Why AI Readiness for Construction Companies Starts With Your Data (Not Software)
Most construction companies chasing AI start by shopping for software. This episode of Construction Hot Takes pushes back on that: Jeff Robertson, Adam Cooper, and Greg Gorman argue the real bottleneck almost always sits one layer down, in the data itself, and the first question they ask every client asking about AI has nothing to do with a tool.
Jeff Robertson opens with the question he asks clients weekly: how good is your data, really? The inspiration was a meme making the rounds: data as a pile of crap, run through AI, and it comes out the other side sparkly and unicorn-colored. Still crap. Garbage in, garbage out did not stop being true because the tool got smarter. Adam Cooper raises a real wrinkle: newer AI tools with built-in connectors can now reach into a CRM, a meeting recorder, and a set of documents directly, acting as the enterprise data layer without a company first building one. Greg Gorman’s pushback is the episode’s hinge, that shortcut still needs consistent inputs, or the AI just gets confused faster.
The back half turns to the job site: training AI to catch “fuzzy match” vendor records typed five different ways, an AI agent that turns a superintendent’s daily log into a consistent report, and drones and robots replacing manual truck counts. The hosts land on a reframe of the whole question: this was never purely a data problem or an AI problem — it’s an organizational readiness problem.
Watch the Episode
In This Episode
Why the first question to ask before buying AI software is about your data, not the tool
Why “garbage in, garbage out” applies just as much to AI as it did to old accounting systems
How AI tools with built-in connectors can act as your enterprise data layer — and where that shortcut still breaks down
Real client examples: fuzzy-matching vendor records, an AI-built daily log, and drone-collected job site data
Why this is ultimately a question of organizational readiness, not just AI or data readiness
Episode Chapters
00:00 — Kicking off: the readiness questions clients ask every week
01:09 — The “garbage in, garbage out” AI meme that inspired this episode
02:47 — Step one: figure out where your data actually lives
05:19 — When inconsistent tool use breaks a company-wide dashboard
06:50 — How AI connectors can become your enterprise data layer
10:13 — Training AI to catch “fuzzy match” vendor records
12:49 — Case study: an AI agent that standardizes daily logs
15:14 — Drones, DroneDeploy, and cutting human error out of counts
18:20 — Why construction paperwork has exploded in 15–20 years
23:59 — The AI hype curve and the “trough of disillusionment”
26:37 — It’s not AI readiness — it’s organizational readiness
Notable Quotes
“Whatever goes into the system, into AI, if it’s bad or it’s inaccurate or it’s not thorough — it’s going to come out. It might be sparkly and pretty and rainbow and unicorn, but it’s going to be crap.”
What does AI readiness actually mean for a construction company?
AI readiness means having consistent, accurate, centralized data before adopting an AI tool. Without it, AI just processes and reformats bad data faster, producing polished-looking output that is still wrong underneath.
Why does garbage in, garbage out still apply when you add AI?
AI does not fix bad data on its own; it amplifies whatever goes into it. If project managers track budgets, quality, or RFI turnaround differently across projects, AI output built on that data will look sophisticated but stay unreliable.
Do you need to build a data warehouse before using AI?
Not necessarily. Newer AI tools with built-in connectors can pull directly from a CRM, meeting recordings, and documents, effectively acting as the enterprise data layer. The data still needs to be entered consistently for that shortcut to work.
Can AI fix inconsistent construction data on its own?
AI can be trained to catch some inconsistencies, like matching vendor records entered five different ways across systems, known as fuzzy matching. It cannot verify facts it was never given, like whether a delivery count logged was actually accurate.
What’s a practical first step before adopting AI tools?
Start by identifying where your data actually lives across systems like Procore, accounting software, and payroll, then check whether it’s collected the same way on every project. Consistency across projects matters more than the AI tool you choose.
How are construction companies using AI for data collection today?
Some companies now use AI agents to standardize superintendent daily logs, while others use drones and camera-based site walks to automatically collect progress photos and delivery counts instead of relying on manual reporting.
About the Hosts
Adam Cooper: President & CEO of Ascent Consulting. Adam is the primary host of Construction Hot Takes and works directly with construction company owners on operations, growth, and leadership systems.
Jeff Robertson: Vice President at Ascent Consulting. Jeff focuses on AI and technology adoption, ERP execution, and the operating detail behind fractional COO engagements with construction companies.
