- Industry Insights
- 19 August 2026
Beyond the Theatre: State of AI in Australian Construction.
A Visibuild research report. 2026.
Seventy-four per cent of the construction leaders we surveyed believe AI will decide who wins in their market within three years. Nine per cent have the data foundation to act on that belief. That distance, between what the industry believes and what it is ready to do, is the single most important fact about AI in construction today.
We surveyed 34 general contractors across Australia and New Zealand, on projects ranging from $20 million to $2 billion. We asked them what they believe about AI, and what they have actually built to use it. The answers were almost unanimous in their conviction and almost unanimous in their unreadiness. Belief has run a long way ahead of capability, and the gap between the two has a name. It is AI theatre: the appearance of progress without the foundation that makes progress real.
This is not a story about construction being behind. It is a story about the whole economy arriving at the same place at once, and about the one advantage construction holds that almost no other industry does.
Thirty years of standing still
For three decades, construction productivity has barely moved. Over that period, manufacturing lifted labour productivity by 3.6% a year and the wider economy by 2.8%. Construction managed 1.0%. The gap compounds into one of the largest pools of lost value in the world economy. McKinsey Global Institute puts the prize for closing it at US$1.6 trillion a year.
AI is the most credible instrument for closing that gap in a generation. Which is exactly why the pressure to adopt it feels overwhelming, and exactly why so much of the current activity is theatre rather than progress.
Labour-productivity growth, construction versus the wider economy. Source: McKinsey Global Institute; Visibuild, Beyond the Theatre 2026.
Construction is not alone in struggling here. Across the wider economy, MIT’s Project NANDA found that 95% of enterprise generative AI pilots deliver no measurable impact on profit and loss. Gartner expects 60% of AI projects without AI-ready data to be abandoned through 2026. IBM found that 77% of organisations say AI is already outpacing their ability to govern it. High spend, thin returns, missing foundations. The pattern outside construction is the same pattern inside it.
The difference is that construction already runs the strongest evidence culture in the economy. ITPs. Hold points. NCRs. You cannot pour a slab without a signature. That discipline, the habit of proving work before it is accepted, is exactly what AI needs. Most industries have to build a quality culture from nothing before AI can be trusted. Construction already has one. It has simply never pointed it at its own data.
What the research found
The detail behind the headline is where the theatre becomes visible.
Belief in AI’s importance sits at 7.9 out of 10. Readiness to act on it sits at 4.8. That gap is the shape of the whole problem in a single line.
Client pressure, the reason most leaders assume they are moving, is not actually there. Contractors rate real client pressure to adopt AI at 4.1 out of 10, and half rate it three or below. The mandate is coming from inside the business, not from the market.
The foundation is missing. Only 38% of respondents have a data lake at all, and of those, only a fraction believe it surfaces anything they can act on. Across the full sample, just 9% say their data foundation actually works.
Nobody owns it. Not one respondent rated the clarity of AI ownership in their business above 8 out of 10. Ownership is the quiet failure underneath the loud ambition.
And the workforce is the wall. 65% call workforce adoption a significant challenge, and even among the minority who consider themselves AI leaders, average readiness still only reaches 5.9 out of 10. The people most confident about AI are not meaningfully more ready than everyone else.
The diagnostic: every contractor is on this chart
Two questions place any business precisely. Do you believe AI will define competitive position within three years? And do you have the data, ownership and infrastructure to act on that belief? Plot conviction against capability and four groups appear.
The four positions. Conviction on the vertical axis, capability on the horizontal. Source: Visibuild, Beyond the Theatre 2026.
47% are at risk of AI theatre.
High conviction, low capability. They believe, they feel the pressure, and they are buying tools their data cannot support. This is nearly half the market.
38% are building.
Conviction and capability are moving together. They are not finished, but they are honest about the order of operations.
9% are yet to start.
Low on both. Behind, but without the exposure that comes from spending ahead of the foundation.
