Nearly seven in ten construction companies have invested at least $250,000 over the past two years in AI, but few accurately assess its value, according to AI in Construction: Measurable Results, Unclear Benchmarks, a report by consulting firm BRG.

“Double-digit cost savings are common across every project phase, but no single yardstick has emerged to capture AI’s full value—and too narrow a definition may limit opportunity,” BRG says. “AI is more likely to deliver better outcomes than faster ones. Across most functions, more respondents reported gains in efficacy—quality, performance, and outcomes—than in efficiency.”

The report outlines five best practices for optimizing AI use:

1. Focus implementation around a specific problem. AI application is strongest in information-intensive work surrounding a project, such as estimating costs, reviewing drawings, scheduling, document analysis, and managing claims. Organizations should begin with a specific decision or outcome to improve, determine whether the necessary information is available, and establish how the output will be reviewed.

2. Invest in talent, data, and integration. Organizations will struggle to scale AI without skilled employees who have access to reliable project data and tools that work with existing systems. Training alone will not solve poor data or disconnected systems, just as better infrastructure will not fully capture potential value without people who can navigate it.

3. Establish accountability for AI-assisted work. Liability for errors is the most frequently cited concern, yet fewer than half of respondents say their policies address human oversight and accountability, and only about one-third address accuracy and reliability. Organizations should define who is responsible for validating AI-assisted work, when to require independent verification, and how to document inputs and outputs.

4. Don’t let narrow ROI measures limit opportunity. Define success before deployment, but don’t reduce every AI initiative to hours saved or immediate cost reductions. Leaders should ask if implementing AI allows the organization to perform better work, manage risk differently, or do something new.

5. Prepare for contracts to evolve with AI use. Contracts may increasingly need to address how AI can be used, who is responsible for validating its outputs, which records must be retained or disclosed, and how liability is allocated.