Ep 36: State of A.I. in Construction (2026 Q1 Update)

 

Join our mailing list & gain more
construction business insights from our experts.

 

 

Episode Transcript:

All right, welcome to today’s episode of Construction Hot Takes.

So, I’ve been getting so many calls and emails and RFPs and inquiries about artificial intelligence. I figured we’d just dedicate an episode to it today.

Yeah, our hotspot is popping off with all of the AI.

Everyone’s coming in through AI post, AI post, AI post. It’s all anybody’s talking about really right now. Construction is construction, but everybody’s wondering, how can I use AI? Are we ready for AI? They’re asking all these questions. So I figured we’d just put together an episode. I’ll share some of the top five trends that we’re seeing across the industry.

I know you’ve got some questions about AI from the skeptic’s point of view, like, is this really a thing? Do we need to be worried about this? Should we be looking at this? I’ll try and answer as best I can.

Everything I’m going to tell you here today is based on the best knowledge I have right now at the end of January of 2026. And that’s a crazy thing, because a month from now, everything we say could be completely different.

Yeah. Well, that’s why the goal of this is that we’re going to try and do this every quarter at least. Every three months, we’ll try and put together an update and let you know what are the hottest trends, what’s the most popular software that’s being used, if we have some names to drop.

I’ll add that we’re not affiliated with any of these software companies other than being Procore-certified consultants. So we do have some certifications from Procore, but other than that, we are just industry friends of a lot of these software companies. We have clients that use them, so we get to meet them all and we get to play with their software working with these clients that we have.

So let’s get started.

Well, you said something interesting to me earlier that I think you should point out, and that is that right now, especially because it’s happening so fast, no one is an expert in AI, right? Even the people who are building the AI themselves don’t quite know everything there is to know, right? It is changing so, so fast.

It’s our job to be two steps ahead of everybody else, which we’re trying to be, and we are. We dedicate a lot of time and resources to it, but I don’t want any of our clients to get snake oil sold, like, “Oh, these are the experts in AI,” someone claiming to sell you these things, because at the end of the day, that doesn’t exist right now.

I mean, this is such a fast-evolving technology. It’s going so fast. There are hundreds of companies out there in the construction space that are launching platforms or embedding AI into their platforms, and people are asking me, “Well, which one’s the best?” or “Which ones are full of it?” And the answer is, I don’t know for sure.

I know which ones are most popular or the ones that are most established that we can share and talk about a little bit today. The ones we’ve seen actually work. The ones that we’ve seen in action. The ones that we have clients actually using. But there’s a lot of stuff out there that we just haven’t seen yet. That’s new products coming to market.

And I don’t know that I’m qualified, or anybody is qualified, to say whether something is legit and powerful and will work well and will actually solve your problems or not. I think there’s a lot of due diligence that needs to happen with a lot of these platforms before you decide to use any of them. Demonstrations, integrations, test accounts.

There’s a whole thing I want to talk about about data integrity and what governance really means. I think that’s a valuable thing because that’s a word that’s being thrown around a lot and people don’t really understand what people mean when they’re talking about data governance. So I think there’s a lot of stuff to unpack here, and I’m going to try and do it in a palatable, digestible format of 15 to 20 minutes.

So let’s get into it.

Right. Let’s do it.

All right. I’ve got five trends I want to share with you.

The first one: copilots are becoming native in the big construction platforms. So you’re probably familiar with Copilot in Microsoft and Gemini for Google. Those are things that are embedded in the software. If you have Microsoft Word, you’ve got Copilot in the corner, and it’s an AI that can help you with your Word document, just like they have it in Excel. Or if you’re in Google Docs, you’ve got Gemini, or in your email, you’ve got Gemini to help you.

Well, they’re embedding AI into some of the ERP platforms now, like Sage. Spectrum’s got one coming out. And then of course Procore just launched Helix, their embedded AI. I was at Groundbreak back in October, November, and they showed it to us. I got to play with it. We got it turned on in our account.

