Adoption Beats Automation: The Future of Preconstruction with a Senior Estimator from Hensel Phelps の全文書き起こし

better at doing this, trying it. But but me personally, like mine okay, I learned the workflows, I see how it's done, okay, what w what's how can we scale this? You know, what what are holding me up? What are the things that are mind-numbing for me? 'Cause I'm the kind of guy that if I something and I can do it and that's repetitive, I'm like, how can I kill the repetitiveness? So and then can we scale this to people? Like like training people so hard. Like just one example is like, hey guys, here's how you do a bookmark to a website URL. It's like, it's even like the little things. but that's what it's about to me. Does this is it scalable? And can we make this easy for everybody? Like that's the ticket, that's the wide open ticket for for for me. Can I learn it quick and ⁓ can else learn it quick? And I'll and I'll you kind of an example. With all these takeoff tools, we we went through. Man, I've looked at so many you'd puke and it's been an adventure. but one e one example, and I and don't not mean in this bashing in any way, but Creo. Creo is a great tool. Like I freaking I I really like that tool and I was like trying to go that direction. But then I started to realize, man, this is a little bit more this is too complicated. I can't train people this. ⁓ It has be OST level O. OST I can learn in twenty minutes and go, right? Boom. Bam bam right. So that's that's an That's that's an example right there. So Luigi La Corte: I see. Andrew Zukoski: What are some of the key things that make a tool trainable or not? Like is it the interface and ease of use? Is it support resources from the company? Is it fit for like how a particular team or a particular company like does a particular workflow? Like what makes something easy to train? Mike Powers: I'm sure. you Kyle Bonde: Well mine is okay, here's our workflow, here's how you do it in their tool. Like i is that is that bridge very short and you know, nice and clean and less buttons and you know and another thing I'd say is like just little ⁓ short videos. Like me give you another example. Edify, if you go to Edify training sites and click on their text, they have they're so text heavy. I think I'm park dyslexic. I'm I'm like, guys, I don't want all this freaking text, like Give me a short video clip and I can learn it like that. And I can suck this stuff up. Like Beck Technology, like they got some great videos and training. And I scoured all their websites and watched all their videos, and they have some great, great training stuff out there. So that that's a tip for me. I like, especially when I'm working with if I'm working with you guys, hey, send me a little video clip back and forth. I f I find that nobody really does that in company. I I just go to Loom, record a video, send it back and forth, and That's how I learned visually. but ⁓ really answer your question, it's ⁓ it's all about our current workflows, can see them in your tool? Or just give me like a concept so I can, ⁓ I can I could probably our workflow in your tool. And I I get I'm I'm the early guy. I because I can like, okay, this looks like crap now, but but I see if we worked it a little bit more and we could scale this company wide. Luigi La Corte: Yeah, that that makes sense. It's really hard to get people to change their behavior. So I get mapping it closer to workflow really helps them as opposed to asking them to change their workflow. ⁓ Kyle Bonde: Yeah, but but but I gotta follow up question with that. That's gotta be hard for you guys. Like we have it how we wanna see it. But then you guys gotta like man you gotta work on your end to make it show like how we want it like then we're like, I don't wanna share our data with you. You probably ⁓ do all this stuff, right? So I mean how do you Andrew Zukoski: All of our customers are very easy to work with on the implementation side. I don't know what you're talking about. Yeah, I mean, it's the classic. It's like a classic challenge of software, Like, SaaS, you to adopt the workflow from the tool that's been polished from, like, the of the industry, or do you want something that fits the way that you already did it? And different vendors choose different along the way, right? Like ERP systems. ⁓ Luigi La Corte: Yeah, it's Kyle Bonde: Yeah. Some somewhat harder. Yeah. Andrew Zukoski: tend to be the most configurable, right, or accounting systems. And then at the other end, I would say probably takeoffs in our world are probably a little more consistent workflow, if only driven by OST's dominance for 15 years. But yeah, not as classic. Kyle Bonde: Yeah. Mike Powers: I think it's one of the things that's changing though. is this like, when I was selling Building Connect in 2016, we were like trying to get people off email and Excel into a platform. And now we're doing pushback into like, hey, can't you just live in email and Excel and it be an headless tool that does things on your behalf. And I think it helps get customers into better positions of like, hey, like we'll meet you where you are. We can do X, Y, and Z to do things. Cause you're breaking it down to do like more of like jobs to be done. Like if you need to do a takeoff. of drywall, like you don't necessarily need a visual application to do that. Cause you just get an email with it linked to the PDF and you like talk back and forth with that email. Um, would that help Kyle's life out because Kyle's still living in Excel. I think there is value there. Uh, I know I've been very vocal and like anti, like I think AI takeoff is commoditized away. You should be able to, you can do it in Claude out of the box right now. Uh, I think there's something there. where it's like, how do you meet the customer you're at? Like you should push back at a technology companies now to do that. not 2016 where it takes six months to release a feature. Kyle Bonde: Yeah, and and I'd definitely add to that. Like have you ever heard of have you heard of the tool lovable? Like, I'm like

in love with lovable. Like, literally it's it's been my jam the last two weeks. I'm ⁓ my god, I can show the I could show this idea instantly and and even change it i at at for me, like I could drop PDS and here's what it looks like. Like you can Luigi La Corte: Of course. Kyle Bonde: You you we could go down a long rabbit hole and lovable, but we don't need to go there right now. But man, the things that the the things that we can do now with software, as me, I'm the end user and I can make mock-ups and scale things to idea anyways. The hard parts underneath the iceberg, right? All the stuff you guys do. But least I could do the kind of apple here. Look. What could be? Mike Powers: We're getting, like, we are getting to that world though. Like there's this, I even this idea of like AI generated UI specific for a customer. So if you can imagine level interface for Kyle on top of Luigi's provision where like he actually doesn't even like there's not a firm provision UI. It's literally just like ask question and build it out. Luigi La Corte: Yeah. Kyle Bonde: Yes. Andrew Zukoski: I don't think so. for a number of reasons. For a number reasons, like, first, like... Kyle Bonde: That's so sounds futuristic. Mike Powers: Why not? Andrew Zukoski: We're talking, like think one of the coolest things that the AI coding tools have done has democratized access to code. He used to have these like low code things and like complicated spreadsheets for building like applications and they were okay, but like lovable and Claude code and cursor and so forth are completely next level. Like my 70 year old dad is hacking nonstop with OpenAI and he hasn't read a line of code since graduate school. Like it's all, it really is amazing. However, two things. if everybody codes their own thing, the complaints that we hear about point solutions, one for each workflow, ⁓ now becomes point solutions, ⁓ for each workflow for each person. Like you have less consistency, you have less plug and play, people are going to interpret information differently if they're given different UIs for it. Like I think it leads to fragmentation, the application layer that like people just don't want. And two, is actually something to be said for learning from tools, There's always this give and take between adapting to what a given customer does best and like. Kyle Bonde: Yeah. Andrew Zukoski: bringing on board best practices that have been refined by others who have gone before. the, what's good example? Like the probability weighting in our CRM. I learned probability weighting by deal stage, which is like a key concept in sales because it showed up in our CRM. And every time I was like, what's our open pipeline? was like, your weighted open pipeline is this much. I was like, no, it's not. It's like three times that. And it was like, can't count it that way. You have to like. You have to grade it by stage. And so I think... ⁓ Mike Powers: But changing though. Like I don't sign into Salesforce.

