New RICS research covering more than 3,100 professionals worldwide found about two-thirds of construction people now use AI in some form, up from just over half a year earlier. Only 19% use it regularly. Commercial property runs pilots at exactly the same rate as construction, 39%, yet converts them into routine work far more often. The gap is not enthusiasm. It is what buyers demand from vendors before signing, and most contractors are still asking the wrong questions in the room.
There is a particular kind of meeting that happens in construction offices right now. A vendor demos something clever. Heads nod. Somebody says the word pilot. Three months later the pilot ends, the champion moves to a different project, and the tool quietly stops being opened. Nobody calls it a failure because nobody called it anything.
The RICS AI in Commercial Property and Construction Report 2026 puts numbers on that pattern, and they are unflattering. Two-thirds of construction professionals are touching AI in some form. Regular use sits at 19%. Between those two figures live a lot of dead pilots.
What makes the data useful rather than just gloomy is the comparison sitting next to it. Commercial property runs early-stage pilots at 39%, the identical rate to construction, but converts them into regular use at 29% against construction’s 19%. Same appetite, better landing. That difference is a buying problem, and buying problems have fixes.
Your firm is probably already past the question of whether to try AI. The live question is what you insist on hearing from a vendor before the trial starts, because the answers you accept in that first meeting decide whether you end up in the 19% or the 39%.
The adoption numbers depend entirely on who is counting
Two credible surveys, wildly different headlines
Before anyone quotes a percentage at you, check the definition underneath it. RICS surveyed more than 3,100 professionals globally and found roughly two-thirds of construction respondents using AI in some form. Bluebeam’s 2026 AEC Technology Outlook, referenced in a recent piece by Bluebeam CEO Usman Shuja, puts the figure at 27% of firms using AI in their workflows.
Both can be true. Somebody pasting a spec clause into a chatbot on their own laptop counts as using AI in some form. It does not count as an AI workflow. The distance between 27% and two-thirds is the distance between individual curiosity and organizational practice, which happens to be exactly what this article is about.
Worth flagging: the PBC Today article carrying the Bluebeam figure is marked as a commercial profile, written by the CEO of a company selling AI tooling. That does not make the observations wrong, and several are sharp. It does mean reading those survey numbers as a vendor’s framing of a market they sell into.
Why one sector converts pilots and the other does not
Same starting line, different finish
Both sectors experiment at the same rate. The divergence shows up afterward, and investment intent follows the same shape, with 47% of commercial property respondents planning to increase AI spend over the next year against 37% in construction.
RICS is blunt about where the problem has moved. Getting organizations to experiment is broadly solved. Embedding AI into professional practice is not, and the body is now telling firms to plan for that transition at the outset, thinking about data, integration, professional supervision and governance before the pilot rather than as a cleanup exercise once it works.
Progress is real, to be fair to the industry. The share of construction organizations reporting no AI use has dropped by more than a third since 2025, and regular use climbed from 12% to 19%. The direction is right. The conversion rate is the weak link. We covered the buyer-side version of this in our piece on why customer adoption pace, not model capability, is the real brake on ConTech AI.
Read the vendor’s own conditions back to them
Three things sellers admit matter, turned into buyer questions
Shuja’s article lists three conditions his company drew from more than 2,000 firms testing Bluebeam Max in beta. Read from the vendor side they are product philosophy. Read from your side of the table they are a checklist, and a fairly demanding one.
The first is that the tool has to live where the work already lives. Their survey put integration complexity ahead of cost, leadership buy-in and skills gaps as the top barrier. Martin-Harris Construction, a Las Vegas firm ENR ranks among the top contractors in the West, used the beta on revision comparison during preconstruction. Their director of preconstruction, James Miranda, described the value as being able to “instantly see what’s changed and where to focus”. That is not a new process. It is an existing one with the manual hunting stripped out.
The second is trust arriving before data does. RICS found privacy and security concerns in construction jumping from 22% to 30% in a single year, so this is getting more acute. Shuja’s argument is that the answer has to be architectural rather than promissory, with heavy processing contained inside the vendor’s environment and only cleaned, restricted information moving beyond it. Whether a given vendor actually builds that way deserves a documented answer from your procurement team, not a reassuring one.
The third is likeliest to survive contact with your P&L: the tool has to give time back. Not improve productivity in the abstract. Return hours. One beta user reported saving four to six hours a month on bookmarking and page labelling alone, which sounds unglamorous because it is. Our earlier coverage of how AI is landing inside the construction workflow found the same pattern: the features that stick target high-volume, repetitive, error-prone work, not the impressive demo.