Greg Gorman: Principal Senior Consultant at Ascent Consulting. Greg leads the show’s conversations on job costing, WIP, and the financial discipline behind scaling a construction company.
Not Sure If Your Data Is Actually AI-Ready?
Most construction companies find out their data isn’t ready for AI only after they’ve already bought the software. Ascent Consulting helps owners find the real gaps first: in the field, in the office, and across systems.
With a 30 minute consultation from Ascent Consulting, we’ll help you figure out exactly what “AI-ready” looks like for your company.
Join our mailing list & gain more construction business insights from our experts.
Episode Transcript
[00:00] Jeff Robertson: Gentlemen, how are we?
Adam Cooper: It’s good, digging the new set.
Jeff Robertson: I see everybody got the memo on blue. I missed on the dark blue,
Greg Gorman: Apparently. Yeah.
[00:11] Greg Gorman: I’m rocking the Navy today, so good. From Old Navy, actually.
[00:15] Jeff Robertson: Actually.
[00:16] Greg Gorman: Sometimes that works out.
[00:19] Jeff Robertson: So we’ve got all three of us assembled. We’ve done a bunch of topics lately on AI. A couple of different angles. I thought we—thought that this—this go-round, we would talk about it strictly from a readiness standpoint. And specifically, there’s a couple of—there’s a couple of questions we ask our clients when they say, you know, we get these calls—all three of us—weekly now. “I think I need AI, I want it, how do I do it?” etc.
[00:45] Jeff Robertson: And we ask them a couple of very important questions. And the first question we ask—the first question I ask—is something along the lines of, “How do you like your data? How do you feel about it? Is it any good?” etc. So I thought we would kind of just play with that alone, that one only. Just that data piece for a conversation.
[01:09] Jeff Robertson: So the inspiration for this is some of you may have seen this on LinkedIn. I sent it to you guys last week. There was a meme and it had data on one side, and it was just a big pile of poop. And then you dropped AI into the middle of it—there were different examples of different types of AI you could put your data through.
[01:30] Jeff Robertson: And on the other side, basically it was just another version of poop. If it’s crap data, you get sparkly data over here after you put—
[01:41] Greg Gorman: It was it prettier, It was like colored and—
[01:42] Jeff Robertson: It had—well, there was one that was a unicorn and had a rainbow. And yes.
[01:46] Greg Gorman: Sparks—sparkles, but it was still—
[01:48] Jeff Robertson: Yeah, it was—it was sparkly poop. So anyway, that’s the inspiration for this conversation. I thought we could talk about what—what that looks like from what we’ve experienced with our clients and—and maybe some ways to help our—our listeners maybe analyze that.
[02:06] Greg Gorman: I have to add one thing to that. So I saw that you sent it to us, and then for three days I saw nothing but that little meme. And in accounting, there’s a famous phrase, right? Garbage in, garbage out. Someone tied the two things together and they literally said, “poop in, poop out,” right? I hate that word, but I said it on the podcast.
[02:28] Greg Gorman: But it’s the same concept. It’s exactly the same concept. Whatever goes into the system, into AI, if it’s bad or it’s, you know, inaccurate or it’s not thorough, whatever—it’s going to come out. It might be sparkly and pretty and rainbow and unicorn, but it’s going to be crap.
Jeff Robertson: Yeah, right.
Greg Gorman: I mean, that’s the big topic, right?
[02:47] Jeff Robertson: So let’s start talking about it from a point of just strategy, AI. Let’s take the assumption that—Adam and I had a conversation about integrations and whether is it—and I think, Greg, you and I talked about is it—is it automation or is it AI? Yeah. Let’s make the assumption that AI is a solution here. That’s our use case, we do want to leverage that. What’s—what is the first thing that we would want to go figure out? What’s the first thing you guys would think about as—is as far as data sources?
[03:08] Adam Cooper: Well, I mean—
[03:22] Adam Cooper: Assuming that you already understand what the end result is—the desired end result. So let’s make that assumption. We already know we want to get these certain reports out, we want to be able to see this kind of data visualized in a pie chart, a bar graph, something like that. So the first thing you have to say is, “Well, where does the data live?”
[03:42] Adam Cooper: Is it in different systems that we need to pull together, or is it already available inside of one system, say, or two systems? So an example we used in the last conversation is the accounting data is in Sage Intacct, and the documentation data is in Procore, and maybe the payroll data is coming from a third-party payroll source.