Ever. I for Kyle Bonde: Bees. cool. ⁓ Andrew Zukoski: It's interesting. Mike Powers: I like, so in this example, so in, I'm sitting in Kyle's seat. Andrew Zukoski: You guys implement, wait, sorry, you implemented Salesforce? Mike Powers: We actually did Monaco, but no one on this call knows what Monaco is, but it's the same idea. Monaco's MTP. You do. But in Kyle's seat, ⁓ the interface is clawed it to all these headless tools, Kyle can show whatever he wants. It's not coding. If Kyle's like, I want to see all the X across all projects we're bidding on right now. Luigi La Corte: Course I know what Monaco is. Kyle Bonde: Yep. ⁓ Mike Powers: Like why does that need to be an interface in any of those applications? Kyle's talked about, I know, like five different tools. Like all those five tools should have MCPs. You should be to go into Claude and ask the question. Kyle Bonde: Well well well let me let me add to what Andrew said, 'cause I kind of agree with him. What this is one thing I'm scared of. Like take love clawed code or lovable for example. I I made this epic I I example. So so I made this awesome tool. We have we have the standard owner report, right? That we deliver it to owners in Edify. It just looks like an Excel file. ⁓ right, broken up by uniformat and bid package, right? CSI. Well so I I I took this in I made made a ep I'm gonna call it an epic dashboard. It's an epic owner dashboard. Now you can take this report everybody's used to using, you drop it into this to tool I built, it brings this amazing owner dashboard. It gives a shareable link to the owner. It's with like a Hensel Phelps URL. It looks amazing. It's all interactive. You can do slicer surfaces. It's like, my gosh, we can make this stuff. But that's but to Andrew's point, I'm a little scared if more people like build these dashboards, we're not gonna look consistent as a company. So we kinda have to really watch what we do if we all can do this. Like, ⁓ man, we gotta have to be consistent on on the way we do this. Mike Powers: Cloud skills, cloud skill to do that exact thing and you'll have consistency across your work. ⁓ Andrew Zukoski: Are you just reinventing a consistent UI from first principles? Mike Powers: No, because Hansel UI wants to be different than Turner's UI. And software provider is not creating a custom UI for each customer. And that's my pushback on that. we're still, in the early, we're in the early innings of this. Like this is like, like we're like, but even imagine in two to three years, like I got, like what Luigi doing at provision. Like Luigi at provision, the output could be ⁓ like ⁓ completely customizable user. Luigi La Corte: Just in time you are, here's the thing. Yeah, for for data data visualization I agree. Yes. Mike Powers: And like, Hey, I'm and like, I love that he's like, again, it's like the harnessing of like, have a specific pencil Phelps data that I want to like harness around the Luigi data, like provision extraction pipeline. ⁓

going be a little bit different than Turner's doing. why force that mold around like, Hey, no, you have to use UI. Like I say it when you're to be like, anyways, that's my Luigi La Corte: I think it really I think it really depends. it really depends on the workflow. I mean, if you're inheriting like IFT drawings, like you should impose a data schema. Yeah, you can represent that schema however you want in the front end. If you're in conceptual design, I could see a lot more flexibility in both how you extract the data and then how you show the data. But yeah, generally I I agree with you. It ⁓ you know, I wanna get your perspective on this, specifically from like a precon perspective. You were at Precon, we saw, we chatted. you saw a lot of takeoff tools. You saw Boone, Bobbyard, maybe there were a couple other ones. I'm sure you trialed a few. I think you told me that you you saw a bunch. where do they work? Where do they not work? And where do you it'll be in in six and eighteen month time? Eight eighteen months time. Kyle Bonde: Man, like like like yeah, it was advancing pre-construction is so fun. Every year it's like a different theme. This year is definitely AI takeoff and scope generation tools, especially with Luigi with what you're doing. It's awesome, but we'll hit on that probably in a little bit. But takeoff, like ⁓ thing that I'm noticing is takeoffs are niching down like full Like, Mike, you guys are in the mechanical world, material automation stuff. Like, one one tool that I'm interested in. is it an example is called Fresco. They do doors, frames, and hardware only. That's all they do. That's all they care about. They source the materials, they find the complex. I'm like, my gosh, if I were a door trade partner, I'd be like rocking this. We're actually looking at it because we're doing a lot of self-performed doors and hardware. but like Bobby art, they're they're dialing in on like I think I think they concrete mechanical and electrical, which is awesome to see in landscaping. Luigi La Corte: L landscaping. Kyle Bonde: And then Boone, they're they're kind of a more general AI takeoff. One of the I've been meeting with that team kind of weekly and they're it's kind of been fun fun working with them. they kind of How they act is like how the human thinks is okay, human goes to schedule, looks at schedule, then they look at drawings. So that's how their AI works. You pick the schedule, hey, go find me the schedule on the drawings, you match those two up and magic happens. So like so they they did this on a demo, generated all the concrete footings, I'm like awesome. then I said, Hey, change the scale me. And when when they that, ⁓ their their footing size they were 10 by 10s initially and their footing size doubled. I'm like, wait, guys, you have that backwards. The image needs to scale with not the not the it always needs to be a 10 by 10 footing. So I'm like, guys, I can't use this right now. then guess what? Like 14 days later,

they're like, hey Kyle we fixed that issue. Look, it works. Like ⁓ things that these and that was a I thought that was a major like, ⁓ crap, you guys gotta go way down in the roots of your code to fix that. Cause that sounds like a ⁓ like a Yeah, they fix but they fixed it and then Luigi La Corte: Tear everything up. Kyle Bonde: Th this is what's happening. You like suggest a feature to a software company and two days later it's deployed. You're like, ⁓ my gosh, this is the pace of change just going like this, right? So so to on like the six months, I think it's gonna be more like trade partners are gonna marry up with the takeoff tools, the ones that survive. And then at months there's probably gonna be a huge drop off. Like guys, ⁓ AI takeoff is just like table Like if don't have it, you're you're done. Luigi La Corte: Mm-hmm. Kyle Bonde: So that's that's kinda where I think it's going and in in to the fact that man these large language models and the virtual language models, like, is is it is this we even gonna need an interface anymore? Is it just gonna ⁓ it for us? I'm s I'm a little bit scared with what's happening in that regard. So who knows what tool we're gonna be using in 18 months for that for that matter, right? Luigi La Corte: Mm-hmm. and what's your current day to day takeoff tool? Kyle Bonde: ⁓ my gosh, I'm I'm literally gonna do a LinkedIn post. I was just th there's a team of estimators out there, they pulled me in, Kyle, ZZ Takeoff's doing something funny. They're all around the the screen working it, working their thing. I mean, our our whole company season's Z