AI in construction in 2026: what the surveys say about adoption, and what holds contractors back

Every conference deck says construction is about to be transformed by AI. Surveys are a better guide than decks. Here is what three credible sources say, with the caveats that go with each, and what it means if you run a contracting or owner organisation.

RICS: use is widespread, integration is not

The RICS AI in commercial property and construction report 2026, published in August 2026, draws on 1,883 responses from its Global Construction Monitor. Its headline findings for construction:

• Around two-thirds of respondents use AI in some part of their work, up from just over half a year earlier.
• Regular use in specific processes rose from 12% to 19%.
• Full organisation-wide integration is below 1%; wide-ranging integration across several processes "remains rare".
• 37% plan a moderate or significant increase in AI investment, up from 30% in 2025.
• Privacy and security concerns rose as a barrier, from 22% to 30%, while high implementation cost fell from 29% to 25%.

Dodge and CMiC: optimism, but weak data

In December 2025, Dodge Construction Network and CMiC released a research report, AI for Contractors. As reported in the release and press coverage, 87% of contractors believe AI will meaningfully affect construction, only 19% have adapted legacy workflows for it, and only 26% rate their data as high quality. Coverage disagrees about the survey's sample size, so read the original report before quoting a number of respondents.

Autodesk and FMI: what bad data costs

The Autodesk and FMI study of more than 3,900 construction professionals estimated that "bad data", meaning data that is inaccurate, incomplete, inaccessible, inconsistent or untimely, may have cost the global construction industry $1.85 trillion in 2020. It is older than the other two, and an estimate rather than a measurement, but it names the problem that the newer surveys keep running into.

What this means in practice

1. Start where the data is clean. Job cost and commitments in an accounting system are structured and audited; they are a better first use than free-text documents.
2. Treat access as the first problem. An assistant that cannot reach your system will give general advice, not answers. Most firms have the data and lack a safe connection to it.
3. Take the security concern seriously, because it is rising. RICS's barrier figure moved from 22% to 30% in a year. Controls that IT can verify (limits per person, exclusion of payroll, an audit trail) are what unblock projects.
4. Measure with questions you can check. Pick a few questions whose answers you already know, and judge any tool by those.

If you want to get your own house in order first, our checklist for preparing job-cost data for AI is a practical start, and the security checklist covers the questions to ask vendors.