[04:04] Adam Cooper: But we can get that into Intacct, and then the timekeeping is coming into Intacct. If we can get the data out of those two systems—or can we get all the data that’s in Procore into the—but where are our data sources? And then we can start looking at the quality of that data. Is it consistent across projects?
[04:23] Adam Cooper: Do I—do I build my budgets the same way? Do I track progress the same way? Do I measure quality the same way? Do we have the same turnaround time on RFIs between projects? Is it 14 days on this project, but only seven days on this project? So consistency in data construction, in data collection, and in data quality is where I would start.
[04:46] Adam Cooper: That’s kind of foundational data practices. You’ve got to be measuring the same things the same way, as I think how I put it before, on the same scale.
[04:56] Jeff Robertson: Right. You and I, we have a client we were working on dashboarding. There’s a little AI piece to this, but the larger question was, even at a dashboard level, if—if not all of the project managers are using the submittal tool the same way, then your dashboard isn’t worth a damn. Forget the AI side of it and being able to analyze that later.
[05:19] Jeff Robertson: If—if—if the three of us don’t process a submittal the same way, it’s not going to help.
[05:25] Adam Cooper: Then a company-wide dashboard is going to be inaccurate. You’re going to get that garbage in, garbage out. It’s good at the project level as a standalone tool, but it’s not good at a broader—at a broader level, at that enterprise layer. So we talked about a new term that I heard from a client this morning. We talk about data lakes, data pools, data warehouses.
[05:45] Adam Cooper: They called it the enterprise layer of data, which I thought was really nice. I like that. So collecting the data from the different sources and putting them at this one layer so that you can then—that’s where you would put the AI. You would point the AI to that level—the data set, a collection of data from different sources, put it into common tables.
[06:04] Adam Cooper: And that’s what the AI would look at. So you’ve got to get the data into a consistent format.
[06:10] Greg Gorman: Yeah, because siloed data is not helpful. We all—I think that’s what we’re saying. But I would go back one step before that and say, does someone getting ready to talk about their data or talk about using AI at their business, do they even know that that’s more important than being able to do it? Because, I mean, being able to do it is a little bit of a lift sometimes.
[06:32] Greg Gorman: If you have siloed data, you have multiple systems, you don’t have a data table, you don’t have a place, you don’t have an enterprise level of data—do you have to—even do you even know that you need that? Right. Do you—and do you have the infrastructure to support putting all your data in one place?
[06:50] Adam Cooper: And I would counter that by saying in recent weeks—just in the last few weeks, and it’s mid-2026—in the recent weeks, I’ve discovered or I’ve been—it has come to my attention that—I’ll speak specifically—Claude now has plug-ins and connectors to multiple data sources.
[07:12] Jeff Robertson: That’s literally in the last two or three weeks. Yeah.
[07:13] Adam Cooper: I now can plug Claude into my Fathom video recorder, I can plug it into my Word documents, I can plug it into my templates, I can plug it into my OneDrive, I can plug it into my CRM. And so I was able to, without pulling all the data into a common location, I was able to tell Claude, “Go look at the contact record in HubSpot, go—go pull the transcripts from the meetings that I’ve had with that prospect, and then open up the Word document.
[07:40] Adam Cooper: We’re going to start writing a proposal based on the scope of work they asked me for.” I didn’t have to pull it all into a common enterprise layer. The AI could then just go grab it, and it became the enterprise layer that I needed. It became the integration, right? Which was pretty cool because I don’t have to necessarily build a data warehouse now for all of these things.
[08:00] Greg Gorman: That—that is true. It can absolutely do that. But I guess what I’m thinking of is, do you understand that you need to have Claude to do those integrations, to have the links and to, you know, to get set up? But then back to where we started, which is garbage in, garbage out. Right? If we’re talking writing a proposal, writing a narrative or writing—fine, that’s fine.
[08:22] Greg Gorman: But if we’re talking about numbers—
[08:24] Adam Cooper: No, I mean, it could be that as well. Like—
[08:26] Jeff Robertson: I’m thinking about the numbers side of things.
[08:28] Greg Gorman: It could be wrong, is my point.
[08:29] Adam Cooper: I’m thinking like, if I’m not putting data into the CRM the same way that you’re putting data into the CRM, now we’ve got disparate data, and the AI might get confused because it’s not—it’s not able to. Like, if I’m recording all the phone calls and all the meetings and putting all the emails into the CRM record that I have—
[08:48] Adam Cooper: But you’re not logging your emails with—with your prospect, then your data is not as complete as mine. And so the AI won’t be able to produce as good a result. So numbers aside, data is data. Whether it’s—whether it’s language data, numerical data, meeting transcripts, it’s all some type of data. We have to be collecting it and storing it the same way.