Takeoff. It's been phenomenal. So Luigi La Corte: So it sounds like you know, I met with Z Z, I like the team. obviously they're they're very experienced. I think it's the founder of of OST. Is that right? Sorry, Plan Swift, yeah, that they were required by Construct Connect, if I remember correctly. Yeah. they don't have automated takeoffs, from what I understand. So you're kind of doing the very futuristic automated takeoffs. You're still spending most of your time in Z Z takeoffs. What explains like how do you reconcile those two things? Kyle Bonde: Plants left. Yeah. Yeah. ⁓ my gosh, I got to sit. Luigi La Corte: If these takeoff if if if these takeoff tools are so good, why are you not just in there all day instead of Z Z Z? Kyle Bonde: Well it's really it's really a simple answer. So really adoption beats automation, like period. take OST or Bluebeam, you know, in twenty minutes I can get anybody falling in love with with Z takeoff if they're from OST or Blue Beam. Like period. If if automation that nobody uses is worthless to me, right? Like let me give you an analogy. OST is a horse. Horse does the job, right? But it's easy takeoff is the car. So we need to get people to move to the next technology, it just naturally happens without any mandates. Like that's that's magical, right? Like horses just kinda went away. But I mean we're still hunting for the freaking

self-driving electric car. All these companies, they're just giving me like they're skipping to the electric car that doesn't have the it just doesn't work yet. We don't have the infrastructure for it yet. Nobody can touch it. So, you know, that that's reason. And and I'm also gonna go down into like what's easy to takeoff does, it's intuitive, but then you can go to the next level. This is where the trades like love this tool. You make assemblies so simple and clean, blood, guts, and feathers. if I I could go into like a a simple wall tile, AI takeoff tool will give me the wall tile square footage, good job. Thanks. crap, gotta do the backer board, the grout, the sealer, the blah blah blah blah blah, right? The waterproofing, it's wet, dry, solder strips. You gotta give me all that stuff so I can actually build the thing, right? Material. so in and this is right after this call, I have a call with Z TakeOp on automated assemblies. They have scripting that's just got lived in their demo environment. The things that are happening with building the scope, automating the scope, like And they just drop image search, right? Like this this is this is it's all about the foundation. Like if you get the foundation right, like now you can do the apple at the top. Like is where I give Edify a bunch of crap. I used to give them crap. I'm like, guys, ⁓ when your AI stuff coming? They're like, we gotta build the foundation first. They just keep saying that over and over and over, Kyle, the AI will come once you get the foundation. It's the same thing with easy takeoff. Like all the foundations all the structure, man. Luigi La Corte: Mm. Kyle Bonde: They they don't have a house of cards. Like all these little AI tools that are out there, they're a freaking house of cards. So the so lo long answer to the shiny apple. Luigi La Corte: Mm. if if you were to Andrew Zukoski: So if I phrase this back, this is like an argument for distribution incremental change in tooling ⁓ revolutionary workflows. Like it's an argument for a very approach to introducing new capabilities. Is that fair? So is there a path from there to the Powers patented nirvana of just sit there and plot and chat with the tools, don't have to UI and. Kyle Bonde: Yeah. Yeah. It's a it's Ha ha ha. Yeah. Mike Powers: But. Kyle Bonde: That's what I'm afraid of. I'm like, we gotta be slowly getting people to level up. Because I'm at I'm at the unfortunate spot where I get to see all these epic tools that Mike's kind of talking about. ⁓ my god. But and I think to myself, how are we gonna get there if we don't go very slowly? Mike Powers: So there's like the running quote joke, whoever it's named towards, like if you asked someone that sold horses to build a better car, whatever, that's a terrible quote. I'm quote. It's the the ⁓ if you ask, if you, so that's what think ⁓ is a challenge, like hard challenge in the industry is like, one, like the technology, Andrew Zukoski: Ha Luigi La Corte: Yep. Yep. Kyle

Bonde: Build a better build a better horse. Faster horse. Andrew Zukoski: If you asked a horse if they wanted a car, they'd say, only if it fits me. Kyle Bonde: Ha ha ha. Mike Powers: Technology is like 20 % of the job of everyone in construction. And like you saw that from the Anthropic report of saying like, oh yeah, Anthropic is going to be able to impact 20 % of construction. That's because 80 % of the field are doing X, Z. So I'm curious, like, you know, we're focusing on this, like, you asked the cart maker, like, to make a better cart, they're going to put things on a cart, but should we actually jump into a cart? Because I would not say. that tooling that was built to just continue to make this infrastructure better and better, that will layer AI on top, that hasn't worked in any other industry from a SaaS perspective. If you look at, again, we always use Harvey as example, the legal software market before Harvey and Legora has gotten destroyed by Harvey and Legora. So that's kind of like a counterpoint to that. And I know people always say construction is different. But like there is a bet of the companies that do make this a larger shift in a business model perspective being able to accelerate faster. Kyle Bonde: Well what what is Harvey Bagora? Is that like a legal software that's just really Mike Powers: Yeah, think of it as like a specifically for lawyers to help them accelerate their workflows so if it has due diligence for you, set up workflows, things like that. And been able to be proven is like ⁓ they've gone to like Lexis Next. So these like historical database systems and the Gore and Harvey are structured differently, they're to action on differently. So I'm saying that like, hey, take off for estimating is like that wedge where that can be drastically different and structurally different. But the Gore and Harvey have made that pitch that we can be structurally different and we're showing value to the customer. not sure if estimating is that bucket. Like there's different things in here that can explored. Andrew Zukoski: plan. Like here's two counter examples. One, Gong lives on top of Zoom and Teams and so forth, right? Like I don't have to completely change my video conferencing and communications stack to start to get the very real AI benefits of Gong, right? Most transformative technology we've ever adopted as a business. Second, relatively, I would say is Pendo, where Pendo is a tool we use to manage feature requests and sort of like ⁓ roadmap, right? ⁓ they introduced a feature, like already had that, they had that store of knowledge, and they introduced a feature that listens to all the gone calls and like creates tickets automatically from it. And like, know, Luigi, you built something in-house that does that, ⁓ but that's an example of an AI capability ⁓ that's real, that's delivering value to us today, that was added incrementally to kind of a crusty, Kyle Bonde: Hmm. Andrew Zukoski: like SAS product from 15 or 20 years ago, quite effectively. I mean, the ultimate question would be, doesn't this mean that Bluebeam should win AI takeoffs? Isn't that what this means? It's either Bluebeam or OST is going to win. Everybody already knows how to use