Know which AI pitches are worth your afternoon
One sharp briefing each week on the deals, tools, and adoption signals reshaping construction.
Join 3000+ Readers
What to actually ask before you sign
Five questions that separate a rollout from a dead pilot
The RICS guidance and the vendor’s own beta lessons converge on a short list. None of these are technical questions, which is rather the point. Your estimator does not need to understand model architecture to know whether a tool fits their Tuesday.
Governance is the part nobody wants to own
The standard that arrived in March
The skills gap keeps showing up in the RICS data, and it is not only a technical one. Somebody has to decide what good AI use looks like inside your firm, write it down, and enforce it. Most contractors have not assigned that job to anyone.
RICS put a professional standard, Responsible use of artificial intelligence in surveying practice, into effect in March 2026, and is pushing awareness of it as AI gets embedded into professional work. Maureen Ehrenberg FRICS CRE, President Elect of RICS, said this is the moment for the profession to “lead, not follow”.
For a firm on thin margins, governance sounds like overhead you cannot afford. The counter-argument is that the 39% sitting in permanent pilot mode are already paying, just in trial licenses and wasted attention rather than policy work. One of those costs compounds into capability. The other does not.
| What you hear from the vendor | Pilot-grade answer | Rollout-grade answer |
|---|---|---|
| Where it lives | New dashboard your team logs into | Inside a tool the team already opens daily |
| Data handling | Assurance that data is secure | Documented architecture showing what leaves your environment |
| Value claim | Improves productivity | Returns a stated number of hours to a named role |
| Failure mode | Accuracy is very high | A defined human review step with an owner |
| Time horizon | Let’s see how the trial goes | Agreed conditions for what routine use looks like |
| Training | Documentation and a webinar | Named internal champion plus a plan for when they leave |
Frequently asked questions
According to the RICS AI in Commercial Property and Construction Report 2026, based on more than 3,100 responses worldwide, 19% of construction organizations use AI regularly, up from 12% the previous year. About two-thirds report using it in some form, and 39% are running early-stage pilots. The gap between those figures is the conversion problem the industry is now facing. (Source)
Because they measure different things. RICS asked about AI use in some form and found roughly two-thirds of construction respondents. Bluebeam’s 2026 AEC Technology Outlook asked about AI in workflows and found 27%. Individual experimentation and embedded organizational practice are separate phenomena, and headline percentages rarely say which one they are counting. Always check the definition before quoting the number. (Source)
It depends which survey you read, which is itself informative. RICS points to a shortage of skilled personnel, alongside privacy and security concerns that rose from 22% to 30% in construction over the past year. Bluebeam’s survey put integration complexity at the top, ahead of budget, leadership buy-in and skills gaps. Both point away from cost as the primary blocker. (Source)
The RICS data shows both sectors running pilots at the same 39% rate, but regular use reaching 29% in commercial property against 19% in construction. Investment intent follows the same pattern, at 47% versus 37% planning increases. RICS does not attribute a single cause, though it does call on organizations to plan for data, integration, supervision and governance from the outset. Treat any tidier explanation as a hypothesis rather than a finding. (Source)
Five questions cover most of it. Which existing workflow does this sit inside. Where does our data go and who holds liability. How many hours does it return, to which role, per month. What does human supervision look like when the tool is wrong. And what has to be true in six months for this to count as routine. The last one matters most, because RICS is explicitly telling firms to answer it before the pilot rather than after. (Source)
Yes. The RICS professional standard, Responsible use of artificial intelligence in surveying practice, came into effect in March 2026. RICS says greater awareness and adoption of it will matter more as AI becomes embedded in professional work. For firms without any internal AI policy, it is a reasonable place to start rather than drafting from scratch. (Source)
Use them, but read the label. The PBC Today article containing Bluebeam’s figures is marked as a commercial profile and authored by the company’s CEO, which does not invalidate the data but does frame it. Vendor surveys tend to sample their own orbit and define categories in ways that suit the product. Independent bodies like RICS carry their own framing too. Cross-checking two sources with different incentives gets you closer than trusting either alone. (Source)
Related Articles
Bluebeam Max Goes Live: AI Lands Inside the Construction Workflow
There Are No More Early Adopters: Construction’s AI Buying Shift Has Already Happened
“You Can Only Go As Fast As Your Customers”: The Quiet Brake on ConTech AI
