Solving Construction Software Integration Problems: Why Your Tech Stack Doesn’t Talk to Each Other (And Where AI Actually Fits)
Construction companies end up with disconnected software not because any one system is bad, but because each tool got bought to solve a specific problem at a specific size. On this episode of Construction Hot Takes, Adam Cooper and Jeff Robertson dig into a construction software integration problem nearly every growing contractor eventually hits: a stack of five or six “good enough” systems that stop fully talking to each other.
Adam walks through a live client case: a contractor that grew from $40–50 million to $150 million and picked up Sage Intacct, ProCore, Kojo, BuildOps, and WorkMax along the way. Sales tax added in Kojo won’t carry into Sage through ProCore. Unit-cost budgets built in ProCore collapse into a lump sum once they hit Sage. The fix, after an assessment and a data-mapping exercise with every software vendor, was custom middleware: a three-to-six-month build to connect the systems at the data-table level. Jeff pressure-tests a simpler, more common version of the same problem: a QuickBooks-and-ProCore shop that won’t give up a favorite standalone daily-reporting app, where the real question isn’t technical at all, but whether people are willing to change once they see the gain.
The conversation then turns to where AI actually fits. Adam draws a hard line: moving data between systems is automation, not AI, that’s just mapping and logic, the same work Zapier has done for a decade. AI earns its place when it’s taught to evaluate data the way an experienced person would: flagging a job as at-risk from a pattern of open RFIs, or turning a veteran superintendent’s gut feel about a job site into a scoreable, trackable number. Using an API-as-electrical-plug analogy, Adam argues construction software has no “USB-C” yet: every connection is still a custom plug, and predicts that within 12 to 18 months, AI placed in the middle of two systems could do the mapping work a custom middleware build does today.
Watch the Episode
In This Episode
Why a growing contractor’s software stack becomes a disconnected “Big Mac” of tools that don’t fully talk to each other
The real difference between automation, integration, and AI — and why most “we need AI” requests are actually automation requests
How Ascent built a custom middleware solution to connect five systems that off-the-shelf connectors couldn’t
Why turning a veteran superintendent’s gut feel into scoreable data may be construction’s biggest AI opportunity
Adam’s prediction: AI could replace custom middleware entirely within the next 12 to 18 months
Episode Chapters
0:00 — Welcome, the new studio set, and today’s topic: the construction tech stack
0:51 — Why a growing contractor’s software becomes a “Big Mac” of disconnected tools
2:52 — Case study: five systems (Sage Intacct, ProCore, Kojo, BuildOps, WorkMax) that don’t fully sync
5:31 — How Ascent assessed the stack and landed on a custom middleware build
7:51 — A simpler stack problem: QuickBooks, ProCore, and a favorite standalone app
10:27 — Two kinds of disconnected data: a technical fix vs. a training and accountability fix
Why do growing construction companies end up with disconnected construction software?
Contractors buy software to solve an immediate problem: an ERP, a procurement tool, a service platform; without planning how each new system will share data with the others. Each tool works fine alone, but a construction software integration gap between them leaves manual workarounds and data nobody fully trusts.
What’s the difference between automation, integration, and AI?
Integration and automation move data between systems using mapping and logic, no intelligence required. AI comes in when you want a system to evaluate data the way an experienced person would: spotting risk patterns, flagging anomalies, and making the judgment calls a human normally makes from experience.
When does a construction company need custom middleware instead of an off-the-shelf integration?
When off-the-shelf connectors can’t move certain data, like sales tax or unit-cost detail, between systems. Custom middleware accesses the underlying data tables directly and typically takes three to six months to build, depending on the software involved.
How can a construction company capture a veteran superintendent’s experience?
By turning subjective “gut feel” into measurable, scored data. Scoring something like job-site organization on a consistent 1-to-10 scale across every project starts to replicate the pattern recognition an experienced person builds after thousands of job-site visits.
Is there a universal way to connect any two pieces of construction software?
Not yet. Every software company builds its own API, so today’s construction software integration work is a custom “plug” for each connection — similar to device connectors before USB. Some believe AI placed between two systems could eventually replace the need for a custom-built connector.
Should a construction company replace a system or keep a favorite standalone tool?
It depends on whether the standalone tool meets a real, well-understood need or is just familiar and comfortable. Before consolidating or replacing anything, ask what the standalone system does that the alternative doesn’t — sometimes the fix is education, not a new purchase.