OST. Yeah, that's true. Luigi La Corte: I think both both are possible. Mike Powers: Yes. Luigi La Corte: If they can sh if they can ship. If they can ship to the desktop fast enough. Yeah. Kyle Bonde: To the to the web. Andrew Zukoski: Yeah. Luigi La Corte: Yeah. Well think Bluebeam is the incumbent and maybe not from like a takeoff perspective. It's probably like OST or Plant Swift. but they have a huge opportunity, which is why I think they're buying startups. They're hoping that they can accelerate that. They they did that with Fermos. you know, the I think the weirdest thing though is that you have to jump on studio and to go to a cloud session. It's just so antiquated when you think about it. But they have all the distribution and people are familiar and they know the product, and that's why people like incumbents. But if the startups can ship really fast and hammer a wedge, then yeah, then then they probably deserve to win. Harvey and Lagoras like surely did. they are probably winning. I think between them there's a well, sorry. Mike, there's tons of audio distortion coming from you. if if like they they've been able to get a wedge and like they grow really fast and raise venture capital. And you know, we haven't really seen that in construction tech that much, barring a few exceptions. ⁓ so Kyle, where would you put your money? in eighteen months, what will you be using for takeoff? Kyle Bonde: don't know. ⁓ I'm agnostic. I'm not attached to anybody. I'm I'm just for the ones that help our people. Like what can get the estimators the smartest, most accurate, stop kill the risk p period. ⁓ And I do have a well, I am s go on the limb. It's gonna be a tool like yours, Luigi, like Luigi La Corte: Mm-hmm. Diplomatic. Kyle Bonde: You guys are more to the source of truth, the documents, the scope generations, the if can kinda figure out how to do takeoffs, you know, y n never know. Luigi La Corte: Yeah, that'd be great. Well we'll get there. I appreciate that. cool. very helpful. And I appreciate you being, you know, so honest. Mike Powers: Can I take one step? I think one step out matter that I want to just like get your opinion on is like, there Kyle in your world, like Hensel Phelps doesn't want to win more projects, they want to win better projects. Is that a fair assessment? ⁓ like. ⁓ Kyle Bonde: Yep. Exactly. Most high highest value projects that we can deliver for our clients and get repeat work and it's success and we don't burn our subs. I mean this is what we do. ⁓ Mike Powers: Yeah. So then the question for you is that like, it's not a speed question because you're not like, I'm trying to bid on a hundred. I'm not trying to do takeoff on a hundred projects next week. the quality of the takeoff or the

context of the takeoff. So do see the ROI driving towards? It's not cause like an easy ROI is like, Hey, instead of one project I did this week, I did two projects this week and I won two projects because I had a faster takeoff. ⁓ In your world, not true. So how do you see and try to wrap your head around the ROI of a takeoff tool or estimating tool where like Hansel Phelps is probably gonna win that airport, Like we're already kind of like, we're already gonna win that project. Is it just caught, like where do you see actual value with these tools? Since you're not trying to be like, I'm trying to do two airports this week, I'm doing one. Kyle Bonde: Well, in my mind we're starting we're scaling more. So so more revenue equals more projects. I I would say that we do we we can use tech to kinda help supplement that 'cause it's so hard to find good estimators. it's kind of a two prong thing. Yeah, we gotta be more efficient, but we also gotta be more accurate with how we handle the risk and make sure we're signing up the right trade partner, make sure we got all the scope. I can't tell you how easy it is t ⁓ crap, there was a sauna, I forgot to buy that out. I didn't have a budget for that. ⁓ crap, that's a super specialty thing. no. ⁓ Like that That stuff ha it's so easy to miss and estimate and pre count for signing things up that are buried is one little note here, like we we need more help with scope holes and double ups and like that is where this all this value is, getting the period and really. Luigi La Corte: if we well that's actually leads to the next question where you know, we see a lot of companies, especially I I just came out from Built Worlds in Chicago, Kyle. I don't know if you were ⁓ planning on on or thinking about going there. ⁓ Th are a lot of there are a lot of tech vendors. I think you would have enjoyed it. ⁓ but what's pretty clear is that people a lot of people are targeting the scope layer. And you I have my my theory as to why why, but it's it's clearly a problem. Kyle Bonde: ⁓ I'm jealo jealous I wanted to go there. God. Luigi La Corte: but what you see repeatedly, well, what I'm seeing it now from a lot of startups, barring it barring a handful actually, is I'm gonna a document as the input and I'm gonna give you the artifact as the output. ⁓ Claude and ChatGPT are actually pretty good at that zero shot, you know, without, you know, training it or giving a lot of examples. Do you do think that this is the right approach that startups should be taking? Do have any commentary on that? Kyle Bonde: ⁓ my gosh, yeah. So it that was the theme at Advanced Pre-Donstruction, scope generation tools. You got Nomek, you got you, you got there's there's a bunch of tools. I've been diving into a ton of them. Down to Bid is actually the first one I really got very interested in. That was probably two years ago. I'm like, guys, this is it. But everybody's catching up. But let me tell you a story. So three years ago Luigi La Corte: Mm. Kyle Bonde: We put AI in into our bid bid bid tabs so we could archive our bid tabs, get the proposal summary from all the bids, we have all this data, but I'm so glad we started doing that. But we we chose that because that was kind of a point in time where we can finally get consistency. But it it really felt wrong to me, to be honest. so our s our CIO ⁓ Her name's Dolores. She she I I kinda talked through her with this and she kinda challenged me a little bit, which was a great question. She she said, Kyle, does the source of truth get generated? And that question, and I love that question. And back in that day, I was like, man, I was like the takeoff, because there was none of this scope generation stuff going on. I'm like, we gotta get the takeoff down. I gotta get OST off the desktop, like we gotta figure that out. So because that's when everything gets generated, and that flows to the estimate, and that flows to bit tabs, and blah, blah, blah. So we we finally got that dialed in, right? But now our layer is solid. So n so this where it's fascinating to me. So now we're Okay, well let's take this further. Back to her question. Where does the source get generated? I'm like, well, it's a document. that point. So now now I just move back a level and now I'm like, okay, I got the takeoff, got the estimate figured out, but ⁓ we gotta get the documents, the scope generation layer dialed Like that's that's kind where where this is going and what why it excites me. It's the source of truth, right? That's the documents. Woohoo! Let's let's see if we figure it out, Luigi, right? Luigi La Corte: Yeah, absolutely. was just interesting to see. I was ⁓ I was chatting with another founder friend of mine about how many companies that we've seen now do scopes and I can count I can count it very quickly here. but I think it's