And for those of you who have Procore, if you didn’t know it, it is available for free in a beta that you can opt into anywhere in North America, as I understand it. So it’s available.

I have a question about Helix.

Yes.

So you use the term “play with it,” and I feel like that’s kind of what a lot of our clients have experienced right now, is they’re just playing with the AI instead of actually figuring out how to make it work for them and do some heavy lifting. Tell me more about it. You’ve actually seen Helix do some heavy lifting and make a difference?

I’ve seen it do some light lifting.

Okay.

I haven’t had a lot of time to play with it for heavy lifting yet, but one, it can run reports. So if you just tell it what you’re looking for, it will generate the report for you. You don’t have to go in and build the report from scratch anymore.

I also created an agent in it. I did this at Groundbreak three or four months ago where I pretended I was a superintendent and I didn’t want to fill out my daily log. I just wanted to talk to my phone and have the AI convert that into my daily log. So I built an agent in a few minutes with their help that did such a thing.

I then opened it up, I talked to it, it started populating my daily log, and then it asked me questions about parts of the daily log that I hadn’t given it any data on. Like it said, “Did you have any rental equipment on site today?” So then I could tell it what kind of rental equipment I had, and then it prompted me for the next section to complete my daily log for me.

So I, as a superintendent, didn’t have to sit there typing on my phone or logging into Procore. I could just talk to it, and it did the transcription for me, and it summarized it, and it was pretty slick.

Nice.

That’s available right now. I mean, that’s a time-saver, especially for people who have superintendents who are a little reluctant to use the tech, or they struggle to get them to do it consistently. These are some of the easy lifts, easy benefits to get.

 

So these AIs are starting to become embedded as free tools, native tools, into these different platforms. So that’s one trend that we start seeing now.

Another one is document intelligence. So the AI is really good at reading and understanding documents, probably better than it is at doing your financial data analysis yet.

So there are platforms out there like Document Crunch. They’re a local company here in Atlanta, but they’re a Procore partner, and they were developed by a construction lawyer. He has trained an AI to understand contract law for construction contractors. And so it can now do the contract analysis, surface risk, and bring that stuff to the surface really easily.

It can do comparative analysis between contracts from your historical records. You can even do this with ChatGPT or Perplexity if you feed it enough construction contracts and train it, and you can use one of the free GPTs that are out there that are kind of pre-trained, but then train it on your contracts. Then it’ll help you draft or analyze subcontracts, prime contracts, all that kind of stuff.

But there are some construction-specific ones that are built, like Document Crunch, that are kind of leading the way on this front. So documents is a thing. We see it can do a lot with a little.

It can also look at plan revisions, look at old plans versus new plans, and help you surface the differences between them, like an overlay. Procore can do an overlay. There are other softwares like Bluebeam that can do overlay. Well, the AI can do that now as well. So that’s pretty powerful.

How trustworthy is that stuff, though? Have you seen it? Have you really dug into it and been like, “Wow, this output is awesome”? Or does it still require babysitting? Because I know a lot of the AI tools do still require a good bit of babysitting. It certainly does steps one, two, three right, but maybe you still have to do four, five, and six. Where’s it at?

Well, as I said, it’s there to surface the anomalies. It’s not there necessarily to resolve them. Although with Document Crunch, it can surface them, and then it can actually provide you with recommended language from a construction lawyer that you can then take to your attorney or your in-house counsel to have them finalize.

But it’s saving a lot of time is what it’s doing. It’s like, “Hey, I spotted this contract language. You may not want to sign up for that. Here’s my recommendation for how you should redline that and send it back.”

Or if your subcontractor redlines your contract to them, it can analyze what the subcontractor’s redlines were versus your original and tell you if it’s now exposing you to more risk or absolving them of something you don’t want to let them off the hook for.

