Architecture’s AI Shift: Faster Design, Same Humans

When Gray Puksand pitched for a Saudi Arabian project last year, the architecture firm produced a comprehensive conceptual design package in a week. Using traditional methods, the same work would have taken three to four weeks. Gray Puksand digital design technology manager Páraic Walsh, a keen proponent of AI in architecture, says “in the conceptual stages, AI is hugely beneficial to us”. “We can quickly test different iterations in a couple of minutes, so the client can explore lots of different options in the conceptual stages, which is great,” Walsh tells The Urban Developer . AI is fundamentally changing how Australian architects work, but not in the way many fear. ARC Arch_Manu Centre director and former UNSW AI Institute deputy director Associate Professor M Hank Haeusler says “AI puts the fear of living Jesus into some architects because they can see that something is happening so dramatically and they know they would be affected”. But as FK chief technology officer Jeames Hanley says “architects are not being replaced by AI—they’re being replaced by indifference”. How leading firms are using AI Gray Puksand’s exploration of AI has shifted from experimentation to strategic implementation. Working with defence and government clients, the firm developed GrAIce, a proprietary large language model with data stored in Australian data centres. The platform includes specialised models trained on National Construction Code and Australian Standards, contract review and three decades of project documentation. More than 100 of the firm’s 140 team members now use it—including founding partner Robert Puksand. “Now I can concentrate on the higher-value strategy piece,” Walsh says. When analysing 100-page tender documents, staff extract specific information in minutes—work that previously took hours. “It’s all about the data though,” Walsh says. “If we have good, clean data and information, this tool is amazing. It’s garbage in, garbage out.” ▲ Gray Puksand digital design technology manager Páraic Walsh. At Architectus, AI integration focuses on technology to optimise solar and wind performance on projects such as Melbourne’s Collins Wharf. Cox Architecture and Atelier Projects have developed similar tools. According to Autodesk’s 2024 State of Design and Make report, 34 per cent of industry leaders now use AI for sustainability—more than any other application. Beyond image generation According to Haeusler, most practices lack the strategy to implement AI effectively. “If an architect were to ask me ‘can you develop me an AI?’, the first question I would ask is: ‘for what?’” Haeusler says. “Without a ‘why’ it’s nonsensical to talk about AI.” “Hollywood misconceptions” of AI are also skewing how the technology can help or hinder the sector, he says. “AI is a one-trick pony,” Haeusler says. “It’s really good in one thing but not at all for others. Many architects think of digital as a tool but it should be a strategy.” The most valuable digital tools often aren’t AI at all, he says, they’re targeted computer programming tools solving specific problems. Hundreds of software companies, many founded by architects-turned-programmers, have identified pain points and coded solutions that help architects more than AI alone, Haeusler says. “Tasks like door scheduling in a BIM model, where you want to make sure all kinds of doors fit the thousands of spaces in a stadium or airport. You can write a script for that and it automates it,” he says. “Some tasks can’t be automated because there are questions of aesthetics. But other tasks that have a mathematical background—physics—because there’s lots of physics in architecture, can be automated.” Strategy is key Haeusler says that “with all respect, really none of the firms have successfully implemented AI yet”. While some larger firms have embedded computational design thinking, AI integration eludes those that don’t have the necessary skills or training within their operation. ▲ Collins Wharf residential development in Melbourne, where Architectus used AI to optimise building performance. A 2024 Royal Institute of British Architects report found 60 per cent of architects received no formal training in AI tools. When the Arch_Manu research centre examined Australia’s training landscape 18 months ago, it found just five CPD courses on digital topics nationwide. “And the majority of architects are in the 50s age bracket,” Haeusler says. “They may have to work for another 10 to 20 years but don’t have access to the training that architecture students have nowadays.” Since 2015, UNSW has educated 30 to 40 computational design students annually. Major Sydney firms now employ these specialist graduates, while firms in other cities largely rely on scattered self-trained individuals and smaller firms can’t justify hiring specialists when they need generalists. Recognising that individual firms lack resources to develop comprehensive AI strategies, major Australian architecture practices got together in October to form a consortium. Haeusler expects a formal public launch within months. The consortium recognises that digital transformation cannot be tackled firm by firm. Its mission is to develop shared AI strategy, governance and sector-specific tools that give a three-person regional firm the same capabilities as major practices, either for free or through subscription. Climate responsibility drives the urgency. “We believe that digital tools should be used for climate change because we’ve got the responsibility,” Haeusler says. “AI is just one component we can employ to address reduction of waste, reduction of carbon emission and reduction of material consumption to deliver projects on budget, on time and for the benefit of the community.” ▲ ARC Arch_Manu Centre director Associate Professor M Hank Haeusler. The consortium is looking to create a “distinctly Australian” approach to AI. “You can’t really copy and paste or take things from other countries,” Haeusler says. “An AI model based on European or US data—it’s metric versus imperial or the sun shines in the north versus the south, so it doesn’t translate.” The consortium would also look to address data ownership for AI training—a crisis many firms haven’t yet recognised. “Architects probably think they own their data,” Haeusler says. “But if they read the end-user license agreement, they might find out that they don’t really own the data to train internal machine learning models. For example, Autodesk prohibits using their training data for independent AI development.” The consortium will deliver AI strategy tailored to firm size, shared platforms and training infrastructure. Arch_Manu has already developed over 50 CPD courses for working architects. The climate imperative The consortium focuses on what Haeusler says is architecture’s greatest responsibility. Cities account for 70 per cent of global CO2 emissions, while construction is the biggest waste generator. The sector also consumes natural resources and energy at unsustainable rates. Haeusler’s team is developing AI systems that convert 2D drawings into detailed models documenting every component—transforming buildings into “urban mines” for material reuse. Buildings can be disassembled, then robots can filter the materials and parts to categorise and file them for later use. “Through computation, we could become urban geologists,” Haeusler says. “Instead of just prospecting for more iron, we could tell you exactly what is in the building and we could reuse that.” Human, all too human When Gray Puksand used AI to generate aerial views of a Melbourne project, it put cars on the wrong side of the road, because the model was trained on US datasets. ▲ More than 100 staff use Gray Puksand's proprietary AI platform GrAIce, cutting pitch times by up to 75 per cent. “We’ll always need human oversight to pull it back in line,” Walsh says. AI’s limitations define where architects remain essential. Trained on historical data, AI cannot invent new architectural movements. “AI is backwards-looking, not forwards-looking,” Haeusler says. “Architects and creatives have a lot of value because they can look forward or they’re in the moment.” Haeusler links AI’s impact to photography’s 19th-century arrival. When cameras threatened realistic representation, artists pivoted to impressionism, surrealism and abstract expressionism—exploring what photography couldn’t capture. The sector faces a similar moment. AI can handle calculation and optimisation but the creative leaps, cultural interpretation and human insight remain the architect’s domain. The practices thriving are those with clear strategies for deploying technology while preserving what makes architecture human. The choice isn’t whether to adopt AI but how to approach it strategically. Haeusler says “this is the moment for architects to think, to really reflect on: what are we really good at? What could we offer and where can’t we compete? Well, in maths we can’t really compete. But humans are adept at understanding other humans. “And an AI would never be able to sit in a conversation with a client and respond to human idiosyncrasies. There are subtleties that we understand as humans—for a machine, it’s debatable.” That interpersonal dimension is irreplaceable, Walsh says. “The beauty of picking up a pen and paper will always be there. AI will never replace an architect because communication and interpersonal skills are the most important factor in delivering a complex piece of architecture for a client.”