one, two, three, five, six, seven, eight, nine, ten. Ten that I just, you know, I've seen between AP and Built Worlds. And yeah, the like the obvious is. Who's going to be the winner? Which one? And and what is winning actually look like here? Right. There's one component, which is like, Andrew Zukoski: Is it naturally a winner take all market? Luigi La Corte: I think it's winter take most. Kyle Bonde: Well I think a winner I think the winner is we're gonna call this a category. In my mind, that's a win for the industry. Scope generation is a category, is a tool section, kinda like Takeoff. Takeoff's got a own friggin' all their people. Well now we got a whole new section that people are like, ⁓ my gosh. Well I got copilot. Well does that count? Probably not. Luigi La Corte: Mm. Well, Copilot's a copilot's a good competitor as well. I will I would argue though that yes, there is a category. And I think that the fact that there's so much so many companies chasing it means that it's a legitimate category with legitimate pain, or else you wouldn't see so many companies funded. I would say though that somebody should win to benefit the

categ to to what to to c to benefit the ⁓ the industry the most. And I I I I love competition. I think it ultimately benefits you, Kyle, and it benefits the industry and people like Hensel Phelps and PCL and LSDON, they're gonna have really good solutions. But what happens is somebody will get the most traction, which will attract the most venture capital dollars, which will be will result in the best distribution. Maybe not the best product, but the best implementation. And as you know, as you just said, training and familiarity probably will be will probably lead to the success of a given product. But distribution is very important. So anyway, I know I'm sure, I'm sure you've seen a couple, but I would love to hear what Mike Powers: you Just before Kyle, Luigi, in your position, that is Procore and Autodesk win because they have the best distribution. They have all the documents. Luigi La Corte: They have the they have one of the best chances of of winning. And and and on that pr Andrew Zukoski: or project site. Luigi La Corte: Project side two. Procore is releasing a a scoping tool. Kyle Bonde: Never know. Sartuol, by the way. Andrew Zukoski: Yeah, that's what Hansel Phelps is on. Luigi La Corte: ⁓ Pro Procore is releasing a scope generation tool ⁓ in their ⁓ bid bid management in their in their bid management functionality. So yeah, they have a good chance they have a good chance too. Mike Powers: I was just curious how you thought. Kyle, off to you and your thoughts. Kyle Bonde: No, I I mean you guys hate it. Like I'm I'm kinda I'm kinda pulling for something like that too. Like Building Connected won this won the trade partners, even though that tool hasn't changed for ten years. they're gonna get taken over, my opinion. But somebody won though, right? And help helped all that. Mike Powers: By who? By who? By who? Nobody has, no one else. Andrew Zukoski: Who's gonna go like tell all their trades to update? Maybe someone who's better integrated with like a now dominant prequel tool, right? Like that's like the rock. No, you don't think so? You would know. I mean, you sold the thing. Why am I talking over you? Mike, let me explain to you why the thing that you know inside and out and took to market. Yeah, go ahead. Sorry, I'll be quiet. Mike Powers: Which is fact that no other precond solution has a network. ⁓ ⁓ ⁓ good. ⁓ Kyle Bonde: Right. That's the no it no no no no Mike Powers: The biggest gap is no one else has built a network. this is befuddling to me. Every pre-construction tool should have a network. that means is you connect nodes between parties. Like engaged with multiple subs, ⁓ multiple and multiple owners on every engagement and pre-con and every tool that's gone to market does not have a network. That's building value. And no, gosh, I just forgot. can ⁓ picture him. Kyle Bonde: That is the Yep. Luigi La Corte: Mm. Mike Powers: ⁓ my gosh, prequel tool that's in market right now is like they're not going to build a big management

software. Luigi La Corte: Precon suite? Like a new one or an old one? Kyle Bonde: Comp, comp, come, come, compass. Mike Powers: No, Compass, thank you. Michael Ho, shout out to Michael Ho. Grinding for 12 years. And the other issue with pre-bid management is the market, the TAM is 15 to $20 million in revenue. It's hard to invest in. Kyle Bonde: Yeah. Yeah. Luigi La Corte: ⁓ compass, yeah. Kyle Bonde: So the one thing the one thing I'll add, I'll one-up you on this. Like you guys did this Gong thing, I don't even know what Gong is, but it's on top of Teams and Zooms and it monitors your calls or whatever. Like just a yeah. Right, imagine if we had that kind of level inside Building Connected where you're getting all this insights and aggregates it up and just to your p to your point, the rock is the network. How do we Andrew Zukoski: Yeah, it just records them all and creates summaries and lets you search them. Kyle Bonde: displace the network. I don't know. Any d you think anybody will kill Building and Connected or compete with them? With with the the pace that they're going though? Luigi La Corte: I think yes. Building connected sold software. Like if you look, the the the current paradigm is not let me create an interface. It's let me automate your existing workflow. So why can't why can't I create an agent that sits across many tools and just sends out emails from your inbox? Mike Powers: So, so I'll say, yeah. Andrew Zukoski: you're gonna have to retrain the subs to go to a new place. Autodesk is doing all this work to migrate building connected into whatever the fuck they're calling it now, Forma. They pointedly did not change any of the URLs that the subs will need to go to. ⁓ Luigi La Corte: Well, they already know their email. Mike Powers: for a month. Luigi La Corte: But the email, just email. Hey, send it to me via email and that's what that's who I got it. That's what they do it anyway. Like most of them still don't respond to Building Connected. You know that Kyle, you know that you're getting responses from subs probably in email. You're getting a lead letter ten minutes before bid close you don't do a lot of lump sum, but still default to email. Mike Powers: So I think the... Kyle Bonde: Shape. Andrew Zukoski: Yeah. Mike Powers: So the white horse is edified as If there's a company that replaces Building Connected, it's edified. Kyle Bonde: Yeah, but they don't have anything that's rated by our management. That's a whole another ball of wax that Mike Powers: What do mean the entire team built it? Andrew Zukoski: That's like a single tendency solution right there. That's not built for network. Mike Powers: Edify is not. Edify is not. Andrew Zukoski: Yeah it is. Kyle Bonde: I w I would say no to Edify, and I'm the Edify freaking champion. But hey Andrew Zukoski: Yeah, yeah dude, it is. interesting. Mike Powers: That's my bet. Want to do a polymarket? That would be my polymarket bet. Luigi La Corte: Polymarket. Andrew Zukoski: Actually, I have a bet. I would bet on ClearStory. Because they built a network. Yeah. Luigi La Corte: ⁓ Clearstar is very interesting. Yeah. I was catching up with Cam at Building Connected and sorry, at ⁓ at Built Worlds, and their network is just so impressive. The amount the amount of the amount of new signups they get from subs every month, it's it it blows my mind. Like I wish I had top a funnel like that. But they've really cracked the network. And it's why it's why