[09:11] Adam Cooper: We have to be recording it the same way.
[09:13] Jeff Robertson: So this kind of gets me to the point. This is the thesis—is the data and the process is the foundation here. The AI is not the solution.
[09:24] Jeff Robertson: It’s—it’s—
[09:26] Adam Cooper: A tool.
Jeff Robertson: It’s a tool.
[09:27] Adam Cooper: It’s in between.
[09:27] Jeff Robertson: The foundation is—in order for it to work, you have to have a consistent, strong, repeatable process, which we’ve beat that drum for 12 years. It doesn’t matter. We’re just applying it to a new, sexy technology that didn’t exist 12 years ago.
[09:43] Greg Gorman: Well, and I think back to that point, I think that’s right. Data is data, and numbers are numbers, and narrative is narrative, and, you know, words are words. But—but like you just said, consistent. That’s the thing. What AI is not—it’s a tool to help you, you know, do X, Y, and Z. What it’s not—it’s not going to take your data and your data and try to combine them.
[10:07] Greg Gorman: And it’s not going to know which one it’s supposed to think is right or which one is supposed to be accurate.
[10:13] Adam Cooper: You know, I would—I would counter that a little bit. Well, how? Because you—because if you train it, it can do that now. Like we talked about—I was having this conversation this morning—they talked about—we were using the term “fuzzy matches.” So we were saying if we’ve got—in this case, we’ve got an ERP system that has a vendor record, and over in Procore, which is not connected, if other people have created that same vendor five different ways—they’ve put a dash in the wrong place or a period in the wrong place, or they’ve hyphenated the name, so it’s got five different versions of it.
[10:50] Adam Cooper: And then over here in their procurement system, they’ve got seven different variations of it. That again, siloed system. You can train the AI like, “Hey, these are all the same thing. Go look for other examples where you have similar but—but not exact matches, and let me train you on which ones those are, if they are different or if they are the same.”
[11:13] Adam Cooper: And then you can train the AI to kind of do those fuzzy matches for you. So there—
[11:19] Greg Gorman: Is more of the ones that are not—that it didn’t learn is—is right or—
[11:24] Adam Cooper: Or ask. “If you’re not sure, ask me as the human and I’ll guide you.” But it’s getting better and better at being able to connect the dots. It can see the gray area and it can clean it up for you. So that’s actually possible now. So I’m pushing back a little bit just to say, like, three months ago, no. But today, kind of. And in three months from now, probably.
[11:41] Greg Gorman: Well, I guess, though, I was—I was coming at it from a—consistent—from—from a consistency standpoint to say what you’re describing is you’re training it the way that you want it to learn. You’re taking someone else’s data and putting it into your trained model. So it’s—of course it’s going to be right for you.
[12:05] Adam Cooper: Well—
[12:06] Greg Gorman: But across the whole organization—
[12:07] Adam Cooper: If we’re treating it at a company level. Well—
[12:09] Greg Gorman: If I was—
[12:10] Adam Cooper: Going to say if it’s a steering committee that is now deciding how to train it, that’s probably the—that’s probably more what I’m leaning towards.
[12:17] Greg Gorman: And I think that’s the key to what I think what we’re talking about. Yeah, AI is going to continue to get better and better and better. I think we all—we’ve seen it in real time. I think we all understand that. But I think there is a—there’s an infrastructure question here about what—what do you want your organization’s baseline to be?
[12:35] Greg Gorman: What do you want your training to be? What do you want your, you know, what enterprise-wide model, you know, as opposed to just one person doing it in one way, one person doing it—
[12:49] Adam Cooper: You know, so much about this for construction that just occurred to me, because I’ve been doing this with some of my clients, is—I’m going to use the daily log as my example, right? Instead of counting on your superintendents to populate the daily log properly day after day, we’ve actually built a little agent that prompts them with all the questions, and they just tell it the answers, and it automates and it formats the data the way it needs to be formatted consistently. So the superintendents can tell it whatever they want.
[13:26] Jeff Robertson: In whatever order they want.
