, but the spend on tokens is meaningful enough where we have to make like real trade-offs around Mike Powers: Mm-hmm. Geoff Lewis: you know hiring additional software engineers. So it's it it's expensive, I would say. Mike Powers: Yeah. end of twenty twenty seven, if you were to guess, what percent of models internally would be open weight that Procore uses versus not open weight? And then the add on to that is what would you think your breakdown would be between the foundational models and things that are kind of like self hosted, self trained internally at Procore? Geoff Lewis: Who? I'll I'll get a martinis, but I think Tiago should buy the martinis and bring his bring his thought to that. W what I would what I would say though, not to totally dodge the question, is I think we're exploring all of those things, right? So we are not Mike Powers: So more general then, I guess, do you see do you see the bet do you see the market trend is your in your thesis, do you see the market trending more towards open weight, or do you think the frontier labs will continue to be iter iteratively valuable to pro construction that you'll continue like you'll see customers continue to use frontier labs? Geoff Lewis: I think there's probably a barbell. I would say that there's certain things that you I know you guys have been talking about this, but there's certain things that Astro can do that are fundamentally fascinating and game changing that prior models can't do, right? And so I think you have to have a component of frontier intelligence to solve new and important tasks fundamentally. Now, at the same time, can you have, you know, an older Gemini flash model solve some things very quickly and very cost effectively? Yes, for sure. And so I think you're gonna basically have this barbell and there's gonna be a management layer. across that that optimizes for the right amount of intelligence for the right task. That's how that's how I'd think about it. Mike Powers: Last one, what's the most overhyped construction tech in category? Geoff Lewis: I I I mean I've said I've said that that's right. That's right. Yeah. Mike Powers: You're you're gonna get to a three T lunch over here. Luigi La Corte: Yeah, you gotta you gotta answer this one. Geoff Lewis: so I I think the AI native Procore category is is gonna be a tough, tough category for people. I think there's tons of AI native opportunities that are really amazing, right? and I think you guys are are all in a bunch of those categories. I think the direct, let me go, you know, compete directly with Forma or Procore, I think that AI Let me put it this way, as long as Procore is able to shift towards becoming leveraging AI to go much more rapidly in software development than startups can, I think it's still gonna be a a tough, tough slog. Luigi La Corte: Jeff, this is awesome. Thanks for Geoff Lewis: Yeah. Luigi La Corte: letting us peek behind the curtain and you know giving us some thoughts on groundbreak and and whatnot. This has been fantastic. Thank you so much. Any parting words for the audience? It's usually GCs. anything from you? Geoff Lewis: No, just want to say thank you to you guys. I love the I love the pod. It was really fun listening to what you guys have done over the past few weeks, especially. So yeah, keep it up and I'm excited to come back one day. All right. Bye. Luigi La Corte: Amazing. Thank you, Jeff. All right, enjoy your Friday. Take care. Bye. Andrew Zukoski: Thank you.