that ⁓ Kyle Bonde: that is it. Mike Powers: adult worlds. Andrew Zukoski: It's huge. Kyle Bonde: Ha ha Andrew Zukoski: ⁓ my God, it's like top of, it's bottom of funnel that feeds top of funnel. It's this Ouroboros like snake eating its tail and creating giant delicious snake states. Luigi La Corte: Yeah. A bigger s snake. ⁓ That's that's how Cam that's how Cam pitched the first VC meeting, just like like a frenetic snake illustration. ⁓ yeah, I I like I like their product. Say it again. Mike Powers: Just remember that did work at Building Connected. We struggled with that growth story of Building Connected. ⁓ Just as a note, we had every subcontractor in the country using Building Connected and we tried to sell them more software and we did not do very well doing that. Just as a note. Luigi La Corte: Yeah, it doesn't mean that it's not possible, Mike. It just means that, you know, previous a previous ⁓ way way didn't work. But in like TNM like CORs, like there is a product that's already working. Like they've already figured out that second product, basically. It's an ops, but if everyone is on the network, that network can be distribu. Yes. Yeah, they've already figured out the second monetization and like by as a byproduct of that, they have the network. ⁓ going back to Kyle, because his his time's super, super precious here. Kyle, when we ⁓ Andrew Zukoski: It is a different product. Luigi La Corte: When we talk about at least the pre-construction AI landscape, you know, one area I don't see a lot of activity in is in the C D stage. And like when we hopped on a demo, this was kind of where you went to as well. Hey, if you can go artifact in, scope out, that's great. But when I'm starting at the napkin sketch, I don't have a great artifact. Do you think so number one, are you seeing companies target this area? To do do you want it to? And then maybe tell us a little bit about Kind of the problem that you face at at Hensel Phelps. Kyle Bonde: Well that that's a good question. So you know, kind of stepping back a second, so you know, the industry was developed on, you know, the hard bid workflows, right? ITB takeoff estimate, bid leveling, subcontract, awesome, we got that down. We're really good at that. We have processes of workflows, train our people, do There's there's in there that we need to get better on, like bid leveling for example, get off Excel. subcontract handoff could be bit better in the ITB world, might be better someday too, like we hit on Building Connected.

But man, look now now shifting. And this has happened pretty recently. It's kind of to your guys' point on change. Like, how do you get people to change to anything? They're stuck on this way, go that way. Well, owners are dictating, and hey, we're not dictating, I mean we're just doing it faster with design build methods, right? So now we're Our Hensel Phelps, we're literally 70% of our work is negotiated now and 30% hardbid. So it's completely flipped over the last few years. So now Andrew Zukoski: That's actually relative to your peers in the ENR top 20. That's actually still a high proportion of hard bid. Yeah. Kyle Bonde: Hmm. Interesting, yeah. Hmm. Luigi La Corte: Yeah, like J E Dunn doesn't do any hard bid, they're just a hundred percent negotiated now. Andrew Zukoski: Yeah, there are, I mean, there are notable exceptions to the Prini, uh, Rudolph and Sutton does only negotiated, but like all the giant infrastructure, like two to Prini's like stated strategies to hard did lump sum projects that no one else has the bonding capacity for it. Like, so you're not, it's not alone, but like it, that's a trend to others that are seeing as well. Kyle Bonde: And it depends on the region too. It's it's all we have very h high hardbid regions specifically that keeps that at thirty percent. But it's like, how do we how come there's nobody there's nobody really messing around in the cause this conceptual world 'cause it's so freaking hard. Kinda what we talked about before, you gotta get your base data up so you can help do predictive models. You know, you have all these cost model tools, you can grab your history and do all that comparisons, but there's nothing that takes the context of your job, looks at all your kind of data, helps helps you give you a a executive summary of what could be, you know, how to predict this and ⁓ I can take off this, but no, we're missing ⁓ XYZ. We we we know we need elevators, right? Like little little things like that. but that that's that's kinda The the other big shift that's happening, estimators they're not counting in corners anymore. It used to be, hey estimator, go count this, but buy it out. Well, no, no. now we're like earning owners' trust, working with the architect, trying to figure out the architect sick or cow so we can do VE decision management, all the stuff you do, Andrew, right? ⁓ estimators are turning into trust agents. so do ya you know what do want startups to know? It's I don't know, it's hard to gotta get the foundation right before you can help do do predictions. Like, I had a perfect tool, here's a perfect tool. Like Luigi, I think I said I gave you this graphic, right? We you AI and specs into this AI monster machine that Luigi makes, it parses all your scopes and materials, and it sends sends it already sends this to a takeoff tool. I'll pre-format it for me to with all the formulas, because setting up your assemblies and it's easy to takeoff or whatever tool you're using, you gotta you can

go measure it, because I don't trust AI to measure anything. And that sends over to your estimating tool. But then this AI tool also ⁓ you're missing this. Do you have this in there too? like that's utopia that can do at early stages. Luigi La Corte: We ended up making that as a as a byproduct, which it would be cool to show you, but r irrespective of that. It's I think people yeah, I think edify like a really interesting opportunity with Edify, even even in join is look I have all these all these projects that we've already ⁓ built together. We already have them broken down by line item. Great, okay. I'm in I'm in conceptual design. An owner has given me a cube on a piece of land for a data center. It's a two hundred thousand square foot data center. Give me a price broken down by by division. and and give it to me to the line item so that I can say, okay, great. yeah, let's add another hundred thousand square feet of of of rack space, for example. Okay, great. Here's here's the impact of that. And you know what I think is a fantastic analog for the UI. It's Claude Design. I don't know if you've used Claude Code. Claude code came, well, Claude Anthropic came out with ⁓ this is for our unin uninitiated listeners. Mike is laughing at. Kyle Bonde: There is. Luigi La Corte: How I'm describing this, but design. If you haven't used it, Kyle, you'd really like it. it helps you design things based on inputs, but the design interface is so interesting. it's upload everything that's specific about your company. So imagine Hensel Phelps, you know, this is what this is our process, and this is what we like. And you know, we like uniform ad to this level, not CSI. Fuck CSI. and then we like this. Okay, great. Now let's start with the design. Okay, I'm gonna do a 300,000 square foot document. Here's the the material that the owner gave me. generate the output in join or edify. great, let's start there. And now I can prompt a change. I think that is a I think that's a winning product. ⁓ I think the best company will do it. Mike Powers: Have you seen, have you seen SnapTree? does