TL;DR: Contractors have largely decided AI matters. Dodge Construction Network found 85% expect it to cut time on repetitive tasks, while only 40% had put a budget behind it. The interesting number is not adoption or enthusiasm. It is that 47% of firms now report difficulty hiring AI specialists, up from 30% a year earlier.
The one-sentence answer: AI in construction today is mostly document handling, estimating support and pattern-finding in data you already have, and the barrier is rarely the technology.
Ask ten contractors about AI and you get ten answers about robots.
The actual deployments are duller and more useful than that, and the survey data now covers them well enough to be specific rather than speculative. So here is what the numbers say, what is genuinely running on jobs, and where the money is going.
What do contractors actually believe about AI?
That it will matter, nearly unanimously, and that it will not take their people.
The Associated General Contractors of America and NCCER surveyed roughly 1,400 firms for their 2025 Workforce Survey. Asked about robotics and artificial intelligence affecting construction jobs over the next five years, 45% expected a positive effect through automating manual, error-prone tasks, and 44% expected better job quality with safer, more productive workers. About a third expected no significant effect at all.
Only 12% expected AI to eliminate jobs. That figure was 14% in 2024 and 17% in 2023.
Read that trend rather than the single number. The people closest to the work have grown less worried about displacement every year for three years, at exactly the point when the technology got dramatically more capable. That is not what you would predict from the general coverage.
What are contractors actually doing about it?
Considerably less than they believe, which is the central fact of this whole topic.
Dodge Construction Network, working with CMiC, surveyed 235 general and trade contractors in September and October 2025. It found 85% expected AI to reduce time spent on repetitive tasks and 75% expected it to help them learn from historical project data. Against that, 51% were actively evaluating changes and 40% had allocated a dedicated AI budget.
Separately, the AGC and Sage Construction Hiring and Business Outlook, published 5 February 2025 from more than 1,100 contractors, found AI was the leading category for increased technology spending, with 44% planning to raise AI investment that year.
So: near-universal belief, roughly four in ten committing money. That gap is not stupidity. It is a rational response to not knowing which specific thing to buy, which is what most of the remaining questions in this series are about.
What is AI actually doing on real jobs today?
Reading documents, drafting things a human then corrects, and noticing patterns in data you already collect.
That is less exciting than the marketing and it is where the value currently sits. Concretely, the categories that show up repeatedly in contractor deployments are document extraction, pulling numbers off bills and submittals rather than retyping them; estimating support, producing a first-pass quantity or price that a human adjusts; photo handling, sorting and labelling site images so they can be found later; and summarisation, turning a long thread or a day of field notes into something a project manager can read.
We build these into our own platform, so this is not a survey observation. By early June 2026 our field operations product had shipped AI assistance in four places: quote drafting, photo handling, project insights, and change order preparation. Each one drafts; a person approves.
The most useful thing we learned building them has nothing to do with model quality. It is that the dominant failure mode is the model not answering in time. We wrote a threat model specifically for AI timeouts and built a dedicated test for it, because a superintendent standing on a roof with a phone does not care why the request hung. Nobody markets timeout handling. It is most of the engineering.
Which AI is worth attention and which is noise?
The test is whether it touches a document, a decision you already make repeatedly, or a pile of data you already own.
Those three categories work because the AI is doing something narrow with material that already exists in a fixed form. A submittal has a structure. A bill has line items. Your last two hundred jobs have costs.
Be more sceptical when a tool promises to predict something with no history behind it, replaces a judgment your best person makes on instinct, or requires you to change how the crew works before it produces anything. That last one is the reliable tell. Software that needs behaviour change before it delivers value usually never delivers value, because the behaviour change does not happen on a live job.
Why does firm size change the answer so much?
Because the same price is a different decision at different volumes, and the survey data shows it clearly.
Dodge found 86% of large contractors believed AI would give them a competitive advantage, against 69% of small and mid-sized firms. On cost, 49% of smaller firms called the investment a major concern, against 26% of large ones.
Nearly double the cost anxiety at the small end. That is the real adoption story, and it is why the "everyone is doing this" framing in vendor material tends to describe enterprise contractors.
If you run a smaller operation, the practical implication is to buy narrow. One workflow, one measurable outcome, cancellable. The general-purpose platform sale is priced for someone else.
What are contractors most worried about, and are they right?
Reliability and data security, in that order, and yes on both counts.
Dodge found 57% citing concerns about reliability and accuracy, and 54% citing data security and privacy risk.
Both are correct concerns rather than resistance. Reliability matters because a tool that is right most of the time still needs every output checked, and checking can cost more than doing. The question to ask a vendor is not whether it is accurate but what happens when it is wrong, and who notices.
Security matters because construction data is more sensitive than it looks. Drawings, contracts, subcontractor rates and client details all sit in the same systems, and the answer to "where does our data go" should be a specific one you can read.
Is AI going to solve the labour shortage?
No, and the same survey shows why it may be adding a new one.
The AGC workforce data is not ambiguous about the shortage. 88% of firms that employ craft workers had openings. 57% said available candidates were not qualified. Worker shortages were the most commonly cited cause of project delays, affecting 45% of respondents, though notably that figure has fallen each year from 66% in 2022 through 61% in 2023 and 54% in 2024.