Architecture’s AI Shift: Faster Design, Same Humans

Want to read more?

From Returns to Ruin: Inside the $230m ISG Collapse

Urban Bushfire Risk Rising as Climate Changes, Cities Expand

Potential Meets Caution as NSW Pattern Book Housing Put to Test

Capital Resets as Scarcity Cycle Approaches Peak in 2026

Fears Red Tape Will Derail Melbourne Metro as Development Engine

Melbourne Long Lunch | Panel discussion with Tim Price, Jeremy McLeod, Katya Crema & Anne Michaels

A fireside chat with Troy Daffy | Managing Director, Silverstone Developments

Brisbane Long Lunch | Panel discussion with Rob Comiskey, Frank Licastro, Steve Laffey & Alexandra Stuart-Robertson

Sydney Long Lunch | Panel discussion with Jamie Durie, Jennifer Macquarie & Michael Romano

Sherpa’s Brisbane Tower Build Begins as Glitter Strip Plans Scaled Up

From Returns to Ruin: Inside the $230m ISG Collapse

Scaffolding Barons Aim Another $500m Precinct at Sydney’s West

Neue Space $55m Mulgrave Hub Adds to $350m Pipeline

2026 Developer Outlook Lunch | Melbourne

2026 Developer Outlook Lunch | Brisbane

2026 Developer Outlook Lunch | Sydney