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.
Not Sure If Your Software Is Actually Talking to Itself?
Most construction companies don’t find out their systems are disconnected until the manual workarounds start piling up. With a 30 minute consultation from Ascent Consulting, we’ll help you figure out whether the fix is a better integration, custom middleware, or simply using what you already have.
Join our mailing list & gain more construction business insights from our experts.
Episode Transcript
[0:00] Jeff Robertson: So, how are you doing? You’ve had a busy morning.
Adam Cooper: What a day already. Yeah. Loving the new set, by the way.
Jeff Robertson: It’s looking right. It’s looking good, right?
Adam Cooper: Like, this is not virtual anymore. It’s real shit right there. We just need the neon sign that’s on order.
Jeff Robertson: Right. Actually, I’m liking this a lot better.
Adam Cooper: Yeah.
Jeff Robertson: We’ve used the chairs before, but now it’s kind of starting to get tied together.
[0:22] Adam Cooper: Yeah, it feels like a real studio now.
Jeff Robertson: Um, so we got a topic for the two of us, and it has to do with tech stack.
[0:51] Adam Cooper: Yeah, you have problems you try and solve with technology, and you either buy something like a platform that has multiple tools, or you wind up buying a collection of tools.
[1:40] Jeff Robertson: So I thought we could talk through that a little bit about how do you kind of untie that knot a little bit.
[1:58] Adam Cooper: Yeah, that sounds great. I’ve got a couple of those projects going on right now.
[2:41] Adam Cooper: Yeah.
Jeff Robertson: And they’re not happy with how it’s performing.
[2:52] Adam Cooper: All right, so let’s take my one of my current clients. So they’ve got a set of softwares that they have purchased over the last few years as they grew. They started buying software when they were about 40 or 50 million, and now three years later, they’re 150 million. They’ve got Sage Intacct as their ERP that they got two years ago. They’ve got ProCore, Kojo for procurement, BuildOps for the service side of their business, and then they had WorkMax for like timecard collection. A perfect example is they create a purchase order in Kojo, and it adds the sales tax. And then it pushes that to ProCore and creates the commitment with the sales tax. But then the integration between ProCore and Sage brings across the line items, but it doesn’t bring over the sales tax.
[4:13] Adam Cooper: Doesn’t support that data field moving over. And so that’s an example of like where the technology kind of talks, but then they have to go back in manually to every commitment and add the sales tax.
[4:28] Adam Cooper: Another example is they create a unit and item-based cost-based budget in ProCore. But then when that imports into Sage, it only comes over as a lump sum. So now you don’t have the underlying data set inside of Sage.
[5:31] Adam Cooper: So for this contractor, how we started was they brought us in for an assessment. We wound up meeting with their Sage VAR, we wound up meeting with ProCore, we talked to the people at Kojo.
[6:22] Jeff Robertson: So you went through and did a data inventory and then started mapping it based on the input from the different software providers.
[6:31] Adam Cooper: Right. And then what data do you need where?
[6:51] Adam Cooper: And how do we all get it back into one place so we can report on it accurately, which we eventually want to be Sage Intacct. That ERP needs to be the source of truth. And so where we arrived was a custom middleware solution needs to be built.
[7:23] Jeff Robertson: Right. So we’re still in progress on that then?
Adam Cooper: It’s being built. Yes.
[7:31] Adam Cooper: And that’s got a three to six-month build process depending on the softwares you have and which companies are building the middleware.
[7:51] Jeff Robertson: Yeah, so let’s talk about something a little bit down the curve on QuickBooks. Maybe they have a third-party payroll timekeeping function, maybe they don’t. They’re on ProCore. And maybe there’s a daily reporting app that they really like and they’re not going to switch.
[8:30] Adam Cooper: I would start by asking them what do they like about this standalone app that makes it so special.
[8:57] Adam Cooper: I didn’t know it did that. And then, oh, we don’t need that extra tool. So now I can get it back into a consolidated platform.
[9:10] Adam Cooper: We have a client right now that we’re finishing up with, and they had their standalone safety and quality platform. And then they decided to buy the ProCore safety and quality tools.
[9:27] Jeff Robertson: I bet.
[9:28] Adam Cooper: He liked his platform and he was very comfortable with his platform.
