Key Takeaways
- Contractors use AI more every year, and the payback lags. McKinsey's 2026 cross-industry survey found nearly nine in ten companies use AI and 37% report any EBIT impact, essentially unchanged from last year.
- Mid-market contractors trail larger firms. One 2026 vendor survey puts production use at 22% for firms with $50 million to $500 million in revenue and about 45% above $500 million.
- The gap starts in the connections. Project data sits in separate apps, spreadsheets, and inboxes, and AI can't reconcile conflicting versions of the same record.
- Construction's productivity decline began around 1970, long before AI. Software closes some of it, not all.
- Firms that are catching up connect bid, estimate, project, and accounting data first. Then they point AI at one job with a person checking the output.
Ask a contractor whether they use AI and most will say yes. Ask whether it changed how a job runs and the answer gets quieter.
Use keeps climbing. The returns haven't followed. Look past the model and the causes are ordinary: a bid inbox that doesn't feed the estimate, an estimate that doesn't feed your project system, and an export someone re-keys into Excel on Friday.
Is construction really behind on AI?
Partly, and it depends on who's counting.
Houzz's 2026 survey of 601 U.S. construction and design businesses found 52% of construction firms use AI for everyday work, up 20 points in a year. The firms are all registered on Houzz, a residential platform, so the sample likely leans residential.
Production use looks different. Space AI, a construction software vendor, surveyed 487 professionals between October 2025 and March 2026. It puts production use at about 45% of firms above $500 million in revenue and 22% of mid-market firms.
Those surveys count different things in different samples. Don't read them as one trend line.
What they share is the distance between using AI and getting paid back. McKinsey's 2026 State of AI survey of 1,719 respondents found nearly nine in ten use AI in at least one function and 44% are scaling it across the enterprise. Only 37% report any EBIT impact, essentially unchanged from a year earlier. The survey spans industries, and we haven't found a construction-only return figure we'd stand behind.
The firms that do see returns look different. About 6% of McKinsey's respondents qualify as high performers, and nearly three-quarters of them are redesigning workflows around AI. About a quarter of everyone else is.
Why does AI stall on a construction project?
Start with the old problem. The Federal Reserve Bank of Richmond found that construction labor productivity fell more than 30% from 1970 to 2020, while productivity across the U.S. economy doubled. The researchers' leading explanation is land-use regulation that keeps projects and firms small, and small firms invest less in technology.
That problem is half a century old, and software won't fix all of it. We'd rather say so than sell you otherwise.
Then there's the part software can reach. Every project assembles a new team of owner, GC, subs, and suppliers, and its information lands in different apps, spreadsheets, and inboxes.
David Homola, a vice president at project software vendor PlanRadar, put the AI consequence plainly in Construction magazine: "If a project has multiple versions of the truth, AI cannot reconcile them."
The survey data points the same way. In Space AI's report, 58% of firms named data quality, fragmentation, and missing standard formats as their top barrier. Firms with a common data environment were 3.4 times more likely to report a successful AI deployment.
We hear it on sales calls too. "Our systems don't talk to each other" came up in 52% of the 215 calls we reviewed. On those calls, Procore often sits next to Sage, Vista, WinEst, and Excel with nothing connecting them.
Fragmentation also explains why tools get abandoned after launch – a topic we recently covered in length, because it’s so important to the final outcome.
Where do the gaps show up in a contractor's week?
At the handoffs. Here are four where we see AI stall, and what changes when the data is connected.
| Handoff | What happens today | What changes when it's connected | Who checks it |
|---|---|---|---|
| Bid inbox to estimate | Subcontractor bids arrive as emails and PDFs, and someone retypes them into a bid sheet | AI reads each bid into a standardized record | The estimator approves every record before it's used |
| Estimate to project system | The budget gets rebuilt by hand in the project system | The approved estimate feeds the project budget without retyping | The project manager confirms the cost code mapping |
| Project system to accounting | Costs are exported to a spreadsheet and matched to job cost by hand | Costs and commitments sync to job cost, and AI flags mismatches | The controller reviews the exceptions |
| Field photos and notes to the record | Photos stay on camera rolls and notes stay in texts | Photos and notes attach to the right job, and AI drafts the daily log | The superintendent signs off |
The first row is real. Skinder Construction kept its system of record and added an AI layer that reads incoming bid emails and turns them into standardized Quickbase records. It saves hundreds of hours a month.
Our guide on AI-powered subcontractor bid management shows how that works, with an estimator approving every record.
The other rows show the pattern. Your systems will differ.
The last row is the one most crews feel. Our piece on AI for safety reporting covers it, with a human making the call on every report.
What does catching up start with?
The foundation. We call it Structure Before AI: connect the data first, then put AI on top.
Connecting these systems is orchestration. The right data reaches the right person, in the right system, with a checkpoint where a human decides. If you run Procore, Procore and Quickbase can work together to carry that data between your project and your operations.
Then pick one job. Choose a task that eats hours every week and has a clear before and after, such as the hours spent leveling bids.
Give it an owner. Let a person check what AI produces before it reaches a client, a sub, or a crew.
Prove it, then add the next one. Our playbook for operations leaders walks through the steps, and Ten Real AI Use Cases in Construction Operations shows where they pay off.
One more thing contractors ask about, and should. Where does my pricing and subcontractor data go? Ask any vendor, us included, who can see it. It's a fair question and it should have a plain answer.
How do you know if your firm is ready?
Answer these five with a yes or a no. You don't need a consultant to do it.
- Can you say where each job's budget, cost codes, and change orders live?
- Does an approved estimate reach your project system without someone retyping it?
- Does one person own each data set, and does someone else own it when that person leaves?
- Can you name one task that eats hours every week and has a clear before-and-after measure?
- Is someone assigned to check what AI produces before it goes to a client, a sub, or the field?
Every no likely marks a connection to fix before AI will pay off. Forward this list to your PMs and your controller. Compare the answers.
Find out where your systems stop talking to each other
If you already run Quickbase or custom applications, start with what you have. Our App Checkup shows where the connections break and what to fix first, so any AI work lands on solid ground.
If you're earlier than that, a discovery call works too. We'll talk through your handoffs and tell you where AI will pay off and where it won't. No pressure, no pitch.
Book a discovery call
Frequently Asked Questions
Is construction really behind on AI?
Partly, and it depends on who's counting. Houzz's 2026 survey found 52% of its construction firms use AI for everyday work, and Space AI's 2026 survey found 22% of mid-market firms and about 45% of firms above $500 million in revenue have AI in production. McKinsey's 2026 cross-industry survey found nearly nine in ten companies use AI and 37% report any EBIT impact.
Why does AI stall on construction projects?
Project information sits in separate apps, spreadsheets, and inboxes, and AI can't reconcile conflicting versions of the same record. In Space AI's 2026 survey, 58% of firms named data quality and fragmentation as their top barrier. Connecting those systems first gives AI one version of the truth to work from.
Is construction's productivity problem a technology problem?
Only in part. Richmond Fed research finds construction labor productivity fell more than 30% from 1970 to 2020 while the overall U.S. economy's productivity doubled, and it points mainly to land-use regulation and small firm size. Software can close data gaps inside a firm. It doesn't change those larger forces.
What does a contractor need in place before AI pays off?
Connected project data, one task with a measurable before and after, and a named person who checks the AI's output before it reaches a client, subcontractor, or crew.
Where should a mid-size contractor start with AI?
With the handoff that eats the most hours each week. For many contractors that's bid intake, where subcontractor bids arrive as emails and PDFs and get retyped into a sheet. Standardize the records first, add AI to read them, and keep an estimator approving each one.