So that’s what it’s really good at, helping you understand where the differences are or where the risk is so that a human can then get involved in a very targeted fashion to go resolve it. It’s not there necessarily to do it all for you, but it’s saving you those first few steps. You don’t have to sit there reading a hundred-page contract anymore.

You can have the AI read the whole thing and give you the ten areas you need to go look at so your time is more focused and valuable.

Oh, that’s helpful for sure. I think there are going to be people, too, that maybe they’re just throwing their documents or their contracts in ChatGPT or Gemini or whatever. What are some of the risks in doing that that people might not understand?

Oh, thank you for the leading question.

So I’ve had this conversation a few times recently. If you are using Perplexity or ChatGPT or one of these widely available AI tools, the most important thing to know is that if you are not on a paid account and you have not made your data private, it is using your data to train its larger language model.

At the bottom, like we use ChatGPT and we have a paid account and it’s protected. It says at the bottom of the screen, “Ascent Consulting’s data is not used to train our models.”

You have to believe that, though.

Well, I mean, it’s the best we can do, right? That’s the best you’ve got right now, is they’re supposedly complying with federal law in that way. We are trusting them to be honest, I guess, is the right word to say.

But just know that if you are in a free account, you are not paying for any of these tools. If you’re on a free account, your data is public domain. You are now sharing that out to the world through the larger language model. It’s using your data and your contract language and your documents.

I think I read something too that it’s a loophole. Were you literally surrendering your intellectual property rights when you put something into the free version of ChatGPT?

I don’t know about that. I would have to talk to a real intellectual property lawyer about that. I don’t think you can violate your own intellectual property claim just by using that tool, but consult your local or national attorney. I don’t have an answer on that one for sure.

All right, we have a lot to get through.

Number three: estimating. So this is one that we’ve talked about before. The AI is not here to replace your estimators yet.

Yet.

Yes. But what it is really good at doing is the nuts-and-bolts takeoff. It can do the counts. It can do the linear foot calculations, the square footage calculations. It can do all of that stuff and give you a very detailed bill of materials that your estimators can then use to finalize the bid.

So think about how much time is spent doing nuts-and-bolts takeoff. If you could save all that time and give that time back to your estimators to spend more time going out and getting quotes for coverage, or getting dialed-in scopes of work for your subs that are bidding, or your vendors that are supplying materials, it gives them more time to finalize the bid, which hopefully makes them better at winning.

It also gives you capacity to get more output. You could generate more bids if you had the AI doing the nuts-and-bolts takeoff for you.

I even see general contractors using this to validate or invalidate subcontractor submissions. I see subs using this to do nuts-and-bolts takeoffs on counts like light fixtures and plumbing fixtures and square feet of sheetrock and linear feet of steel stud that are needed in a building. It can do all of that really well. It’s really good at counting, and it can read plans.

There are specific softwares that have that built in.

Yeah, I was about to say, is this just ChatGPT?

No, this is a trained software. There are ones for most of the trades out there already. You should go check with your estimating software provider. And if you’re still estimating in Excel, then I recommend you go look at estimating platforms that are available now. I think you’ll be really shocked at what’s available.

All right. The other thing that they can do, what we’re leading up to with estimating, is that if you can start dumping in historical records for how projects actually went, it will start to update its database for what things actually cost to buy and install. So it’s got a feedback loop that’s starting to be built as people put more and more data in. It’s getting smarter and smarter.

Eventually, it’s going to lead up to being able to put together the draft of the whole estimate and the bid proposal for you. A human is still going to be able to validate and finalize those bids, but it’s going to simplify the estimating process. It’s going to standardize it, too.

If everybody’s using the same takeoff software, the counts are going to be equal, and now you’re competing on productivity and labor and safety ratings and things that should make a difference, not “did you miscount the number of light fixtures and toilets in the building?”

So if we could standardize the counts with AI, then we could have a little more level playing field in the estimating and bid process.

And if you’re not using AI, think about how much more competitive your competitors will be if they’re using the AI and they can generate two to three times the number of bids that you can, and they have more accuracy. So eventually, I think the whole industry is going to wind up with AI doing takeoffs.

