Smarter, Faster, Still Human: How AI Agents are Changing Design-Build

People usually fall into one of two camps on AI: it’s either the most exciting tool since BIM or it’s a technology to fear. In this episode of the Design-Build Delivers Podcast, we look past the hype and into the jobsite, where AI agents are already reshaping how design-build teams work.

Brought to you by ARKANCE (Learn more at arkance.us/dbia), this conversation features Brian Skripac, DBIA’s Director of Virtual Design & Construction, and Paul Hedgepath, Director of Virtual Construction at MJ Harris. Together, they share how AI is moving from speculation to application and delivering real value in scheduling, safety and specs.

From waterproofing details resolved in seconds to healthcare owners eyeing AI for facilities management, the stories highlight both the promise and the practical barriers of adoption. Along the way, our guests tackle questions of trust, integration and what it takes to bring new tools into collaborative teams.

Access all our free design-build resources and learn more about Design-Build Done Right® at dbia.org.

DBIA members are shaping the future, one successful collaboration at a time.

Chapters

• 0:00 Meet Brian and Paul
• 1:17 AI's Adoption Curve
• 3:37 Real-World Use Cases
• 8:51 Beyond the Hype
• 15:15 Agents Explained
• 27:05 Bridging Office and Field
• 33:42 Barriers and First Wins
• 39:40 The Future of AI in Design-Build
• 42:14 Wrapping it All Up

Transcript

August 2025 podfull

Fri, Aug 29, 2025 1:06PM • 42:58

SUMMARY KEYWORDS

AI adoption, predictive scheduling, BIM integration, data grid, AI agents, construction efficiency, AI in design, AI in field, AI collaboration, AI training, AI barriers, AI future, AI productivity, AI integration, AI benefits.

SPEAKERS

Erin Looney, Brian Skripac, Speaker 1

00:00

DBIA,

Erin Looney  00:08

people usually fall into one of two camps. On AI, it's either the most exciting tool since BIM, or it's the start of the robot uprising. In this episode of the design build delivers podcast brought to you by our cons. We we aren't going to make that decision for you, but we are going to talk about how AI is already actually reshaping design build today, no matter which camp you're in, I am your host. Aaron Looney from the DBIA headquarters in DC, joining me for this conversation. Are two people with a much more level head about AI than most of us. Brian skripak, DBIA is director of virtual design and construction and Paul hedgepath, Director of virtual construction at MJ Harris, together, they share how AI agents are changing the way teams work, from predictive scheduling to solving problems in the field, we are starting today's episode by taking a pulse of the industry on artificial intelligence. Are teams, firms and owners viewing AI adoption as an exciting opportunity, a necessary evolution, a source of concern, or something else entirely?

Speaker 1  01:17

Since Brian, both me and Brian, are kind of from the VDC world, I would compare this to Bim 20 years ago. So when we're starting out, I'd say 2005 and trying to get Building Information Modeling and 3d coordination on job sites. Everyone is a little hesitant, right? These are different types of methods in terms of coordinating that they're not used to, used to doing a lot of it on site. But once we got a couple jobs under our belt, and then the superintendents and the people were starting to see that benefit, then you started to get the need to have it on projects. So I see the same thing for AI, is people are still hesitant, just like they were 20 years ago with them. But then when it answers a question for you, when it solves a problem, then you start to get

Brian Skripac  02:01

more buy in. I totally agree. I see the same thing. It's in a level of excitement and apprehension. It's kind of the fear of the unknown. A lot of times, the building industry and the construction industry see things as, if it's not broke, don't fix it. Not that there's not a lot broke with with our current processes. But hey, it's just like you said, Paul, right. As soon as you show value in that, everybody's kind of head turns pretty quickly and they say, Well, wait, how did you get that information so quickly? Or how did you find that issue? Oh, I did it from Ai. Then everybody's kind of ears perk up. You're like, Hey, why aren't we doing this everywhere? And it is exactly the same as Ben. You definitely see the same cycle, so I'm not quite sure yet, but prove it to me, and then I'll jump in.

Speaker 1  02:45

And what's cool is I'm using AI with BIM right now, so I had a coordination meeting, and we had issues with a pad we're about to pour, and we had to figure out if we had the right pumps in middle for this stuff, because we have Procore LinkedIn with our AI solution, I was able to go in here, and it told me all the pumps that have been submitted, the dates and everything, with one question, right? And one question I get instantly get the answer. So I don't even have to open Procore, open other things and dig through it, and I'm not even the assistant pm dealing with the submittals. I wouldn't even know. I kind of know where to look, but it would take me a while, but because I just knew what to ask, it knew where to look, and it was able to give me that information, just small things like that that continuously happen, whether it's in BIM with a coordination meeting, or guys in the field that are able to text just like they would text me, or someone else, they can text the AI, and then they're able to get solutions based off that as well any kind of information they need.

