For two years, the question construction leaders have asked about AI has been blunt: will it take jobs? A new survey suggests the more useful question is how it will change the jobs we already have.
In a survey of 109 builders and owners released this month, Berkeley Research Group found that nearly half of North American respondents had changed their team structures because of AI. Close to four in ten had retrained staff. Only about one in nine said they had cut headcount, while roughly one in five had added people in some areas. In short, AI is reorganizing work far more often than it is eliminating it.
That matches what many contractors are seeing in practice. The tools are arriving fast, and the hard part is fitting them into the way crews actually build.
Where AI is landing first
BRG found the earliest impact falls on information-heavy work: estimating, procurement, design review, and schedule and contract oversight. Eight in ten North American builders surveyed now use AI to help evaluate and plan projects. Some tools check drawings for missing details. Others draw on past projects to forecast cost.
The field is catching up. Recent coverage of job site AI highlights progress-tracking systems that compare camera and lidar captures against the 3D model and schedule, flagging work at risk of slipping. This week, EarthCam released AI connectors that let teams ask general-purpose AI assistants about conditions on their sites, drawing on pre-analyzed camera imagery. The company also added models for trench safety, housekeeping hazards and delivery verification.
The pattern is clear. AI is moving from the office toward the point where work is installed and inspected.
The gap between buying AI and using it
Momentum doesn’t mean maturity. Only 15% of BRG’s respondents said AI was fully integrated across their projects and operations. Construction also trailed retail and financial services on governance: fewer than half of construction respondents had policies covering human oversight and accountability, and only about a third had policies on accuracy and reliability.
The barriers will sound familiar to anyone who has run a technology pilot:
- Skills. Four in ten respondents cited a shortage of people who know how to use these tools well.
- Liability. Over 40% worried about who is responsible when an AI-generated output is wrong.
- Habit. BRG noted that much of the training happens in-house and only sticks with constant use and visible support from leadership.
In other words, software alone doesn’t change outcomes. It changes outcomes when it fits into a crew’s daily routine and someone stays accountable for every decision.
What this means in the field
For superintendents, foremen and QA leads, the practical lesson is to judge AI by whether it shortens the distance between the plan and the work in place. A dashboard in the trailer helps. An answer delivered at the moment a worker is about to set a sleeve or anchor a bracket helps more.
That is where augmented reality and AI start to reinforce each other. Progress cameras and drones tell you what happened yesterday. A worker wearing AR glasses can see the approved model laid over the actual space today, before the concrete is poured or the drywall closes the wall. If the software spots a deviation from the plans, the fix costs minutes, not a change order.
The survey’s findings on structure and oversight matter here too. Tools that keep a person in the loop, show the evidence behind each flag and log what was accepted or overridden are much easier to govern than tools that act on their own.
Four steps before your next AI pilot
- Start with a costly, recurring problem. Rework from misread drawings, layout errors and missed inspections are common candidates. Pick one, not five.
- Measure a baseline first. BRG recommends tracking estimating accuracy, schedule performance, rework and change orders. Capture those numbers before the pilot so you can prove, or disprove, the value.
- Plan the role changes. Decide who reviews AI flags, who signs off, and how field staff will be trained. Treat it as a crew process change, not an IT install.
- Write the oversight rules early. Define what the tool may suggest, what a person must approve, and how decisions are recorded. That answers the liability question before it becomes a claim.
How Augvator fits
Augvator is building AR overlay software for construction guidance. It runs on AR glasses and combines BIM interoperability, AI decision support, and voice and gesture controls. Teams upload or scan their plans, and the software flags where work in place deviates from the approved design. Workers stay in control: the system surfaces the issue and the evidence, and people decide.
We’re at the prototype stage and working with industry partners to test the platform on real projects. We’re looking for contractors who want to measure the impact on rework and layout accuracy with us, using exactly the kind of baseline described above.
Want to see how field-level guidance performs on your projects? Request Pilot Access and tell us about the problem you’d like to tackle first.
Sources
- AI reshapes construction teams more often than it cuts jobs: report, Construction Dive, 21 Oct 2026 (BRG survey)
- EarthCam Launches Construction AI Connectors, GlobeNewswire, 20 Oct 2026
- AI is moving onto the construction site, Stacker via KIFI Local News 8, 14 Oct 2026