6% are sceptical pragmatists.
Capable but unconvinced. They could act and have chosen to wait.
AI theatre, then, is not a failure of ambition. It is ambition without the foundation to hold it. The 47% are not the least serious contractors in the market. Many are the most serious. That is what makes the position dangerous.
Five barriers between belief and readiness
Underneath the diagnostic sit five specific barriers. Each has a number, a reading, a move that addresses it, and a test that tells you whether the move worked.
One. The phantom mandate
Teams rate client pressure at 4.1, but internal pressure at 5.4. The mandate is coming from inside. Most AI programs begin because adoption feels overdue, not because a client asked. That is fine, right up until the pilot has no owner and no measure, because nobody ever wrote down what it was for.
The move: write the mandate before the pilot.
The test: it fits on one page.
Two. The data that is not there
Everyone has data. 38% have a lake. 9% have one they believe works. The lake itself is not the requirement. What matters is whether a business has ever deliberately structured its data, the way an ITP structures an inspection. Tools bought onto a five-out-of-ten foundation run on incomplete data at best, and produce confidently wrong answers at worst.
The move: run the data foundation like a quality system.
The test: you can answer three cross-project questions in under an hour.
Three. No name on the door
53% rate ownership clarity at five or below. 29% rate it three or below. Nobody rated it above eight. AI without an owner is a pilot that quietly dies. Someone has to be accountable when it goes wrong, not merely credited when it goes right.
The move: name one owner, then test the naming.
The test: the CEO can write the accountable owner, the threshold and the KPI in thirty minutes.
Four. The workforce wall
65% call workforce adoption a significant challenge, and 32% call it critical. The tool that never reaches the prestart never reaches the project. Adoption is not a training problem you solve at the end. It is a sourcing problem you solve at the start, on site, with the people who will use it.
The move: take the pilot to the prestart.
The test: the site names the problem the tool solves, unprompted.
Five. The signal gap
Executives hear client pressure at 2.1 out of 10. Their own quality managers hear it at 6.4. The people closest to the client and the people making the AI decisions are reading different signals. The executive sees calm. The delivery team sees demand. Both are right, and that is precisely the problem.
The move: reconcile the signal monthly.
The test: the executive can say what clients asked the delivery teams in the last ninety days.
Five shifts: from theatre to discipline
The businesses crossing from belief to readiness are making the same five shifts.
From AI as a productivity tool, to AI as a quality and risk discipline. From innovation theatre, to outcome accountability. From data lakes, to data discipline. From buying tools, to building foundations. From a top-down mandate, to bottom-up validation on site.
None of these is a technology decision. Every one of them is a quality decision, and construction already knows how to make quality decisions.
Four questions to ask before you spend another dollar
- Who owns this, and what happens when it goes wrong?
- Would our data pass the inspection our concrete does?
- What would the site actually use?
- What measurable outcome, by when, and who checks?
If a business cannot answer these four, it is not ready to buy. It is ready to build the foundation that makes buying worthwhile.
The bottom line
The front-runners two years from now will not be the contractors with the biggest AI budgets. They will be the ones who treated AI the way they already treat quality: owned, measured, and proven on site before it was scaled anywhere.
The revolution is real. It just has not happened yet. Whether it happens to you or for you is decided by what you build this year.
About the research
The findings above are drawn from a survey of 34 Australian and New Zealand general contractors, conducted in May 2026, on projects ranging from $20 million to $2 billion. Respondents spanned executive, quality management, and operations and technology roles. The cohort includes ten of the top fifty contractors in the Hubexo Construction League Australia.
This is directional research, not a population study. It is weighted toward Australia, self-ratings tend to skew positive, and Visibuild is a vendor publishing findings that run against its own commercial interest. We think the picture is worth having anyway.
The full report, including the complete methodology, the readiness index, and the barrier-by-barrier data, is available as a PDF.