Erin Looney  03:37

And so far, we have had aI without the arrival of a Borg cube. So I'd say it's a win. No Skynet. Everything seems to be working the way it's supposed to. The whole conversation changes when that happens. But for now, we're safe. Let's get into some of the things it can do, from predictive scheduling to safety monitoring, document automation. There are a lot of recent industry reports that are spotlighting some of these things, these use cases of AI across the AEC landscape. I read one recently from sphere partners, and another one from NetSuite, who talked about emphasizing dynamic real time scheduling. They were shifting crews. They were shifting materials based on conditions that were changing in real time. Let's get into some examples of what you've seen.

Speaker 1  04:22

You mentioned schedules. So there are schedule specific applications out there that just deal with scheduling, but we also have a scheduling agent built in with our solution that allows us to take the p6 XDR export we can also directly integrate link it with p6 which is for those unfamiliar, is a construction scheduling software that most general contractors use, and then we're able to analyze this. We have our two week scheduling meeting and analyze it to see any types of risk, and they're involved in the schedule. So how these scheduling meetings typically work is we'll have all informing around the table. We'll have our superintendents, PMS and everybody, a lot of people in the room talking. About task. But as opposed to spending that time trying to figure out that risk, why not use an agent built out with all the other agents we have in one solution to find that risk for us, and then we can take that risk and put those in agenda items up front. So talk about those things first, as opposed to it being less of a discovery meeting during the scheduling meeting. And for everything that I'm mentioning has to do with data grid, and it's a specific agents that are built within data grid. So I'm trying to keep it all in one ecosystem, right. As far as this whole AI ecosystem that has all these different purposes, but then also is separate from a management system or something that would keep you married to one thing. So the good thing about them that I like is the integrations and the connectors. They were a Connector API company until they were an AI company a couple of years ago. They connect to everything, and that's really the first step, I feel, in really getting something to work, is to have it actually talk to the things that you use, whether it's pro core or Autodesk build or whatever you're using, and then have the ability to talk to lots of different things, if you ever do switch management systems or those types of things. And then another thing that I was specifically interested in was how they use your data. Are they training their main model off your data, or are they keeping that localized within your own space? Today, maybe it's not a competitive advantage, but maybe a week from now, you might have an RFP agent or a hard bit agent that someone else is using, then their data is trained on yours. And it sounds far fetched, but if you think about where we were with AI, not just three years ago, it's not so having those methods in place now will protect us in the future to where agents are still smart, but your proprietary data stays with you. This might be,

Erin Looney  06:43

might be a bit of a stretch, but as you're saying, then, as I'm watching tools come out, what this starts to feel like, if it doesn't speak to one another and it doesn't integrate is a lot like streaming. If you were early, early adopter of streamers, like I was every streaming service that came out, I was right there. Now I have all these separate accounts, and everybody's bundling them together for about a quarter of what I'm paying for one and I'm thinking, well, but then the problem is, I lose the account that I have. It all has to start over. It sounds like these tools, like data grid that you were just talking about, and then all the other agents that you bring in are aware of the importance of avoiding the stream bloat. Yeah,

Speaker 1  07:22

that's a great way to put it, because, yeah, if you got one tool that does one thing, then you got five other tools that do different things. I mean, it's all AI that stream bloat can go to your field team, and there's already tech fatigue already nowadays, even with BIM on some project teams, some trades, still have that fatigue that may not be exposed to it, even though we're 20 years in, keeping that tech fatigue at a minimal is huge, and I think the more you can keep it in one app that doesn't marry you to a bigger app, this is management system or other things, is better. The third point that I was looking for was something that allowed you the ability to build your own agents and have an agent studio. Some things limit you, and you can only use the agents they have. We have OAC meeting packet, agent that I'm working on, and this is something that I recognize, and it's probably an issue with a lot of GCS, is you got to pull all these things from different places to do a owner, architect, contractor, meeting packet, and it takes time. So why not just make an agent that does that for you? And it's pretty easy.

Erin Looney  08:20

So Brian, let's talk to you a little here. You know you're out there going to baseball games at every park, telling us you're working on stuff, but when you are out there,

08:32

I know at the baseball game,

Erin Looney  08:36

oh, we could have a whole conversation about robots. So where, Brian, when you're talking to people, are you seeing untapped opportunities for AI, and are there any processes that design build teams should be rethinking with what you're hearing in mind?