[13:27] Adam Cooper: In whatever order they want, but the AI is smart enough now to build the daily log in the same format across all their projects for everybody. So now you’re using the AI to structure your data.
[13:38] Greg Gorman: The questions are the same and they’re consistent, which is exactly I think I’ve got a—
[13:40] Jeff Robertson: So here’s a—here’s a question. This goes back to this a little bit.
[13:44] Adam Cooper: Of structuring its own data for its best use case. This is pretty cool.
[13:48] Jeff Robertson: Does that mean—
[13:51] Jeff Robertson: I don’t know—I don’t know really where I’m going with this. But does that mean that maybe the “crap in, crap out” isn’t true if you could train a model to filter your data and find the crap, recognize that three different people are entering things three different ways, and in some cases, hundred—maybe thousands of people are entering it wrong or different?
[14:13] Greg Gorman: Let me tell you why.
[14:14] Jeff Robertson: Could it get cleaned up?
[14:14] Greg Gorman: It’s a great question. So if we go to the daily logs, talk about—let’s say one of the questions is, “How many loads of sand got delivered to your site today?” 50 loads got delivered. It’s going to take that. It’s going to ask you the question, “How many loads got delivered?” But if you put in 40—that’s right—
[14:36] Greg Gorman: That’s the wrong number. Sure. But it doesn’t know it was 50 unless you took a photo of it. Maybe.
[14:43] Jeff Robertson: Maybe.
[14:43] Adam Cooper: Maybe you just gave it the delivery tickets and it figures out how many loads of sand you got delivered for you.
[14:48] Greg Gorman: Well, that’s just one example, though, of where—what—what goes into the system, in spite of all of these systems—
[14:54] Jeff Robertson: Okay, I see your point.
[14:55] Adam Cooper: Its thing, process thing.
[14:56] Greg Gorman: Right. That’s when it goes back to process and in making sure the data that goes into the system and then letting it do all the things we’re talking about still has to—it’s going to do that, but the data going in still has to be accurate on a level of—let’s use truth and not truth. Is that—is that what happened or is that not?
[15:14] Adam Cooper: So here’s the other thing is I have a client right now that’s using DroneDeploy, for example. So they’ve got drones flying around their job sites. Those drones are seeing the trucks, they can count the deliveries, and you don’t even need the superintendent’s report on the metrics anymore. You can let the—you can let the AI and the technology do the counting and all that stuff.
[15:37] Adam Cooper: What you’re asking the superintendent is, you know—it doesn’t even have to ask you, for example, “Did you do a punch walk today?” You can go look in the punch list tool and see if you created any new punch list items. It can go look at the quality logs and see if you created any quality issues.
[15:52] Adam Cooper: If you’re—if you’re just using the tools in your daily routine, the AI can collect all the data for you out of those tools for the reporting purposes.
[16:01] Jeff Robertson: Purposes, which goes back to—I repeat this over and over again. In the old days when we were filling out Excel logs or counting trucks, the—the report was part of the process. You had to do the thing and then log it. A good—a well-designed system, the—the log or the report is just a—it’s just a—the product of the process.
[16:28] Jeff Robertson: If you’re—an artifact, yes. Thank you. If you’re—if you’re entering the data, then the report just happens. You literally press print or run or whatever. Or maybe it’s running automatically on whatever. What we’re saying here is there’s a couple of different use cases here. Yeah. Your point is valid that if—if your system is that you don’t have drones—
[16:50] Greg Gorman: That’s exactly right.
Jeff Robertson: You’re a $10 million company and you’ve got a guy standing at the gate going, “Check, check, check.” And then they got to go meet with the trucking rep, who’s also counting trucks for the owner, and go, “Did you get 10?” “No, I got 12.” “Shit, how are we going to deal with this?” Yep. That happens real-world every single day.
[17:10] Jeff Robertson: Right? After you argue about how many tons or how many yards are in each—
[17:13] Greg Gorman: Truck, and you talk about your kids and you’re ready to talk for the weekend, so there’s—paying attention.
[17:18] Jeff Robertson: There’s a—there’s a couple of different use cases. There’s—there’s, you know, Future Boy over here that’s got like—he’s way out. He’s thought-leading out there. Then there’s probably somewhere in the middle of, you know, there are people out there that are still counting trucks. But I have to—
[17:33] Greg Gorman: That’s—that’s what I meant by infrastructure. I think we can all probably agree that in—in some future, whether it’s tomorrow or five years from now, there will be a way to almost do everything you’re talking about. There will be a future way to have less reliance on human error or bad data. But right now, you have to have the infrastructure to have the drones, to have—
[17:57] Adam Cooper: And it’s getting cheaper and cheaper.