that ⁓ a cool way. It has a good example of like what you're talking about. Luigi La Corte: cool. Kyle Bonde: Ooh. Luigi La Corte: And Andrew disagrees. Andrew Zukoski: Yeah. Mike Powers: You can try it. You can try it out. You can get a free account. You can do that exact thing. It's cool. Andrew Zukoski: I have. Kyle Bonde: Well well the well th what you're hitting on is is is key though, right? You're you're dropping in whatever you think you want, describe the outputs and one of the things you hit on ho we we've been focused on kind of high level comps, right? Grab a bunch of hotels, compare there you go, that's your cost. Well, now we actually have the power to get down to the line item and adjust it to their specifics and really make a custom estimate for the owner. Like that if we could get ⁓ high high level estimates earlier that We're all gonna be better. And like got these this tool that will suggest V E's on top of that. Andrew Zukoski: Yeah, it's interesting the like, cloud design for like one of the AI use cases we've heard from a few. companies that we work with is really good search over the SOP documentation. ⁓ like there's a, I've a lot about the Hensel Phelps way. There's the way for other companies as well, but making it easier and easier for people to understand like, what do I do next? Or potentially automate like portions of that. I don't know if it gets all the way to throw it in, generate a building. does seem to be good things in design, more power to people, displacing Revit. I know on the click a get a building side. Kyle Bonde: Like the the things that are coming there, like throw in PDFs, generate a three D model, like, ⁓ my gosh, the I I c I've literally built that tool. I drag and drop my two D take up in there and it builds a three D model of the page I'm on. I'm like, ⁓ my gosh, this is working Luigi La Corte: Does it give you the break like the scope breakdown by division? Kyle Bonde: No. I mean I'm just talking Luigi La Corte: I think that's what you need to get to. That's the level of granularity, right? Like, okay, what do you have for finishes? Okay, we want carpet. Okay, no, actually no, we want we we want vinyl. ⁓ okay, great. Let's swap that out. what did you choose for your what did choose for your your air handlers? We use Diork. No, don't use that. Use something else, use Tyco, I don't know. Mike Powers: What? Luigi La Corte: How many tons of of how many tons of cooling do you need? I need one ton. One ton for my data center. that's that's just an example. Mike Powers: Chill. Kyle Bonde: Spi Spe speaking of which, back into getting to trying to use your tools, like one of the things I love that Mike did is, hey, treat Mike as Mike Powers as a bidder. Like, I'm just gonna invite you to my building connector project. It's a public project. Hey, what can you do with this? Whoa, cool, that's our project and your tool. Like, cow, I don't have to go through all these hoops of security. It's I You know, the fact that you did that a while ago, like, ooh, it it shows me the possible. So keep doing that. Mike Powers: We actually don't want anyone signing into our application. Kyle Bonde: But but but you could but you can show people is the point. Like you're not don't give them access, but hey, here's what you could look like, right? Luigi La Corte: These days. These days. Mike Powers: I'm saying our major users don't sign the application. It lives in your email. Luigi La Corte: Yeah. Kyle Bonde: ⁓ you d you just kinda go in, you get in, like just Mike Powers: Forward me an email, we can sign into your bids at email. We live in your Outlook. We call it the Net. We merge all your projects together, structure all the data. Hey, do you want a bid or not in this project? Based on the insights that we're providing, here's your line card badge for the manufacturer. Andrew Zukoski: Yeah. Kyle

Bonde: ⁓ that's interesting. Yeah, yeah, but then the next line I'm gonna get is, well, is that secure? How do we make sure our data's protected? Like I'm a little worried that I can't sign you know, we're all Mike Powers: You can sign in if you want to. Oh, you can do whatever you want. But I'm saying like, hey, you're living in your email right now. You're communicating with your subcontractors. You're back and forth. You can still do that same exact experience all the data gets structured back and it gives you insight. if in your world, it'd be like, let's say John, the mechanical subcontractor said, hey, we're missing a chiller on this drawing, which should be extreme. Kyle Bonde: Hmm. Mike Powers: then that could automatically feed in and draft an email for you to send it to the other three subcontractors that are bidding on that project saying, hey, there's a missing chiller on three and you don't have to do anything. Kyle Bonde: And you just hit approve, you just yep, send it. Cause you have this gnarly integration with the connected, I'm assuming, right, you're talking about? Mike Powers: We have everything. We have 75 bid management integrations right now. Kyle Bonde: So so that's why you're saying y you like don't think building connector will go away too, maybe. Mike Powers: I think it's again, it's a substrate that actually doesn't need to be engaged with and that will continue into the future. ⁓ You're gonna live in your email. You're gonna live in it. Again, I'm just like fanboying provision right now, but like if you were doing a scope analysis and provision, cares what the bin management tool is? You need to get the documents. It could your email, it could there's a company called Luigi La Corte: Mm. Kyle Bonde: Mm-hmm. Mike Powers: that has been around since like 2002 that's a bid management software. Like, cool, like, like it doesn't matter. Like, it's like, that's. Kyle Bonde: Yeah. ⁓ Luigi La Corte: Mm. Kyle Bonde: Well well Well me give you this crazy example. So I'm a lovable nerd. They had they added aggregate peers on this project kind of last minute. I'm like aw craps. So I took all the geotech notes, the draw my takeoff, the drawings, the requirements, dropped into lovable. This is a public project, so don't worry, I'm not doing anything illegal. and then I shared that link and I shared that link with all the s ⁓ all the trade partners and where you get they could download all the detailed documents that's specific for them, gives them little mini packages. think that's what you you do, Luigi. You give trade partners mini package, just the stuff they need, but man, the dashboard it made was epic. And and I want that for every trade partner. Here's all your stuff, consolidate it, all the risk there, get all this stuff for me and life's good. We'll have a successful project. ⁓ that that's that's that's where I'm hoping we get to. Luigi La Corte: Future is here. Mike Powers: Live in your email, live in your Excel. in your SharePoint, live in your, wherever you wanna live. Luigi La Corte: D the dark horse here is you know, co-pilot. Kyle, I know you like copilot. I know you use copilot and you build your own agents. It's another it's sometimes I hear like, hey, you build scopes, but like I built one in co-pilot. So like with the same product, right? And then we know we obviously explain the difference and why auditability is important and why accuracy is important, but still it's brought up in the conversation. As long as brought up in the conversation, it's a competitor, whether you like it or not. tell us about your experience of copilot. And I think you mentioned as well you built your own agent. it sounds like it was a very