Brian Skripac  08:51

I think data grid's a great example of one of those things there that you know, Paul's talking about the connectivity to all of these other programs as design build. It's collaboration, it's connectivity, it's transparency across those different outlets, which is key to be able to look for ways that we can access data, have that immediate response finding where there's discrepancies. Look up something about air handler too, right? Well, wait a minute, I got two different responses here. Where am I finding this. We're also looking for opportunities to be preventative about issues in the field. And I think that's that's a really big one that I'm excited about, and seeing where it goes. Whenever you hear a conversation at different conferences or something, there's different levels of where you're going to with some of these discussions. It's like, oh, chat. GPT. Bad. It's going to steal all your information, and this, that and the other thing, and it's like, I feel like people just get tunnel vision on one tool or approach and not understand the real breadth of what AI is starting to do, and the preventative opportunities, looking for gaps in schedules, looking at safety issues. On site, quality issues of comparing what's installed versus what's there. The way that things are now integrating into building information models, like Paul talked about, those are the real opportunities that people see. It's not about rewriting an email. And, you know,

Speaker 1  10:17

although it's very good at that, it is very good at that. You know, I

Erin Looney  10:20

use it for that all the time. Tone checking, I will defend it's such a frustrating thing because the M dash is a glorious piece of punctuation. And not to cut you off here, Brian, but you did make me think of a question kind of within another question, which is, are there any similar issues to that in construction? So anyone who's been paying attention to chatgpt specifically knows everyone thinks they're an expert at identifying something created by AI, and it's even more interesting how bad people actually are at it. Is there something like that Is that something the AEC industry might need to think about the

Brian Skripac  11:04

same way you talk about generative design, right? I think back to Phil Bernstein's presentation at our conference a couple years ago, where he talked about AI and design build, and he said, design me a library. You know? He's like, this is a pile of crap, right? This is garbage. It's not able to rationalize all of these different interactions of things. I think a lot of times you look at quote, unquote design images that come out of it, and they all look the same. They all look like some version of a Zaha Hadid curvy building, right? And it's not, I think, to another conference that we were at, and this was a water wastewater presentation, and there was this generative design AI based tool, where you just plug in some base criteria and it's just spitting out iterations. And you know, the presenters were like, hey, you know, this is what we're doing. This is what we're seeing. Here's an owner talking about the feedback they got in a request for a proposal where, typically they only may get two, maybe three options over a course of multiple months that people were able to crank through. And he's like, Hey, this other company came in and they gave us they gave us 15 and, you know, the person was talking, yeah, a lot of the stuff was garbage. But if I can go through and sift this and maybe I start to see things that I didn't think about, I say, hey, if I take this little nugget from over here and put it with this piece, what if this starts to do things, and the opportunity to output multiple ideas that somebody's not really able to process, and then merge those together into a really unique, high performing, sustainable solution. Is where it's at the same time, the real buzz kill at the end of the presentation, I remember this guy stood up in the back goes, you want me to stamp drawings that were done by a computer. I don't think so. I'm not putting my name on it was just like, totally took the totally took all the air out of the room, right? And everybody was getting excited about what this opportunity could be and start to look at things differently. And there's a poof, and then it was gone. So it's a balancing act, right? And we need leaders like Paul, who are out there doing it and sharing these stories and sharing these opportunities, and where they're driving value, how they're solving problems, how they're mitigating risk, how they're, you know, increasing safety and performance on the job site. That's what we need today to really get past that. And people are embracing this. Those are the really great stories and the champions who are out there helping to move this forward. One

Speaker 1  13:21

of the exciting things I think about it is the base models. So as these models keep coming out, once a model comes out and you have all your agents built out, or whatever you are using in your company, the model automatically makes it better, as opposed to say, if we're a designer waiting on a new Revit update, right? You gotta wait till next year, and then maybe they have the feature you want. Or for a general contractor, maybe it's pro core Autodesk, build all these solutions that take programmers time to do it, whereas when you build out an agent, like there's automatic video capability with one of our agents that were built out because Gemini 2.5 came out, and then you're able to walk the job site, process the video and see any type of safety updates, or just general updates from the project with the same agent because it had the different base model. So you're seeing this happen overnight, and the speed is there from all these Microsofts and these Googles making all these base models for it, but then your agent just takes advantage of that. So the sooner you're able to get and embrace this type of AI mindset and use it in your company. You're going to have an advantage because of the speed of AI in general.

Erin Looney  14:26

It's sort of like hacking your Roomba, if you have the iRobot brand, it says we encourage you to hack the

Speaker 1  14:32

room I have not hacked my Roomba. I'm going to go try though now. Oh well,

Erin Looney  14:36

when it comes to life, I did not give you the idea. I got one upstairs

Speaker 1  14:41

and downstairs. I remember, downstairs. Well, then they're going to team up putting my bed in the middle of night. We're here to overthrow you, Master,

Erin Looney  14:52

yes, thank you for the upgrade, and I'll die. Brian, you and I have talked a lot about the standpoint that AI is. Not going to take your job. It's the person who knows how to use AI who will take your job. You know, I'm talking to both of you here, but that is a conversation Brian and I have had. So what Brian does that mean for DBIA members? Think about where firms maybe should be focusing up skilling efforts.