All right. Trend number four: schedule.

So this is an interesting one. The AI cannot build a critical path method, a CPM schedule, yet to my knowledge. However, it is really good at doing schedule analysis. If you’re feeding it data about where you’re at in production versus the schedule, it can help you spot pinch points. It can start to help you put together recovery plans. It can help you adjust all that stuff.

So there are several softwares I saw at Groundbreak that were fantastic at this. I’m not going to drop any names, but there are plenty of them out there that are playing in this scheduling area, because production and time and sequencing, the AI is really good at understanding sequential activities.

So we can understand that you’re supposed to be halfway done with an eight-day activity in four days, and if you tell it you’re only at 30%, it can recalculate and redo the whole schedule and redraft it for you. So pretty slick. So take a look out for that.

And my last emerging trend right now is on the finance side.

Easy lifts that we’ve seen: there are multiple platforms like Field Materials and Kojo that can do your purchasing, your invoice matching, and even do a three-way match with a field delivery ticket if you’re collecting those. But if you’re not, at least it’s matching your purchase order with your invoice.

And it’s there to spot the anomalies. It’s not there to take humans out of it. But it takes a lot of time for accounting people to receive an invoice, pull up a purchase order, match things up line by line for price and quantity and description, and then find the anomalies. “Well, we bought it at $1.50 and they’re charging us $1.75 per foot. What’s the problem?” And then going back to the vendor.

So what the AI is able to do is look at the invoice versus the purchase order, and if everything matches, just pass it through. It looks clean. It’s ready to process. It spots the anomalies and surfaces them so that your people can focus on just going to resolve the problems with the vendors or with the purchase orders or with the invoice, whatever the problem is.

So it’s simplifying that AP process.

Also, the three-way match. Very few companies do the three-way match of getting the field delivery ticket, matching that up with the purchase order and the invoice. The AI can do it pretty much instantaneously. You issue the purchase order through the platform, somebody in the field takes a picture of the delivery ticket with their phone and uploads it to the platform, and then the invoice comes in and you drop that in, and the AI goes to work and it just tells you if everything matches, go ahead and pay the invoice.

And if there’s a problem, like a back order that wasn’t captured, like half of one line was backordered but they invoiced you for all of it, it spots it. It surfaces it. I’ve seen them where it can even prep the email to the vendor and you just hit the send button if you’ve got the vendor’s information in the system. That’s pretty slick. Big time savers. Those have been around for a year or two now.

Yeah, we’ve seen those. We’ve worked with contractors on both platforms, so I’ve seen those firsthand.

The other thing that it’s getting better at is doing the data analysis. So we talked about Procore Helix a little while ago. If you have clean data, and this is going to lead to my governance thing, if you have clean data in the system, the AI can actually do the trend analysis and generate certain reports for you.

So what do I mean by data governance?

At its most basic form, data governance is: are we collecting the same data across all of our projects in the same format and putting it in the same place?

So the example I’ve used recently was RFIs. If everybody is putting RFIs into Procore, for example, it’s the easiest one to talk about because we’re talking about Helix, then it can do RFI trend analysis.

But if some people are still sending them in emails, and some people are generating them and putting them in a spreadsheet, and not everything’s in Procore, then you do not have standardized data, and therefore any reports that come out of Procore on that topic would be skewed to just the projects that have that data.

So when we talk about data governance, again, we mean collecting the same data the same way and storing it in the same fashion so that the AI has a consistent place to look.

If you are doing that with commitments and subcontracts and purchase orders and change orders, and you’re doing them all the same way, then you have data governance, and now you can run actual reports that won’t be skewed.

The AI can only work with the data it has. There’s an old saying from the ’80s or ’90s: garbage in, garbage out. It still holds true. If you put garbage in the system, the AI is working with garbage, and you’re going to get garbage out.