Now the part that gets no coverage. Among salaried roles, difficulty filling AI personnel and specialist positions jumped to 47% of respondents from 30% in 2024. That was the largest single-year increase in the survey.
So the sector is short of electricians and now also short of the people who would deploy the technology meant to compensate. Anyone selling AI as a workforce fix is describing a solution that requires a scarcer worker than the one you cannot hire.
The realistic version is narrower and still worth having: AI reduces administrative load on the people you already employ, which is not nothing when a project manager spends their evening on paperwork.
What should you actually do next?
Pick the single most repetitive document task in your office and price fixing only that.
Not a platform. Not a strategy. One task, where you can state the current cost in hours per week and would notice the difference. Estimating support and document extraction are the usual candidates because the before-and-after is measurable.
Then insist on three things before signing: what happens when the output is wrong, where your data goes and who can read it, and what it costs to stop. A vendor who answers all three plainly is worth continuing with.
The rest of this series takes those questions one at a time, starting with how to evaluate a vendor and what the technology genuinely costs.
Frequently Asked Questions
Do I need to hire someone technical to use AI tools?
For narrow tools bought as software, no. For anything requiring integration with your existing systems, you need someone accountable for it, and that is exactly the role 47% of surveyed firms reported struggling to fill. If you cannot hire it, the practical alternative is buying tools that work without integration, or contracting the integration rather than staffing it.
Will AI make my estimates more accurate or just faster?
Faster first, and more consistent before more accurate. AI is good at applying the same logic to the hundredth bid as to the first, which removes a class of human error caused by fatigue and rush. Genuine accuracy improvement depends on the quality of your historical cost data, which is usually the real constraint.
Is it worth waiting for the technology to settle down?
Waiting is defensible for large platform commitments and expensive for narrow ones. The cost of trying a single-workflow tool is now low enough that the information you get from trying usually exceeds the cost. The thing not to do is a broad, multi-year commitment while the category is moving this fast.
What happens to our data if we stop using a tool?
Ask before you start, get it in writing, and treat a vague answer as an answer. You want to know whether you can export in a usable format, what the vendor retains, and how long deletion takes. This is the same question as with any software, but it matters more when the tool has been reading your contracts and drawings.
Are the survey numbers here going to age badly?
Some will. Adoption figures move fast and every percentage above is from 2025 surveys, so treat them as the state of play entering 2026 rather than as current. The structural findings are more durable: the belief-versus-budget gap, the size split on cost, and reliability outranking every other concern.
If you can name the single task in your operation that eats the most administrative time, that is enough to have a useful conversation about whether AI helps or whether better configuration of what you own would do it cheaper.
See ClickWerxs AI services or get in touch.
Sources
- Associated General Contractors of America and NCCER, 2025 Workforce Survey Analysis — approximately 1,400 firms; 45% expect robotics and AI to positively affect construction jobs by automating manual, error-prone tasks; 44% expect improved job quality and safer, more productive workers; about one third expect no significant effect; 12% expect job elimination, down from 14% in 2024 and 17% in 2023; difficulty filling AI personnel/specialist positions reported by 47%, up from 30% in 2024; 88% of firms employing craft workers report openings; 57% report candidates not qualified; 45% report project delays from worker shortages, against 54% in 2024, 61% in 2023 and 66% in 2022. agc.org
- Dodge Construction Network with CMiC, contractor AI survey — 235 general and trade contractors surveyed September and October 2025; 85% expect reduced time on repetitive tasks; 75% expect help learning from historical project data; 51% actively evaluating AI-related changes; 40% allocating a dedicated AI budget; 86% of large contractors versus 69% of small and mid-sized expect competitive advantage; 49% of smaller firms versus 26% of large cite investment cost as a major concern; 57% cite reliability and accuracy concerns; 54% cite data security and privacy risk. Reported via constructiondive.com
- AGC and Sage, Construction Hiring and Business Outlook, published 5 February 2025 — more than 1,100 contractors; AI the leading category for increased technology spending with 44% planning to boost AI investment. sage.com
- ClickWerxs field operations platform — AI assistance shipped in quote drafting, photo handling, project insights and change order preparation by early June 2026, with per-tenant usage metering, a written threat model for AI timeouts and a dedicated timeout test; 318 commits, 71 threat models and 19 test files as of 4 June 2026. First-party operator data.
Third-party survey figures cited in this post are attributed to their publishers with sample sizes and dates; adoption statistics move quickly, so verify at the source before relying on them. ClickWerxs sells AI implementation services and earns revenue from those engagements. This post reflects our direct experience building AI features into our own platform, not independent third-party research. This is operator opinion and not legal, financial, or professional advice.
Kaleb Dickhaut — Founder, ClickWerxs. Kaleb built ClickWerxs from the ground up, from payment processing ISO to the Command Center platform to the AI SEO methodology the blog runs on. He has onboarded hundreds of small businesses onto payment and CRM systems. linkedin.com/in/kaleb-dickhaut