Brian Skripac  15:15

The first thing is, you gotta do it right. You need to start small. Do something, find out what your big pain point is, and find out how AI can help address that. One of the things I was actually going to ask Paul was going back, because we're talking about this and trying to bridge the gap across multiple perspectives. We keep talking about AI agents. You know, we talked about AI in general, AI agent, Paul, do you want to talk a little bit about what that term means. You're talking about, I have multiple agents that are doing this, and it's people might think I have aI doing this. So before I answer that, I think it might be good for Paul to me mention that just raise awareness to that terminology as well.

Speaker 1  15:54

When we talk about agents, we're talking about specific thing that's a tool for a specific thing. There's also AI agency, right? So working on behalf of someone now, you can do automations with these agents to where it'll work on a schedule and do that type of thing, as opposed to it, it kind of going off on its own and doing its own thing without your instruction. That's more of a, in my opinion, like a Artificial General Intelligence type of thing that's not really around here yet. AGI is a term that a lot of people use, really, I think we're still in a narrow AIS, what they would call that world right now, to where you have aI specific things, or tool specific AIs. I like the scheduling agent, those types of things. So people, when they say agents, sometimes they'll talk about it in different ways, right? An agent like I talked about, or agency and working on behalf, which is partially right with automations, as you can do in data grid and that

Brian Skripac  16:44

kind of stuff. We just want to make sure that our members understand what it is, and you know something that they're hearing and being able to understand. So I think that's that's a big deal,

Erin Looney  16:53

and we're trying to avoid the uncanny valley. Tell people we are not in the uncanny valley at this point in the construction industry, you said artificial general intelligence, which they just did the update to chat GPT recently, that will almost do things for you. It's a little creepy, and people immediately, oh, no, I don't want it paying my bills. It's going to steal my money. No, it isn't going to do that, because it isn't at that level yet. And I think using that comparison of, you know, we've created a villain out of chat GPT, and I think that helps understanding what it can't do. Might be able to bridge to what you all are talking about, specific to construction.

Speaker 1  17:30

Yeah, you're right, because it is in the end. I mean, the agents that we're using are based on the models like GPT. I mean, you can use GPT four, oh, and other ones with the agents I'm using, you just you can pick all the agents, whichever one you want. So you're right. It's the same thing. It's just with the software that I'm using, it allows you to have the direct connectors into the programs that we use every day, and then you can build the agent on top of that, within that program that utilizes those connectors. If you try to do the same thing, say, with a chat GPT, you're going to be limited in what you can do, you can't dump a whole set of drawings into chat GPT, and it can't analyze it all. There's a limit to a lot of stuff to whereas with the data grid, or, you know, the software that we're using, you're able to do the connectors. That's another kind of difference between the two. But when you think about AI in general, you know, the base models and things are the same

Brian Skripac  18:18

for DBIA, this is one of the things that we want to continue to explore, right? We just did a blog post on it. You know, we've had this as an opportunity at the the VDC Leadership Exchange. We've had presentations on it the last couple years. We're going to have other companies that will be talking about AI at the expo, at the annual conference. We want to keep bringing this information to our membership and, you know, using the networks and the colleagues and the peers that we have out in the industry like Paul to help spread the word and educate this is another opportunity to collaborate and be more efficient. I think that's a big thing, right? Productivity and efficiency on the job, access to information, continuing to find better ways you can, kind of go back to the BIM analogy, right? This was, hey, how can we use this technology to better understand the buildings we're designing, constructing the roads, whatever they are, you know, how do we do a better job at accessing the amount of information we have on projects in disconnected, dispersed locations and silos? You know? This now starts to bring it all back together. We can access information and understand it and translate it and communicate more effectively. So just continuing to make us all better at the work that we're doing.

Speaker 1  19:26

And what I'm excited about is the designers using AI or starting to use it. And what are those use cases that are out there? I'd be if anybody wants to reach out to me on LinkedIn or email, you know whether you're using it during conceptual design, design development, or even during the CA process and how that might work out for you is, I think there's a lot of opportunity, because there's a lot of information that you have to manage. Sometimes your design cycles are less than two years, right? And you're going through different iterations of different things and a lot of stuff back and forth, just like we have in construction. So I think there's huge opportunity in the. Design world for this, and I'm excited to see where it goes.

Erin Looney  20:03

The design build delivers podcast is brought to you by Archons, your technology optimization partner, helping design build teams streamline workflows with Autodesk solutions, expert support and real world training. Ready to work smarter and faster, get tools insights and schedule your AI readiness check at our cons.us/dbia.

