Breaking the Productivity Paradox: How AI Can Reverse Construction's Declining Efficiency

Speaking at Offsite25, the annual conference of prefabAUS on 28 August 2025, Associate Professor Dominik Holzer from the University of Melbourne delivered a compelling presentation on how Artificial Intelligence could revolutionise the prefabrication industry. His talk, "From AI to Factory," outlined the immense opportunities and significant challenges facing the construction sector's digital transformation.

The Mounting Crisis

Holzer began by highlighting Australia's stark housing reality. The nation is falling dramatically short of its National Housing Accord targets, with actual dwelling approvals continuing to diverge from desired targets. Perhaps most troubling is that Australia has actually become slower at building homes over time—the average period between commencing construction and completion has extended from around six quarters a decade ago to more than nine quarters today.

When asked to explain this paradox in an age of technological advancement, Holzer was unequivocal: "It is often not just what we can do within the design and delivery side of things that determines how effective we are," he explained. "It is often also constrained by contracts; it is constrained by government overregulation associated with application and approval processes."

This decline in productivity occurs against a backdrop of severe skills shortages. Australia currently faces a deficit of 83,000 tradespeople, creating a perfect storm that threatens to derail housing delivery targets.

Why AI is Different This Time

The construction industry, traditionally one of the slowest adopters of new technology, must now confront whether AI can provide solutions to these systemic challenges. Holzer believes AI represents a fundamental shift from previous technological advances.

"There are numerous patterns of information and knowledge management that can be boosted by the capabilities of AI," he said. "It can really affect a whole different range of parts of your information flows and knowledge systems."

This breadth of application sets AI apart from previous technologies like Building Information Modelling (BIM), which primarily benefited spatial coordination but had limited impact on project finance, procurement, fabrication processes, or logistics.

AI Applications Across the Construction Workflow

Holzer's presentation mapped out comprehensive applications across the entire prefabrication process. In finance and feasibility, AI can assist with market analysis, demand forecasting, site suitability assessment, and dynamic financial modelling that updates projections as conditions change. For production management, AI enables predictive analysis for optimising schedules, reducing production cycle times, and automating compliance checks when interfacing with BIM systems.

On the factory floor, AI applications include risk forecasting, inventory management, and automated quality inspections. "You have computer-supported visual checks where you have quality checks that would allow you to detect any defects or any deviations from what was designed," Holzer explained, noting this could "speed up your QA processes dramatically."

The technology extends to logistics through process simulation, supply chain vulnerability detection, and route optimisation. For site management, applications include minimising construction risks, supporting crane movement optimisation, and facilitating augmented reality installations.

Avoiding the Implementation Trap

Perhaps even more impressive than federal recognition is the systematic state-by-state adoption of Smart Building priorities. Queensland has set a 50% MMC target for government projects while preparing for the 2032 Olympics build.

At the conference, Assistant Minister for Planning, Housing and Better Regulation, The Hon Rebecca Young, announced that a dedicated MMC sub-group would be added to the Queensland Building Ministerial Advisory Council (BMAC). This provides Modern Methods of Construction with focused attention within an influential reform body where prefabAUS already serves as a council member.

New South Wales launched a $10 million Modular Housing Pilot with Pattern Book fast approvals for modular construction.

Victoria committed $50 million to a Future of Housing Centre of Excellence, Western Australia allocated $50 million to its Housing Innovation Program, and both South Australia and Tasmania established dedicated MMC social housing programs.

The financial sector has also shifted dramatically. CommBank now offers prefab-specific products allowing 80% contract price access before home installation—a crucial breakthrough that addresses the sector's unique procurement requirements. A Federal Treasury Working Party is actively addressing remaining MMC finance barriers.

The Data Foundation Challenge 

Central to Holzer's message was the critical importance of data preparation—a step many organisations overlook in their rush to implement AI solutions. "The biggest investment probably would have to be this sort of data cleansing, data tagging exercise that needs to happen to then really reap the benefits over time," he explained.

This data work involves several key components. First, organisations need to establish in-house reference databases that are "specific to your organisation, specific to what your core business is about." The goal isn't about homogenisation of approaches across the industry—Holzer stressed: "We don't want AI to lead to everyone doing the same thing and just hoping to do it quicker and cheaper."

The process begins with historical project analysis. "For many, it has to do with information tagging, where you look at your organisation's history and projects, you standardise some of the data, and you allow the AI to harvest information and extract some trends and analysis."

This historical data mining can reveal patterns in everything from financial performance to technical details that proved successful on previous projects. However, Holzer noted a fundamental industry weakness that makes this challenging: "We are such a project-by-project-based industry. We have been really bad in knowledge management and knowledge sharing, even within organisations."