[10:27] Jeff Robertson: So it’s a matter of working through to design a future state that doesn’t hurt as much as the current state. They imagine the change to be worse than it will be in the end.
[10:44] Jeff Robertson: So you’re stuck with siloed data. So you’re making a choice to live with the stuff that pisses you off as opposed to choose to change it, upgrade it, fix it, to get to a better place.
[11:17] Adam Cooper: Uh, I don’t know if I would wholeheartedly agree with that statement. I don’t find too many of my clients that are unwilling to go through the change if they understand the gains that they’ll get at the end.
[11:31] Adam Cooper: If they well, Tony Robbins once said that people don’t change for no reason. They either change to avoid immense amounts of pain, or to experience and acquire a lot of gain.
[11:55] Adam Cooper: So you have to be in a lot of pain to be willing to invest money to fix it, or you have to think that there’s a lot of upside that you’re willing to invest to get the return on investment.
[12:13] Jeff Robertson: You have to imagine that future state, and it has to be better.
[12:16] Adam Cooper: Or incrementally better. It has to be significantly improved for you to be willing to invest that money.
[12:32] Jeff Robertson: Nobody does that for a hobby.
Adam Cooper: But there’s an upside to getting through it. So people will make that investment of time and money to go through them to get out the other side.
[13:05] Adam Cooper: So sometimes it’s an education thing, and there’s also you bump up against the limitations of is this an easy integration or is this a difficult integration.
[13:46] Jeff Robertson: So that’s a great segue. Let’s talk a little bit about the difference between an integration and where AI may or may not sit in that spot. Make the case for me that AI is the solution for that, or disprove the case.
[14:22] Adam Cooper: Okay. I don’t think AI is the case for that. That’s not what artificial intelligence is really designed to do. What you’re talking about is automation. You’re talking about moving data from one place to another, which does not require intelligence. It just requires mapping and logic. We’ve done that for the last ten plus years with Zapier.
[15:03] Jeff Robertson: Perfect.
Adam Cooper: That’s what we’re talking about. That’s an integration. That’s what the integrations do.
[15:08] Adam Cooper: So where AI would come in, let’s say I have these five systems, maybe seven or ten systems. People call them a data lake or a data pool.
[15:33] Jeff Robertson: A data warehouse.
Adam Cooper: Yeah, the one I heard today that I loved, an enterprise layer.
[15:40] Adam Cooper: So what they were talking about is setting up some type of an automation that goes and queries all these different systems and pulls into common data tables all of the different information.
[16:22] Adam Cooper: Where AI comes in is now I want to teach it how I evaluate that data. This is how I look at the data. If the RFIs were overdue and there’s a lot of pending change orders, this project might have a little more risk, and I should probably raise a yellow flag.
[16:57] Jeff Robertson: Compare this data to that, that seemingly are not connected at all, and tell me if there’s a correlation.
Adam Cooper: Right? That’s what the AI can do, that it can replicate. If you teach it how you think and how you analyze, the AI can then start to interpret the data.
[17:34] Adam Cooper: It kind of helps us analyze the data, synthesize the data.
[17:40] Jeff Robertson: If you’re in the books of a company daily, weekly and monthly looking at different KPIs, over time, a human brain develops and understands the patterns.
[18:04] Jeff Robertson: Whereas if you have a large dump of data, doing the analysis, AI can see those patterns in a matter of minutes, seconds maybe even.
[18:30] Adam Cooper: And it can also ask it to look for patterns that you haven’t seen yet. A lot of times what companies are doing now is put in all the historical data. They just give it everything, and it’s able to synthesize and process all that data looking for all the patterns and starting to build those correlations to give you a predictive model.
[19:12] Jeff Robertson: So an example would be we had this conversation with a client, we’re building some dashboards. An experienced project director or vice president of operations has spent 25 or 30 years in this industry building things. And that person knows, because of 10,000 repetitions, that when you have a whole bunch of open RFIs on a job, that’s probably a bad thing.
[19:56] Jeff Robertson: You don’t know what’s bad about it yet. But you know it’s a flag, because you’ve just been there, that a lot of RFIs means one of two or three things: I have a PM that’s not paying attention, or I have an engineer or architect that’s not responding in a timely manner.
[20:26] Jeff Robertson: If you have a good set of data, you can have it dig down one or two layers deep in those columns to say, this is where you should probably be looking.