Erin Looney  20:31

So you mentioned LinkedIn, and now you're going to get a bunch of requests from everyone who listens to this show, other other people can click on your account like I did you in a LinkedIn Post recently, you described choosing an AI company as more like hiring a teammate, and I'll be honest with you, that's how DBIA is comms team. That's how we think of our chat GBT account. Our chat GPT is called Charlie. He is kind of like an intern, useful. Speeds things up, but he doesn't have the depth and knowledge and commitment to DBIA or our members that a human teammate has. So by comparison, how did MJ Harris approach that mindset shift of understanding it as a teammate and not as a tool, but without getting into that creepy territory of calling it by name, like us, we

21:20

do name our agents. So,

Erin Looney  21:22

oh, good. We're not the only

Speaker 1  21:24

I'll call the scheduling agent speedy or something. Yeah, I say that because you have to really understand the instructions of the agent and what you're telling it to do. When you're doing instructions of agent, you're almost giving it a job description, like you would a person in the HR department. So we're telling the scheduling agent, Hey, you are a scheduling expert. You're going to look for these types of things. And this is all text based, just written out. And then you can also tell the AI to write the instructions for you. And so it's like giving itself instructions based on your description, and then it makes itself even better. And so it's that's kind of weird, but that's what it's doing. And then so you're training this thing to work on your behalf, to some extent, as a expert in something. So that's where I kind of relate that whole HR partnership type of thing, and then going more wide and saying, yeah, in terms of the company that you're working with, it's doing the agents and everything you want to make sure that it is a good relationship, because it's more than just saying, Yeah, I'm going to use Revit instead of ArchiCAD, or I'm going to go use this. You're actually trusting a company to help your entire company and your entire process and really make everyone better in your company. And every process you have not just modeling or not just scheduling everything when you're talking about that, it really does matter the type of AI company that you partner with, and making sure you know the security is in place, making sure that they're not training on your data, all those checklists that go in place to make sure it's a good fit, right, just like you would with any employee that you hire, because it really is acting more like a employee than it is a piece of software.

Brian Skripac  23:02

I don't know, you know, how much you're seeing this, like programs like test fit, from a large scale planning exercise. I think we all think on the design side, we think more about, you know, the visualization aspects of it. But I think that's another one that you start to look at from large space planning efforts are getting pretty significant and saying, Hey, I need this many departments on a spot this size, and the ability to start dragging things around and having all of these complex configurations and spatial relationships being done in the background and laying out an optimized solution is pretty powerful. Yeah,

Speaker 1  23:37

it's amazing just what that one product alone is able to do. I mean, I used to work for an architect too, in late 90s and just programming phase. You probably remember this too, Brian. I mean, you're just coming out with different schematic design layouts, blocks with colors, diagrams, yeah, or they actually use markers and color the different ones, yeah, the diagrams. And now you just pull this thing around, doing whatever you want. It's giving you dynamic feedback, which is amazing,

Brian Skripac  24:03

that increase in productivity right there, because by time you lay everything out and copy it around and move, okay, there's one. And then you do the back in the CAD days. It was saved as option two, yeah, redo everything. Option three, final.

Erin Looney  24:21

Final, final do not change. So my dad actually in this, as you were talking about that, it made me think of he first he was an engineer, and at some point, just arbitrarily, he said, You need to learn how to use AutoCAD. And I was like, for what? And so he really made it his mission to think this was some kind of universally applicable tool that was going to follow me through life, and Little did you know,

24:45

today, you're here talking about 2025,

Erin Looney  24:50

Brian, how does delivery method, or does delivery method affect a team's ability to bring in AI in any way? Are there characteristics that help. Or hinder adoption depending on your delivery method.

Brian Skripac  25:02

Well, I think the individual team members can use things much in the same way that Paul talked about, but I'll relate it again, back to them, right? If it's a hard bid scenario, you know, Paul gets on the job, is he going to have access to any of that past information that was on that traditional the architect and engineers contract? So no, he's not going to be able to go through that information and see it, or he's just going to have the drawings and now he's got to feed them in there. And there's still a level of historical knowledge and information and decision making and things that the builder isn't going to have access to, just based on those contractual silos that are there. So I think the individual team members still have the opportunity to do things on their own, but there is another level of collaboration once you open things up and you're in that larger team focus moving forward, like you would see in design build or progressive design build job, and having all those team members there at the beginning to be able to share that information access, it is a different scenario.

Speaker 1  25:58

I think the design build scenario would lend itself to more collaboration among the teams earlier on, I think the sooner, the better. I mean, with, even with BIM, like you said, being able to collaborate and talk about, you know, maybe the GCS being brought in at DD, or, I guess with design build even earlier, the longer we are, the designers and the builders are teamed up, the better, and the more that lends us to longer use of AI?

