Robots can lift bricks—but they can’t fix the deep productivity problems holding back Australia’s housing sector Speaking at The Urban Developer’s Urbanity -25 on the Gold Coast on July 30, CSIRO chief research consultant Stefan Hajkowicz makes a clear prediction. “The world of robots on construction sites is coming,” he said during his keynote, An Autonomous Future .  The image is easy to picture—machines laying bricks, drones scanning builds, demolition handled by bots.  But Hajkowicz, who has a Ph.D in economic geography, warns the hype around robotics risks distracting from a deeper issue. “Productivity shortfalls in construction are a major factor in the housing crisis,” he says. “It takes more people, more hours, more person-hours to build a house today than it did 30 years ago.” ▲ At Urbanity 2025, CSIRO’s Stefan Hajkowicz explores where automation and the construction industry intersect. The numbers tell the story. In 1990, Australia built 636 homes for every 1000 new residents.  By 2023, that figure had plunged to 276, rebounding only slightly to 367 in 2024.  The gap translates into longer waits for new homes and higher prices, even as population growth continues to climb. Also, the Housing Industry Association estimates the country will need another 90,000 tradies by 2029 just to keep pace with demand. Robots may change how work gets done, but they won’t, on their own, fix the inefficiencies that have built up over decades. A promise technology struggles to keep Hajkowicz draws on decades of economic research to explain why technology often under-delivers on productivity gains.  “You see computers everywhere being productivity statistics,” he says, referencing the Nobel Prize-winning economist Robert Solow.  “There’s no shortage of AI or computers. But when we add it up, we don’t see this productivity gain.” Companies chasing AI breakthroughs run into the same wall.  “They think they’re going to push this button called AI,” Hajkowicz says.  “And then the company will implement it and hey, bonza, we’ve got an AI-enabled company. “We’re going to be productive and our share price will—well, the share price sometimes does go up when they just announce that—but it doesn’t translate to productivity that easily.” ▲ Source: The Urban Developer He cites research by Evan Shellshear at QUT showing 80 per cent of AI projects fail to deliver results.  In Hajkowicz’s experience, the pattern is similar across industries. “Most of what we set out doing doesn’t work,” he says. “We do it again and we keep trying and the 20 per cent that do make it, totally worthwhile, because they can transform the company completely.” The problem often starts before the technology is even deployed, with too many projects chasing tools instead of outcomes.  “Start with the business problems first,” Hajkowicz says. “Don’t even think about AI. Then think about how AI and data might be able to make a difference. But choose the right use cases.” Focusing on the wrong problems, having poor data quality or lacking the infrastructure to deploy models effectively are common reasons why AI experiments stall.  “It’s challenging stuff to get value out of this,” he says. ▲ One does not simply implement AI and get productivity uplift: Hajkowicz challenges quick-fix thinking at Urbanity 2025. People, not just programs, will drive value For all the talk about algorithms, Hajkowicz argues the bigger variable is human skill.  His team analysed 73,000 job ads mentioning AI, expecting technical expertise to dominate the list. The results were surprising. “The top required skill set to get an AI job wasn’t machine learning,” he says. “It was communication, followed by machine learning, artificial intelligence, management, then Python.” The finding underlines how success with AI depends on people who can bridge technical and strategic domains.  “There’s going to be a lot that AI doesn’t do. The future is about moving from the areas that are going to become not needed to the ones which are more increasingly needed.” Where robots might make a difference In construction, Hajkowicz sees several areas where AI could help, even before robots become commonplace.  “A couple of key areas I think would be development assessment … we can speed that up a lot,” he says.  “[Also] rework and testing whether the building is built to spec ... to reduce the need for rework, is another one.” Robotics is already starting to take on heavier jobs—demolition construction bots are an example.  “They don’t have to care too much, as they say,” Hajkowicz says. “They can just go and knock it down.” But more sophisticated machines are not far behind.  “What we’re breaking through with is vision language reasoning robots,” he says.  “We say, pick up all the water bottles off the table and put it in the kitchen.  “It doesn’t know what a table or a water bottle or a kitchen is, but because it can access large language models, it can figure out what these things are, it can then look at the room, and determine a set of actions.” Those capabilities could eventually change how physical work gets done onsite.  “That’s the sort of thing that we’re working towards … watch this space in robotics,” Hajkowicz says. “The breakthroughs we’ve seen in the large language models that are so good at reasoning is coming to the physical world and that will be relevant to it.” But he stresses that robotics won’t be replacing trades any time soon.  “The idea that you’re going to just be able to get rid of the plumber and put the robot in,” he says, “it won’t look like that for a long time.” ▲ Hajkowicz says robots will not be replacing tradespeople anytime soon. Experimentation over instant transformation If Australia wants AI and robotics to make a difference in housing delivery, Hajkowicz believes it needs a culture shift. “This culture to start to happen in Australia,” he says, describing CSIRO’s approach.  “The sort of thing the culture that is in CSIRO is all about experimentation, learning, trying something, but not expecting too much of it. --> “We will have that 80 per cent failure rate,” he says. “The thing is, that’s OK. The 20 per cent that doesn’t will come to the surface and I think that is the pathway to where we want to be … the returns to experimentation have just got way, way greater.” Robots may one day help build houses faster and cheaper. But for now, the deeper problem remains unresolved.  Australia’s construction sector isn’t short of technology—it’s short of time, skilled people and strategies that turn ideas into measurable gains.  Without tackling that, Hajkowicz suggests, no amount of robots will close the housing gap.