Navigating Legal and Technical Obstacles

Legal issues present complex challenges that the industry is only beginning to address. On the critical question of liability when AI systems make errors, Holzer was direct: "You can't blame an algorithm for making decisions that led to wrong outcomes... You will not win a legal case if you say, Sorry, it's not my fault, it's the AI's fault."

He argues that the solution lies in "expert-in-the-loop systems" where human oversight remains integral to AI-assisted decision-making processes.

Data ownership and intellectual property protection create additional complications. There are substantial risks of inadvertent competitive intelligence sharing, particularly given AI's appetite for data. "AI is very hungry and keen to grab information from everywhere," Holzer noted, emphasising the need for careful data governance and new AI-inclusive policies by insurers.

The Low-Hanging Fruit

For organisations ready to begin their AI journey, Holzer identified several accessible starting points, all of which depend on proper data preparation. The biggest opportunity lies in addressing the industry's knowledge management failures by "establishing systems that allow you to bring up information in real time, and to help you make decisions based on more informed background information that's available to you much quicker than you were able to do in the past."

Other accessible applications include automated quality assurance on factory floors, transport logistics optimisation, and improved site management through crane movement optimisation and automated visual checks. However, these applications require standardised, tagged data to function effectively.

The Foundation: Data Strategy

When asked what construction professionals should do immediately to start their AI journey, Holzer's response was emphatic about data fundamentals: "They should set up an AI steering group that discusses their data and how they deal with information clearly. AI can only work if it's clean, consistent and high-quality. It's the same trash-in, trash-out conversation about data we’ve always had."

This requires what he called "a homogenisation of data structures across your business" that links different processes—from supply chain integration to design, fabrication, scheduling, transport, and finance—under unified logic. "You want AI to be able to query across, and so that really depends on how clean and purposefully appropriate the data that you deal with is."

The data strategy must also address IP and security concerns. Organisations need "a group of people within the organisation who look at this IP question, who ensure that whatever data you train your AI on is yours, or it's publicly available to the point where you're not infringing on anyone's IP."

From his consulting experience, Holzer emphasised that "data safety is the number one priority. You don't want to expose yourself to any risk." This involves establishing "very solid back-of-house data and data integration system, and a very clear understanding ethically and legally, on what data sets you’re drawing on."

Skills Transformation and Collaboration

The workforce implications are profound, with production planners, quality inspectors, and material handlers likely to see significant changes. However, rather than simply eliminating jobs, AI-enabled prefabrication requires workers who can collaborate with evolving, non-human intelligence systems.

Holzer advocates for coordinated responses involving government, education, and industry. This is also where the prefab sector needs to assist government institutions in finding more efficient ways to process building applications using AI. The problem isn’t overregulation as such, but the time it takes to manually check submissions against the myriad of compliance requirements in place. AI could play a significant role in automating a substantial number of tasks here. The prefabrication sector, being relatively small and cohesive, is well-positioned for collective action, though he acknowledges the challenge of balancing data sharing with competitive advantage protection.

Environmental Consciousness

Holzer concluded with an important warning about "meta-cognitive laziness"—the need for awareness regarding AI's environmental impact, particularly the substantial energy demands of data centres. "Not everything you could do using AI... you should do using AI," he cautioned, emphasising the importance of thoughtful, sustainable implementation.

The Path Forward

The transformation of Australia's construction industry through AI represents both an unprecedented opportunity and a complex challenge. As Holzer noted, the technology is evolving rapidly—"ChatGPT hasn't even existed in any useful way up until, let's say, 3 years ago"—requiring urgent but strategic responses.

However, his message was clear that success hinges on proper preparation: "Don't be naive, don't think it's like a tool that you plug in and it will suddenly boost your productivity. You really need to be quite strategic about what it is that you want to use it for."

Success will require unprecedented collaboration between government, education, and industry, built on a foundation of clean, well-structured data. For the construction sector to realise AI's potential in addressing Australia's housing crisis, the focus must shift from viewing it as a simple technological tool to understanding it as a comprehensive transformation of how the industry captures, manages, and leverages information to make better decisions and deliver projects more effectively.

The message is clear: AI adoption in construction isn't just about technology—it's about reimagining an entire industry's relationship with data, knowledge management, and systematic learning from past experience.

Associate Professor Dominik Holzer is deeply involved in researching the impact of AI across the AEC sector, and he lends his expertise on strategic AI adoption to industry via his consultancy AEC Connect.