[21:10] Adam Cooper: I had that conversation this morning as well, talking about the builders who are aging out and how do we get that 30 or 40 years of experience in their heads and into a system that the younger generation can still use.
[21:42] Adam Cooper: How a senior person thinks, and then make sure that that is a structure that is now followed and then you’re collecting data around is the process being followed.
[22:09] Adam Cooper: Things start getting overdue. Yeah.
[22:16] Adam Cooper: I think there’s a huge opportunity to capture the way that the senior people think. A seasoned superintendent can walk onto a job site and walk around for an hour and come back and tell you if that job is running well or not, just because they’ll look at how the material is organized, how the subs are organized. You can feel it as you walk around and they’re doing some type of statistical analysis in their head, but they think of it as gut feel.
[22:49] Adam Cooper: We’re not keeping a scorecard in our head of how many points they’re earning, but in essence, that’s what we’re doing. So how do we set up some type of a collection system? What I said was you need to figure out what you’re measuring and what’s the scale you’re scoring it on. So how is the material organized on the job is a thing we could measure. One being complete disarray, ten being pristine with labels on the shelves.
[23:33] Jeff Robertson: And there’s the rub. As long as everybody’s doing it the same way, that’s where the challenge is.
Adam Cooper: That’s where the process and the system, and then that turns into culture.
[23:53] Jeff Robertson: I used to have a boss who would consistently repeat, we care about what we measure.
[24:07] Adam Cooper: One other thing I was thinking about, where I think we might be going in the future, is APIs. So APIs are the protocols that computers use to talk to different softwares. It’s your connector to plug into a piece of software.
[24:44] Adam Cooper: The API is like the door that opens up or the plug that you plug in, and it allows you to access the data tables for us. As an electrician, I always look for analogies that are germane to me, there’s different plug configurations based on European or American.
[25:08] Adam Cooper: So we were talking about you have to build a plug that’s configured to connect to the piece of software. So that’s the API.
[25:44] Jeff Robertson: Yeah, they’re gonna gatekeep it.
Adam Cooper: But there are plenty of things that you can access, and that determines what you can read and write from that.
[25:57] Jeff Robertson: So when you think about an API, I think about a plug that plugs into the piece of software and it tells you what you can and can’t see and what you can and can’t read and write.
[26:23] Adam Cooper: And that’s where the middleware shows up. So they build these off-the-shelf connectors. And what I said to somebody last week is there’s no USB, there’s no universal API plug right now. Everything is a custom plug. So it’s like 20 years ago when you’d have a lightning plug for this and an HDMI adapter for that. And we came up with the USB-A, and that allowed us to plug in keyboards and mice.
[26:51] Adam Cooper: Now we have USB-C. USB-C allows you to basically plug anything into anything. So I think eventually APIs will not be these unique configurations. I hope they’re going to start moving towards a more standard plug connector.
[27:15] Adam Cooper: The other thing I was thinking is
Jeff Robertson: I hadn’t thought about it that way. That’s a really interesting thought.
[27:18] Adam Cooper: And the other thing I was thinking about is where AI comes in is once you build the plugs, put AI in the middle, let it figure out all the mapping for you, and then instead of having to build custom middleware, you stick AI in the middle and you tell it what you want it to do and it figures out what it can and can’t do.
[27:40] Adam Cooper: I think the AI could figure that out, just like we would. We’re hiring a middleware company to go through the data mapping.
Jeff Robertson: But you’re manually designing that.
[27:48] Adam Cooper: I think the AI, I don’t know if they’re leveraging it yet, but I think that’s where we could put AI in the middle and it could figure out here’s all the things I can do. He’s basically the AI I envision in my head that just plugs into a system and starts fiddling around and then figures out how to control everything. I think that’s where we might be going maybe in the next 12 to 18 months.
[28:34] Jeff Robertson: You build an agent that is capable.
Adam Cooper: Of building an agent that’s capable of figuring out all the data tables I can get, and maybe connect that to a SQL database. And then the AI goes and pulls all the data and puts it into the table, and then runs its own analysis, and that becomes an autonomous agent AI that can run for me every night. I don’t have to build custom middleware every time, I just plug an AI in.
[29:20] Jeff Robertson: Well, this is a good topic. We covered kind of two wider ranges here today. But this is a good conversation. Thanks for listening. We appreciate it. You can find us wherever you find your podcast. Please look for us on YouTube.