[17:58] Greg Gorman: It’s getting cheaper and cheaper, but you still have to have it or else you can’t use it.
[18:04] Jeff Robertson: I think there’s—that goes back to—we had a conversation this morning about this. You and I talked about this this morning. It is all about finding the—the need. What is the need? What do I want? What problem am I solving? And then you apply the tool to that. Yeah.
[18:20] Adam Cooper: So— so one thing that has occurred to me over the last couple of months and—and I’ve been reading about it and I’m reminded of it again right now—over the last, let’s call it 15 years, 20 years, construction—the level of complexity of running construction projects has gone up. There’s more data collection, there’s more forms, there’s more reporting, there’s more—there’s more paperwork involved in building a project than—than there used to be. Yeah.
[18:53] Jeff Robertson: And in some places, just because you can, unfortunately.
[18:57] Adam Cooper: Yeah, in some places because it’s required because of, you know, for whatever reason. And that complexity, the complicatedness of running construction has gone up and up and up and up. And I think we’re at a precipice now. I feel like the technology and the AI and the automations is going to start to bring the complexity down because it’s going to start taking care of the paperwork side of the business for us so we can get back to the business of actually building things.
[19:24] Adam Cooper: I think—I think—I think the pain has gotten so great. I was talking about this with clients just this morning, and they’re like—like, “We spend so much time doing the paperwork, we’re not out on the job sites walking the job sites and building the job.” And that’s why the young generation is having such a hard time learning how to build, because they’re so inundated with having to do so much paperwork.
[19:47] Adam Cooper: And we push all that onto the junior people in the office.
[19:50] Jeff Robertson: Always have.
[19:51] Adam Cooper: And I think if we could—if we could automate or use the AI to bring the complexity or the amount of time it takes to get that work done, we get them back out in the field more. And that’s how we get that time back with these—these—this younger generation, and teach them how to be builders again.
[20:07] Adam Cooper: And not just construction paperwork processors.
[20:10] Jeff Robertson: I’m all for that. You want to run for president of construction? I’ll vote for you. I think that’s the right idea.
[20:14] Greg Gorman: I think it’s a really interesting point because I’m just sitting here thinking about, if you use the classic example, right? The pyramids were built 5,000 years ago. People have been—
[20:25] Adam Cooper: That I’ll just do that.
[20:26] Greg Gorman: But people—I would love to hear that. No, probably not on this. Yeah. When—hopefully when no one’s recording you, because it’s going to be embarrassing.
[20:37] Adam Cooper: Watch Graham Hancock.
[20:38] Greg Gorman: People have been building structures that—that are—that where—like where there was nothing and now there’s something, right? The Empire State Building, right? 1931. Call it what you want.
[20:51] Jeff Robertson: Did we build those? Yes. Okay. Just wanted to check. Got it.
[20:54] Greg Gorman: Okay, so—so those were built with no laptops, no computers, everything, right? I do believe that whether it’s—
[21:02] Adam Cooper: Barely had gasoline engines back then.
[21:04] Greg Gorman: So—so it’s—it’s the more data is what we’re talking about. Yes. It’s not just the more paperwork, it’s the more data, more things can be reported on now. You can have more data points, you can have more output, you can have more inputs. You can write things down more. You can track things differently. There’s just more information to fill a bucket than we’ve ever had on—
[21:29] Adam Cooper: What’s interesting is that information is almost always, always, always been available. We just didn’t bother collecting it.
[21:35] Jeff Robertson: Yeah, I was going to say, there’s a certain matter of big tails wagging dogs.
[21:39] Greg Gorman: Here, collecting it was too high to make it valuable. Now, the cost of collecting that data is very, very, very cheap. It doesn’t mean we don’t want it or we don’t—we shouldn’t use it, just means there’s more of it and it’s cheaper to get. And so what we’re talking about on this episode is it starts and ends with that data.
[21:58] Greg Gorman: We’ve like what, quintupled the level of available cheap data that we can have on a construction project now, probably in the last—
[22:06] Adam Cooper: 20 years.
[22:07] Greg Gorman: You know. Yeah, 15, 20 years. And now there’s more of it, so it needs to be more thought of—thought of.
