Closing the construction AI blind spot: new research urges consistent national data
While technology vendors and industry commentators frequently herald artificial intelligence as the solution to Australia's construction productivity slump and housing shortages, new research from the Faculty of Architecture, Building and Planning (ABP) reveals a critical data vacuum.
In a newly released white paper, The Construction Blind Spot: Why Australia cannot yet measure how construction uses AI, and what planners and policy makers need instead, Senior Lecturer in Construction Technology Dr M. Reza Hosseini leads a cross-sector research team auditing national AI adoption metrics.
The report was co-authored with senior leaders across tier-1 contracting, engineering consulting, contech advisory, and state infrastructure:
- M. Reza Hosseini (ABP)
- Tim Mumford, Acciona
- Alex Afshar, Dolanto
- Dr Amir Soroush, Victorian Infrastructure Delivery Authority (VIDA)
- Belinda Hodkinson, AECOM
The authors highlight that Australian sources report construction AI adoption anywhere between 6 per cent and 44 per cent. According to the Australian Bureau of Statistics (ABS) Characteristics of Australian Business 2024-25, just 6 per cent of construction businesses used AI, placing the sector 14th out of 17 national industries. Conversely, commercial surveys of corporate enterprise leaders report adoption rates as high as 44 per cent.
"The confusion comes from mixing three entirely different kinds of measures and calling them all 'construction'," explains Dr Hosseini. "The ABS captures the vast subcontracting pyramid where 98 per cent of firms are small businesses, while enterprise surveys focus on well-resourced corporate headquarters."
Invisible on the Job Site:
The white paper also examines live usage data from Anthropic's Claude generative AI platform. Despite Australia leading the world in per-capita Claude use, site-based construction jobs ranked 21st out of 22 occupation groups, accounting for just 0.11 per cent of conversations.
While engineering professionals (surveyors, civil engineers, architects) show up in chatbot queries, physical trades such as carpenters, plumbers, electricians, and machine operators barely register. Furthermore, national surveys completely fail to track AI embedded in domain-specific estimating, 4D scheduling, and building information modelling (BIM) tools.
Recommendations for National Policy:
To ensure Australia's $200-billion infrastructure pipeline and national housing targets are supported by sound evidence, the authors propose a five-point reform agenda:
1. Standardising business data using official ANZSIC codes;
2. Tracking embedded construction software rather than just chatbots;
3. Surveying site workers and subcontractors on the tools;
4. Reporting adoption by firm size and state; and
5. Mandating simple, lightweight AI reporting on major publicly funded projects.