Erin Looney  26:23

A lot of what we've talked about today has hinted at this question, more or less, for better or worse, a common theme in specifically AI conversations in the AEC industry is its potential to overcome everyday inefficiencies that drain time and drain talent. That is universal outside of the industry, but I'm bringing this back to LinkedIn, stalking you again, Paul, you also talked in one of your posts about the absurdity of paying highly skilled professionals to spend hours searching for one spec in a sea of PDFs. So from your vantage point, how are AI agents solving that problem? And more broadly, how should the industry be thinking about these tools as this type of productivity multiplier? We

Speaker 1  27:05

have a agent called Deep Search, and that's exactly what it does. It just finds stuff. It's probably the most commonly used one. So the ability to find something instantly is pretty substantial, and then able to find it in the field, because we have questions all day long, because we're the experts at building, but we have to build per the set of drawings. The drawing specs are the contract, right? So we're having to build and adhere to that. But also you have a whole bunch of asis and design changes that come out all throughout the project. Maybe you have over 100 RFIs or more coming out to answer things that we find as we go through this process of being an expert builder and then asking the design team, you know, Hey, these are the things we think need to change. All that has to be tracked. And there's all these different members that are all involved in it that may not have exactly written the RFI. They may have not have been exactly notified about the ASI. Everyone is notified, but maybe the guy actually putting it in didn't get the email. And there's all these different types of ways that people can miss information once we've gone from, say, server based and siloed management systems like having prolog, then the architect had new forma, then the owner had something else. We're getting closer by that by using things like pro core build and other stuff. But AI is gonna take us to the next level, to where, not only do we have centralized information, still a whole bunch of information, but now we have centralized information that everyone can get to easily, and it will actually answer questions specific to the problem that you're having, as opposed to you being the librarian and trying to find something now you're actually the construction expert getting the information You need when you need it to where you can move on and add higher quality tasks to other things. When it gives you that information, it's not only giving you information, it's actually citing where it came from and showing that exact document that it pulled it from, to where you can not only get it, but you can trust it. There's no hallucinations or anything like you would have in chatgpt when you're dealing with a construction specific, a genetic AI solution, because it's going to pull exactly what it is that is getting the answer from

Erin Looney  29:08

librarians everywhere are wincing it. You basically just making them obsolete. But, like,

Speaker 1  29:15

librarians are still around. But, you know, they had the Dewey Decimal System and they had a little card, I remember the card things and the flipping catalog, yeah, the card catalog,

Erin Looney  29:25

the cabinets are people's furniture now. And like, Oh, look at this cool thing I have. And if you ask somebody, what is that they they don't know a lot of times. Brian, are you seeing some of the same things Paul just talked about in your conversations?

Brian Skripac  29:39

Yeah, we are. This idea of being able to focus on high value tasks and using technology to do the more mundane aspects and find information is a big issue. It's a it's a challenge that we have, listening to some of the things that Paul's mentioning and thinking back to where we started with this. Well, when we were back in the BIM days, right? We've. Ever started this like, how would a project manager, you know, God forbid they open up a Revit. I don't want to open the Revit model. I don't want to open the Navisworks model. I don't want to go in there and try to find anything I'm going to write spend 10 minutes typing an email to send to somebody to go do something. Now they're just opening something, typing a question, getting all that information immediately, enough to find out what sheet it is and find out what folder it's in. They don't find out what bid package it was from. It's all right there. And you know, thinking about the time to get an answer and answer a question out on the site, whether you're an architect, an engineer, a builder, a trade contract, or whoever, it's just right there. It's accessibility, it's speed, and it's accuracy.

Erin Looney  30:40

This is not a completely new question. This has been a theme throughout our conversation, but let's focus a little on how AI powered assistants do help bridge gaps between the office and the field. Maybe you've got a real world example where an AI tool solved a coordination issue before it became a bigger problem.

Speaker 1  30:58

One example would be waterproofing that you have around have to have around, have to have around the edge of a concrete wall, and then whether tape has to be on that corner or not. So this is a question whether or not a spec required, whether or not it's a requirement in submittal. You're out there in the field, you text the field agent, and it tells you yes, there needs to be tape required. It not only pulls the spec up, it shows you the location, but also pulls the submittal up from that waterproofing and it shows a cut sheet with a piece of tape on it. You can see, instead of walking in the trailer and spending 30 minutes trying to find it or asking an assistant, pm or somebody to dig it up for you, with our field guys, most of their value to us is going to be in the field, managing those guys and making sure that there's the highest quality work being performed and using their expert constructability brains out there all the time make sure everything needs to be happening. So if they're able to get something answered and not be that librarian, then it's more value to the next thing, because they get 100 problems a day. That's what they do. They solve problems all day long.