[22:14] Jeff Robertson: Well, it’s—considered, well, it goes back to the—how are you collecting it. So the drone idea is out, yeah, yeah. It’s an outstanding example of, do you have the means to collect it in a way that is unobtrusive? Let tech collect the data so that—so another—
[22:35] Greg Gorman: No human error —
[22:36] Jeff Robertson: Example could be— yeah, I was going to say another example could be static photos or maybe video over time, and technology analyzing your schedule progress as opposed to a human. We talked about this earlier—a superintendent walking through a job. Give him an hour, he’ll go, “I can tell you everything that’s wrong. I’m not sure how to solve all of it yet, but I can.”
[22:59] Jeff Robertson: It’s because he’s walked 10,000 jobs. Or maybe not 10,000.
[23:03] Greg Gorman: That computer is his brain.
Jeff Robertson: Right?
[23:04] Jeff Robertson: But the point—
[23:05] Adam Cooper: Is 10,000 hours.
[23:07] Jeff Robertson: But the point is, there are some things—there are some pieces of data that we can collect very cheaply, and there’s some things that maybe the tech just hasn’t caught up to being able to make it cheap yet.
[23:18] Greg Gorman: But I would say either way, that data that we collect, the data that we choose to collect, the data that we decide is important—back to the—I think that’s the thesis of the episode, right?—has to be somewhat usable. It’s got to be consistent. It’s got to be—it’s got to be usable to put into a model that you built, that you trained, in order to get output out of it, or else you’re just putting data in for the sake of data, whether it’s good or bad.
[23:46] Greg Gorman: I think now—now, I think, you know.
[23:49] Adam Cooper: I start by saying, let’s assume that we already know what outputs we want to get.
[23:55] Greg Gorman: Yeah exactly.
[23:56] Adam Cooper: Which is in mind. Start with the end in mind. And then.
[23:59] Jeff Robertson: Well, it’s interesting because you were—you were describing the state of things, an inflection point. What you drew was—and I always forget the name of this curve. It’s a bell curve, but there’s a specific name attached to the change curve. And you hit the top and you drop into the trough of disillusionment, and then you level back up and you actually start to get productive gains out of it.
[24:21] Jeff Robertson: I think that’s what you were describing.
[24:23] Greg Gorman: Trough of disillusion.
[24:24] Jeff Robertson: The top of the—
[24:25] Greg Gorman: Curve novel.
[24:26] Jeff Robertson: The top of the curve is like—under the top of the curve is named like unrealistic expectations or whatever, because you’re climbing the curve going, “Look at all this crazy shit it can do.” You get to the top. You’re like, “Well, you can do a lot of it, but some of it’s more expensive than we thought. Shit.” And then you level—then you kind of curve back up and you level out.
[24:44] Jeff Robertson: Here’s where we—here’s the middle. Like it’s a little bit like that, a little bit.
[24:48] Adam Cooper: Kind of why I drew this—this way 12 years ago. It was like, you’re going to have all these gains. There’s going to be a dip because you have to go through some pain.
[24:56] Jeff Robertson: The trough of disillusionment.
[24:57] Adam Cooper: That’s probably what that is. But then.
[24:59] Greg Gorman: It’s called back on.
[25:00] Adam Cooper: What is now. I didn’t call it that back then. I just called it the Dip. Right. But there’s always a dip before you accelerate again. But that—the stock market does that too.
[25:10] Jeff Robertson: But that was—that was what you were describing of, you know, specifically to the AI is this idea of maybe we’re starting to go down that—the backside of the curve is happening.
[25:21] Adam Cooper: It’s like things are going to start to get easier because we’re going to start letting the AI or the automations take care of some of the time-consuming, repetitive tasks that don’t need a lot of human—I don’t want to say they don’t need thought, but they’re kind of automatic. They’re just like collecting the data and putting it in a report and spinning it out.
[25:41] Adam Cooper: I think if I take up a lot of my time updating logs and this thing, if I can have that automatically done, then I get more time to spend on other things.
[25:49] Jeff Robertson: I think about it in terms of we’re finally coming to a point where we’re getting real, practical, everyday, affordable use cases for this new sexy tool that two years ago, a year ago—in some cases, we started talking about this about two years ago. It’s really gotten crazy in the last six months to a year. We’re hearing a lot about it.