Oh, I think we’ve seen that happen with the prospective clients that come to us and they want solutions now, and they want to buy this cool software and they want it to start working, when it’s like, whoa, whoa, whoa. Your data has to be the first step. You’ve got to make sure that that’s good.

I think a lot of people forget that that step needs to happen, and if anyone’s going to sell you something, “Oh, it doesn’t matter…”

The thing about the data is it starts with humans. We have to be the ones putting the data in the system for the AI to do something with it. It doesn’t just create data out of nowhere. It can generate output data, but the input data, the commitments, the RFIs, the change orders, those start with humans putting data into the system.

And if everybody’s putting the data in the same way, then the AI can really be leveraged to do a great job for you.

So those are my big five trends.

And a few highlights: integration beats innovation. We don’t need to be creating custom agents for everything right now. There are so many off-the-shelf tools that can be leveraged that you can just buy, that somebody else has taken the time to build and customize. They can then be configured for your systems, integrated, but you don’t have to be innovators at this point. You can just become users.

Well, and that’s a key part too, is having your entire organization actually adopt using it in the specific and agreed-upon way. Just that is going to set you two steps above your competitors who aren’t using it.

A lot of what we’re doing right now is helping our clients figure out what is the right tool to use and what do we need to do first to make that tool useful for you.

So this happens a lot with the AP invoicing software. Not everybody creates purchase orders in the system the same way, or not everybody’s creating purchase orders, or maybe they have a construction side and a service side and they do things differently on the two sides. So we have to come up with a way to standardize the data so that we can use the AI to process invoices.

That’s an example. Service is a lot different than construction operations, and taking into consideration those two different functions and those two different departments is where we’re finding a lot of opportunity to help our clients.

So I have a question for you that’s a little more existential, but that is: do you feel that the art and the science of construction, and actually really knowing how to build a building nuts and bolts all the way from the ground up, like you and Jeff know, that knowledge is dying? But how fast is it dying? And is it still worth having when the AI can do all the things?

Do these young kids even need to learn all of that to have the theory? Or should they just rely on the AI?

No, you can’t rely on the AI, and you won’t be able to rely on the AI for that stuff, I don’t think, in our generation, in my lifetime. There’s just too much craft to it.

What the AI is good at is the paperwork side of construction, the data side of construction. But the actual art of building things and understanding sequencing and coordinating people, because it’s people who are building these things and managing the people and organizing the teams and taking the safety precautions and ensuring the quality, we can try and automate a lot of that and we can offload a lot of that to the AI.

But right now, in my mind, the craft of building is not going to be replaced by autonomous robots with AI that know how to lay bricks, for example. It could be done, but I don’t think that that’s where the leverage is.

The leverage is in everything that goes into supporting those folks in the field that are actually doing the work: the paperwork, the purchasing, the deliveries, all the sequencing, the scheduling, the invoicing, the billing, the money collection, the finance side of the business. That’s where the leverage is.

Now the field side can benefit from that because now the data is there. The knowledge is there for them to go do their jobs. It’s on demand. They can get to it quickly and they can have a lot of confidence that that data is accurate. Like they have the most recent set of plans. They can see the RFIs were asked and answered. It’s all available now so they can go do their job with their tools. I hope that’s the way that it goes.

Well, until Skynet becomes active and Terminators show up.

Right. I did just rename my home network Skynet just to freak out the neighbors. You’ve got to have a sense of humor about these things.

All right, that’s my Q1 2026 AI update for the construction industry. I hope you enjoyed it. I hope you learned something. Please subscribe, follow us, like us, check out our YouTube. We’re up over 17,000 views on YouTube now. That’s shocking. That’s like half a basketball auditorium.

Is it shocking? I mean, you hired me to do my job well.

Shocking to me. I’m surprised that many people want to listen to us babble on about industry good things to say.

Thank you. Appreciate it.

Anyway, check us out. Apple, Spotify, YouTube. We’ll see you on the next one.