Construction has an AI pilot-to-production gap: two out of every three organisations now use artificial intelligence in some form. Fewer than one in five use it routinely. That gap — between trying AI and running on it — is the most important finding in the industry’s newest evidence base, and it is where advisory work now sits.
The Royal Institution of Chartered Surveyors published its AI in Commercial Property and Construction Report on 18 August 2026, drawing on 3,148 responses across two of its global monitors. It is the largest recent dataset on how the built environment is actually adopting AI, as opposed to how vendors say it is.
What the data shows
In construction, AI use rose from roughly 51% in 2025 to about 67% in 2026. The share of organisations reporting no AI use at all fell by more than a third in a single year. On the face of it, that is rapid adoption.
The composition tells a different story:
- 39% of construction organisations are running early-stage pilots
- 19% use AI regularly in specific processes — up from 12%, but still under a fifth
Commercial property has moved further along the same path. More than 75% report some AI use, pilots sit at the same 39%, but regular use reaches 29% and around 6% describe widespread or full integration.
The pilot-to-production gap
RICS names the problem directly: both sectors pilot at an identical rate, yet commercial property converts pilots into routine use far more effectively — 29% against construction’s 19%.
Analysis: identical pilot rates with divergent production rates rules out the most common explanation. This is not a difference in appetite, budget or awareness. Both sectors are trying at the same rate. Construction is converting worse, which points to something structural in how construction work is organised rather than to any reluctance to experiment.
The most plausible structural difference is the unit of work. Commercial property manages standing portfolios — the same assets, the same data, the same processes, month after month. A pilot that works in March is still relevant in September. Construction manages projects: temporary organisations, assembled per job, disbanded on completion. A pilot proven on one project has no institutional home when that project closes.
The barriers have shifted
The reported obstacles in construction, in order:
- Lack of skilled personnel
- Integration with existing systems
- High implementation costs — falling, from 29% to 25%
- Privacy and security concerns — rising, from 22% to 30%
In commercial property the leading barrier is different again: data quality and availability, ahead of integration and skills.
Analysis: the movement matters more than the ranking. Cost is receding as a barrier while privacy and security climb eight points in a year. That is the signature of a technology moving out of the innovation budget and into the operating estate — nobody runs a security review on a pilot nobody depends on. The rise in security concern is evidence of seriousness, not of hesitation.
What closes the gap
The following is PGAN’s advisory view, not a finding of the RICS report.
- Own the pilot above project level. If a tool is sponsored by a project, it dies with the project. Sponsorship has to sit with a function that outlives the job — commercial, engineering, or the PMO.
- Define production before piloting. Most construction pilots have no stated threshold at which they graduate. Without one, a successful pilot and a failed pilot end the same way: quietly.
- Treat the security review as a graduation gate, not an obstacle. The 30% now citing privacy and security are the organisations closest to production. Build the review into the pilot design and it stops being the thing that blocks rollout.
- Fix data before adding tools. Commercial property’s leading barrier is data quality, and it is the sector converting better. That ordering is not a coincidence — it is what a sector discovers once it gets past pilots.
- Buy skills as capability, not headcount. The top barrier is skilled personnel, in a market where those people are scarce and expensive to hold between projects. Fractional and advisory engagement models exist precisely for this shape of demand.
Frequently asked questions
How many construction organisations actually use AI regularly?
About 19% use it regularly in specific processes, up from 12% the previous year. A further 39% are running early-stage pilots.
Is cost still the main barrier to AI adoption?
No. Cost is falling as a cited barrier, from 29% to 25%. Skilled personnel and systems integration now rank higher, and privacy and security concerns are rising sharply.
Why does commercial property convert pilots better than construction?
RICS identifies the gap but does not attribute a single cause. PGAN’s view is that portfolio-based work gives a pilot a permanent institutional home, whereas project-based work disbands the organisation that proved it.
Sources
All statistics from the RICS AI in Commercial Property and Construction Report 2026, published 18 August 2026 by Tom Storey MRICS, Anil Sawhney FRICS and Katherine Pitman. Based on 3,148 responses — 1,883 to the Global Construction Monitor and 1,265 to the Global Commercial Property Monitor. Figures cited are as reported by RICS; readers making investment or operational decisions should consult the full report.
PGAN advises infrastructure and construction organisations on moving AI from pilot to production — governance, data readiness and the capability model behind it. Explore our advisory practice, our work across technology and digital and construction and real estate, or request advisory. Read our briefing on Dubai Law No. 3 of 2026.