Brian Skripac  31:54

Paul, have you seen, you know, the same way that you're talking about a field. I'm thinking, you do a lot of healthcare projects, right? Your healthcare labs, yeah, we're 80% healthcare. So the owners and our operations teams. Are you seeing them starting to utilize this? And they gotta keep equipment up and running. It seems like a natural fit for them to start extending this information. If there's a piece of equipment that's down, they gotta get it back up and running, because the research department might be down or something else isn't functioning. You know, is that the same opportunity for them?

Speaker 1  32:25

I think it is. I've had some owners that have started to ask me about it, that are looking at it, some big, big healthcare owners. They see the value, not only in operations, because you think about it, we're building, we're designing and building this thing in four years, and then these guys are maintaining it for the next 30 so just like you would have any type of Facilities Maintenance Solution, you want to have the same type of solution maintaining all your data. And why not use agent, AI to do that? But at the same point, what I've heard from owners, and what they said to me is actually managing the design process too, with it, as they're hiring these architects early on, using AI to be on the same page about program management, about the changes that are happening codes. There's a lot of things that they're concerned about early on during the development of a hospital, and that just carries on throughout the project, during construction. And that's another value of having, if you're an owner, having something that works with everything right, works with all different types of softwares, because most owners don't require a GC to use. Sometimes they do, but they're kind of getting away from that a good bit to where we can all use different management systems, or an architect can use a Revit or whatever else they're using, but having something that connects into all that, that comes into a main database, that allows you to manage that stuff through an AI is is important to an owner.

Erin Looney  33:42

So we've talked a lot about the good things for the most part. Today, let's look at barriers. Now let's look at things like data silos, workforce training gaps, integration challenges. This is the type of stuff that's consistently cited in the articles that I've been reading about. You know, the biggest obstacles to effective AI adoption? So both of you different perspective. Paul, what kinds of groundwork, whether infrastructure or culture based, are you seeing and what needs to be laid before these agents and AI in construction can thrive?

Speaker 1  34:12

You have to solve a problem for somebody, and then you have to have it all set up and try it yourself before you just go out on job site, or go to your design team or whatever, and say, Hey, try this, because it has to have the right instructions and the right data connected. So there is a little bit of setup. It's not like a massive amount of setup, like me trying to learn Revit when I went from ArchiCAD to Revit 2007 that took a little while. It's not like that. So it's really couple weeks you make sure that your agents are the way you need them to be. Get with whatever provider you're using, you know, from from my case, it was data grid and getting with those guys to make sure that it's giving us the output that we want. Because the last thing you want to do is get in front of somebody and ask you something and then it tells you something that you didn't expect, not that that's going to happen a ton if you have your agent program, right? But if you don't. It will. It can only pull in the data and give you what you're asking it to give you. So just having a good understanding of that, because, along with tech fatigue, specifically with AI, you can have lack of trust in the AI. It's so popular in culture right now that people are hesitant to go all in and saying, Oh, this thing can't build our building, or this, you know, just like what Brian was saying when the guy stood up talking about digital plans. I'm not going to stamp that right. It's not a big lift. It's just something you gotta make sure you do. Just don't go in head first, not knowing what you're doing and getting in front of somebody with it. It

Erin Looney  35:36

goes back a little to what you were saying about hiring a teammate. You wouldn't hire somebody without really thinking through this person needs to be able to give us these things. And it's similar in a lot of ways to that. Brian, are you seeing the same type of roadblocks and obstacles and conversations when you talk to DBIA members or prospective members?

Brian Skripac  35:57

Yeah, I think it's the same, right? And again, finding these commonalities of other things is people, process technology. What challenge are you trying to solve? Adding trust is a big issue, right? And that's, I think, one of the things you get into. We talk about that all the time at DBIA, with trust, but you can't just throw software at something and expect it to fix it. Like Paul said, what challenge are we trying to solve? Is there an opportunity to use an AI agent to solve an issue that we consistently have on projects, build the process around it and implement it. That's how you make things successful. It doesn't matter if it's AI, it doesn't matter if it's BIM, it doesn't matter if it's a scheduling software or whatever. You have to know why you're doing something and apply it. And I mean, I think just one of the big reasons we always say, you know, it always comes back to people, process and technology. I think the other hurdle is, is, again, just the apprehension right? People need to see value in what you're doing. And I think anytime we can show an ROI to a task, that's the biggest thing to making the pessimists get on your side. Hey, we used to, used to take us this long to do it. I just did it in this time frame, our documents aren't getting any smaller, our drawing packages aren't getting less, the specs aren't getting smaller. We're getting more and more and more data and information. You know, we always talked about data structure, too, and now AI starting allow us to not worry as much about data structure, right? Because now we have this ultimate access to everything that's there to be able to go out and search and find it. Technology keeps evolving and innovating and helping us get past the challenges that we've created for ourselves. We went from drawing to digital, digital to BIM, and there was more and more information, and now we got to focus on data structure to get it there. Now al lets us sift through everything. You got to trust and innovate move forward.