[26:11] Jeff Robertson: We’re finally getting to a point where the technology—we’ve talked about this internally two years ago, ChatGPT does not do a 10th of what we can make it do today. We can—we know that internally. Yeah. So the point is that technology’s changing over time, we’re learning use cases in real time. And then you do start to go, “Well, maybe it wasn’t as sexy as I thought it was, but holy—wait a minute.”
[26:33] Jeff Robertson: “But we could use it for this.” That just continually iterates over time.
[26:37] Greg Gorman: So maybe it’s not a data readiness. Maybe it’s not an AI readiness, maybe it’s an organizational readiness. Are you ready to—back to our example—have drones take photos for you so you don’t have to have human error? Are you ready to build agents or to have someone in your company who can build agents for you to make sure your data is consistent?
[26:59] Greg Gorman: I mean, that’s a whole different level of organizational structure that we haven’t had before in the world.
[27:06] Jeff Robertson: Really find the problem. What is the problem you want to solve? I think a lot of people have fairly—go, “Well, it’s what’s called the easy button.” But it’s this—like it’s this gigantic thing I can’t get my mind around. It’s the elephant that I can’t even figure out the first bite of how and where and when. Yeah, I don’t even know where to start.
[27:24] Jeff Robertson: Yeah. So I think that’s a great approach. What is the problem I’m trying to solve? Can I use—leverage AI or technology and AI coupled together to solve it? Is—is that—can I afford that? Is there a cost to that? How much is it? What’s my ROI? What’s it going to save me? I mean, I love the counting thing.
[27:43] Jeff Robertson: I mean, nobody likes doing that and doing that for a long time, and nobody likes doing it great. Anything—anything else? I think we beat this one to death.
[27:55] Greg Gorman: No, but I would say that was—
[27:55] Jeff Robertson: That was my strong close, by the way.
[27:57] Greg Gorman: I’ve said this in a few of these episodes where we talk about AI and AI. I think there is more to say that every time we do one of these episodes, something new has either just happened or is about to happen. And in six months of—.
Adam Cooper: I love this month.
Greg Gorman: It’s exciting. It’s so exciting.
[28:12] Adam Cooper: There’s always new things to learn.
[28:14] Greg Gorman: And there’s always more to talk about.
[28:15] Adam Cooper: As I said it, it’s a rapidly changing landscape, and what we talked about today as what’s available now, what might be in the future six months from now might be available today, and what’s the next thing? It’s going to be available, right? Yeah. I mean, they’ve got the Boston Robotics, the dogs people, and those things are walking around job sites now.
[28:36] Jeff Robertson: The little—still freak me out when.
[28:37] Adam Cooper: I know.
[28:37] Greg Gorman: I know, I know they’re weird,
[28:38] Adam Cooper: But it’s creepy. But they are walking around job sites now, and they’re tracking progress. They’re taking pictures. You know, you just have to have a person walk around.
[28:47] Jeff Robertson: Their actual guard dogs, like they’re doing security.
[28:49] Adam Cooper: It’s that—what’s that service that people use? They go in and they go to the same spot in the building every time, they take all these 360-degree pictures. Because the model.
[29:01] Jeff Robertson: Yeah. Matterhorn. Matterhorn. Yeah.
[29:03] Adam Cooper: No, Matterport.
[29:04] Jeff Robertson: Matterport, thank you.
[29:04] Adam Cooper: So now they’re putting those cameras on the dogs. And the dogs walk around the whole job site.
[29:09] Jeff Robertson: Which is a much, much more efficient way to do it than having—because they were making the PIs do that at 5:00 in the morning or 10:00 at night when nobody’s there.
[29:19] Greg Gorman: I mean.
[29:19] Adam Cooper: And it’s not that expensive anymore.
[29:21] Greg Gorman: Yeah. Everything in the real world comes to construction. It’s just happening a little bit faster now, right? And Google has been driving around our neighborhoods for years, you know, like photographing every corner. This is what you’re describing. We just—can—I mean, those dogs can walk a site for 24 hours as long as they’re charged. Yeah. And that’s just going to continue to get shorter and shorter and shorter and more data.
[29:40] Adam Cooper: Yep.
[29:42] Jeff Robertson: Thank you, guys, for sitting in on this. This is a good one for you.
[29:45] Greg Gorman: Yeah, it’s fun to talk about this stuff.
[29:47] Jeff Robertson: Excellent. Good. Well, we appreciate everybody joining us and tuning in. Please like and subscribe on YouTube or wherever you find your podcasts. And we’ll see you next time.