impressive one, for like a quite a large federal bid. How how did that go and what did it teach you? Kyle Bonde: it was it was a fail. I love talking about failures. So we may maybe I shouldn't talk about failures, but hey, that's all we learned. ⁓ ⁓ know, we we have the ability to skip I learned how to make an agent and share it with up my team. Hey guys, ⁓ out this agent. You can use it because I I build the instructions. I'm awesome. Luigi La Corte: You can talk about failures. Andrew Zukoski: how Kyle Bonde: So then we actually took this up to another level, talked to IT, hey, can we deploy this agent company wide? And and we're doing that with other tools. We don't have one in precon yet. We're getting close, but we actually had a a a note in there in our instructions to this AI agent that says deprioritize drawings, because we we have this assumption that the VLMs and it can't can't really read the drawings very well. so we put that prompt in there. And I w we deployed it to a few people and they tried it and this thing was just kind of giving them really crap information. I mean that was the fail. Like we said deprioritize the drawings because we don't trust the VLMs it can do, it can't read drawings. We have this impression. Well, it's actually now that we got we have a few models in there now, we're finally getting GPT and Opus in there, so it's it's getting smarter. so now that we that was the fail, by the way. So we we did that and and now we're Now that I removed that and said don't depri prioritize drawings, we flipped it. Because that's where all the information is, right? And now because that's what I did on my agents that I built. I'm like, man, the ones I built work so freaking good. Why is this one so bad? ⁓ it was that direction that we said to deprioritize the drawings. So it's a it's it's that that's that's that. then I'll and then I want to pivot it a little bit too with Cold Pilot. So I was on the Andrew Zukoski: Hmm. Kyle Bonde: pilot team to roll out co pilot and I was like, guys, what are we doing? This is it it felt like it wasn't that good, right? But this was way back when Chat GPT was starting. We were kinda one of the f first companies I feel like to start rolling out Copilot, which I felt cool. But I'll like

I'll try it in Excel, I'd ask it for something, it wouldn't do anything. I'm like, what are we doing? s but but now I'm I'm to but that was like two years ago I feel like. But but now every day I I I hear a story, somebody s comin came into my office yesterday and told me about how they use Copile to help them. I'm I'm hearing all these stories. We share stories company wide. Like for example, I dropped in all of the Edify our company codes into Edif from Edify, dropped them into Excel and said, hey, cross coat all ⁓ all of these line items from the takeoff ⁓ all our company codes so I can import them back into Edify and just and boom, boom, done. Like the stuff that's possible now with Copilot and Excel is just It'll blow your mind with what it does. Andrew Zukoski: Is that adoptable? Like, you, like you've got a, you've got, you've, you've clearly got like a really high bar and a lot of taste around what can be rolled out to you know, your peers and your teams at, ⁓ Phelps and probably beyond. Do you, do you think that you can roll out copilot based workflows? And I don't know the answer. trying to ask, I'm not trying to challenge it. I'm asking a very genuine question. I don't know the answer to. Luigi La Corte: Mm. Andrew Zukoski: Like, you think other people will learn how to open up copilot, have it do things instead of a spreadsheet, have the taste for and the feel for when it's right, when it might be wrong, when it needs guidance, how to prompt it, that sort of thing? Kyle Bonde: Man, it's I think the seeing it work. Like that's the cool part. Specifically Excel thing, since we live in Excel in a lot of our worlds to help help tune it to what you need. I'm okay just change that model to GPT 5.6 and then it'll really work for what you need to do. like the fact that people are just picking it up, you hit the button on the bottom right, I mean it's really it's trainable. It's in Excel, people feel great, it it highlights what it changed. ⁓ Andrew Zukoski: Yeah. Yeah. I did install Excel just so that I could use Copilot. Like it drove business to Microsoft from me. Like, yeah, it's a little scary to admit that. Kyle Bonde: Mm it's Yeah. But but like back the counterpoint of what you just said is like ex emails, I'm like, I don't trust this thing to draft for me. I don't trust like other pieces of it, I'm like, let me give you another example. Like, I do we ⁓ have our agents point to the drawings, we have a drawing specs, all that stuff the Ouija does, right? And and you ask it a question, give it some prompts, and you finally learn what it is, like, okay, cool. I've had some crap answers too that I really lost faith in Copilot. Luigi La Corte: Okay. Mm. Yeah. Kyle Bonde: in that regards, but pretend it was good, right? You you got a good answer. but then a week a couple weeks go by, you're like, crap, what was that? I forgot about that. Where is that chat? ⁓ you got a million chats, you can't find it. You lose track of these things so quick. And it Lu Luigi, that's what I like about what what you're doing is you're gonna have like specific questions for each scope ⁓ out for every project. And it it'll just go mine those questions for you and they're all there. Ask same freaking questions at every job. Why why is there not just a central spot for all this Luigi La Corte: Mm-hmm. Kyle Bonde: Agent data mining for me. Like, let's go people. make this for me. Luigi La Corte: Yeah, there's so much possibility and I think it's just gonna get way better. You know, we're still kinda early innings and how ⁓ these LLMs are are being used on site and yeah, I think there's gonna be some really magical products. I was just telling the the the team here before, I really do think that the next unicorn of construction will be in pre-con. There's just so many so many pains. ⁓ I don't think I don't I I don't think Mike agrees, but Kyle Bonde: So glad there's so many people come into pre-conto. It was it was it was dead. robots somewhere. Luigi La Corte: Yeah, pre construction robots. That's that's the that's the reality that we don't want to face. Yeah. Pre construction ⁓ I know, I'm just I'm just kidding. But robot it'll be it'd be the most fun. Kyle Bonde: No no no no. Yeah. Yeah. Andrew Zukoski: Well, you got like, got the like six axis robot and like a highlighter pen on printed drawings, like. ⁓ Kyle Bonde: ⁓ ⁓ ⁓ Luigi La Corte: That's right, yeah. That's where robots are deployed in office, just sitting next to Kyle. Andrew Zukoski: Like two of them using a slide rule. Luigi La Corte: robot's using C Z takeoffs, no API, just the via keyboard. All right, let's cut it here. Kyle, this has been this has been fantastic. Thanks so much for telling us about your opinions and how you're using AI and technology. I think the industry really loves reading your posts. I like reading your posts. but it's it's it's great to see someone care so much about about pre-con. I care about pre-con a lot. So anyway, I appreciate you and thanks so much for for hopping on. Andrew Zukoski: Yeah. Kyle Bonde: Yeah, appreciate you guys too for pushing us. We gotta push each other. Like don't say yes and don't stop. Luigi La Corte: Exactly. Yeah, always keep pushing. All right, we'll we'll call it there. Thanks everyone. Andrew Zukoski: Thanks folks. Kyle Bonde: Yeah, right on.