Erin Looney  37:43

Yeah, let's say a design build team is just beginning their AI journey. They want to be part of this. What is one early, low risk win they could pursue to sort of build that comfort and confidence without needing a full scale transformation. Maybe they're not ready for the Borg cube.

Speaker 1  38:00

I think just finding stuff, right? Search and Find and that's something you can do, and connecting with the documents that you have and search through things, I think, for design build, from a designer standpoint, just looking for commonality between specs right when you go into the next job, and that kind of stuff, from a construction standpoint, just reviewing documents and being able to find things quickly when getting acquainted with a project quicker, but asking it thanks. Tell me about the plumbing requirements on this job. Tell me about this. And just like you would sit down and ask an expert that's been on our project for a year, and then you're you're the new person on the project, if you just do the same with AI and just ask you questions, that's really kind of the first rung, and there's ways to do it. Even with the solution we have, there's ways to do that. There's not a big lift. And anybody that can do that. That's one thing I like, is they make it easy, easy access for anyone to really start, I think just using AI in your everyday life too, whether it's GPT four, oh, on your phone, and just asking this stuff in general, and then you'll find that it's giving you a lot of information just about daily life, to where you get used to using AI, and then you can take it and then use that project specific AI tool as well. And you're like, oh, it kind of relates, right? Because you're kind of getting in the habit of using it

Erin Looney  39:13

sort of builds the credibility as well. Yeah, yeah. Final question, this is our speculation question. If we fast forward five years, what does an AI, empowered design build project look like now? To kind of set it up, there was an article that framed 2025, as a turning point for agentic AI, which you've talked about, tools that behave more like collaborators and software. So what does that look like in five years? It'll be,

Speaker 1  39:40

man, it's hard to tell with AI. I don't know. I'm not gonna say dream scenario AGI, but I don't know if that's that's happening or not in five years. It could at the minimal I think you're going to have agents to where you don't even have to give it instructions. It knows there'll be an. AB list basis for construction, as well as architecture and every profession in our world of design, build, construct kind of stuff to where we can just pick from it. And you can kind of do that now with marketplace and some of those things. But everything you can imagine, there'll be an agent for already built out that works perfectly, and then you'll just plug it in, and there'll already be workflows in place, just like there are for BIM. Everybody kind of knows the coordination workflow now, right? Or they know that how we do design documents in Revit, because we've done that over 20 years. But with AI, 20 years is five years, whether or not we're at the AGI point by then or not, I don't know, but I think there won't be any questions. There won't be any proving anymore. It'll just be something that'll be commonplace, that the early adopters will have an advantage because they'll have more of a AI infrastructure built out within their company, but everyone will be in agreement that it is something that everybody needs to use in our industry. Do you think we'll

Brian Skripac  40:56

be able to get to the point where, and I'll take this back to the modeling side too. When you put that door there, can't be there because it's a code violation. You can't put that toilet there because it doesn't meet. Ada, I think

Speaker 1  41:07

the information part is going to be first. What we're doing, what we're kind of dealing with right now, is processing information and schedules and words and documents. The relationship between things is going to be next. And not only don't put that door there in the design phase, but I see it going as far as with the pipe fitter out there in the field with AR, this is getting with ar, ar safety glasses. It'll be cheap. It'll be connected to the AI that'll know everything about every code in the world, and this has to do with your jurisdiction, and knows everything about that project to say, don't do that fitting that way, or don't do that waterproofing that way. It'll get to that point too, to where you have aI helping us with design, with codes and AI helping us with installation and means and methods. Five years, I don't know, but maybe

Brian Skripac  41:56

cool think about, Yeah, it's cool to think about.

Erin Looney  42:01

There are people out there right now going, is it?

Speaker 1  42:04

Is it? No, you're not going to be have a job. Don't worry, it'll be fine. It's just it's gonna make us more efficient.

Erin Looney  42:14

So AI is not here to take your job, but the person who knows how to use it, or at least the person who can hack their Roomba without triggering a takeover just might. And judging by what Paul and Brian said today, it doesn't matter if you're pro AI, anti AI or unsure, because it's already here on our job sites. Thank you to Brian and Paul for sharing their insights, and to you for listening, of course, and huge thanks, as always, to Fred Yee for his technical expertise in making this podcast possible. I am Aaron Looney, and this is the design build delivers podcast brought to you by our cons. Learn more at our cons.us/dbia. You.