How can AI help with sustainability in the built environment?

Bringing together Australia’s foremost sustainability thinkers, the Sustainability Summit explores the ideas and innovations reshaping the built environment. Across 10 expert sessions, architects, designers and engineers gain inspiration, practical knowledge and valuable connections to help drive smarter, more resilient and sustainable design outcomes.

At the 2025 Sustainability Summit, a group of architects and sustainability specialists discussed how artificial intelligence (AI) is reshaping sustainability across the design, construction and operation of buildings, and where its greatest impacts may yet emerge. Moderated by Farbod Fathalipouri, Associate Director – Design Technology and Innovation, GroupGSA, the panel featured Julian Sutherland, Head of the Sustainable Assets team at JLL, Jordan Mathers, Associate Urban Designer & Innovation Leader, SJB Architects, Chris Serrano, Principal Environmental Consultant, ARUP, Dr Donna Wheatley, Partner, Gray Puksand, and Dave Pigram, Associate Professor, UTS.

The built environment sits at the centre of the global sustainability challenge. Moderator Farbod Fathalipouri opened the session with some hard numbers on the scale of the problem: 40% of primary energy use, 38% of global greenhouse gas emission, 12% of potable water use, 40% of global material use, and 30-40% of global waste. The alarming figures that highlight the enormous environmental footprint of the building industry also present an equally significant opportunity to drive change.

So, is AI moving the needle on sustainability in the built environment?

Speaking from the perspective of his work with computational design, evolutionary algorithms and optimisation systems as against the more contemporary large language models, UTS Associate Professor Dave Pigram elaborates on how “old school AI” has been quietly delivering measurable sustainability gains for years, specifically mentioning projects that demonstrate AI’s capacity to radically reduce energy use and embodied carbon.

The first one is an adaptive lighting system in which each light fixture contains a sensor that communicates with others to negotiate appropriate lighting levels based on occupancy. Without needing traditional commissioning, the system self-adjusts in real time, achieving energy savings of around 70 per cent while improving occupant experience.

In a very different application, Pigram describes a patented design system for offshore floating wind turbine platforms, which uses topological optimisation to minimise material use while maintaining stability and structural integrity. The outcome is a reduction in material and transport costs, and a lower levelised cost of energy – both critical for scaling renewable infrastructure.

Yet another project involves an innovative air-conditioning duct system that is robotically 3D printed from recycled plastic and delivers up to 89% reduction in embodied carbon.

Across such projects, Pigram notes, reductions of 70-90% in embodied carbon are increasingly achievable against climate targets that demand 95% to stay on track.

JLL’s Julian Sutherland argues that AI is influencing every stage across the lifecycle of buildings. Property, he points out, is a data-rich sector, encompassing everything from leases to energy use and carbon performance. AI’s power lies in its ability to interrogate these vast datasets and support more robust, evidence-based decisions.

AI is also informing investment decisions right from the very beginning: whether a new asset is needed at all, where it should be located, and how changing use patterns such as hybrid work influence demand. Even before design begins, AI-driven analysis can shape briefs by evaluating whether refurbishment or adaptive reuse might outperform new construction in carbon terms.

In repositioning older assets, Sutherland says, AI enables granular insights into how people actually use buildings and surrounding spaces, tracking movement patterns, access points and connections to transport and amenities. These insights can fundamentally reshape how existing buildings are reimagined, repositioned and reset within evolving urban contexts.

In terms of design outcomes, AI-enabled tools are now analysing the structural frames of these buildings to optimise facades, daylight, energy, heating and cooling. Crucially, he notes, AI accelerates access to best practice, helping teams benchmark designs against industry-leading standards and explore more ambitious outcomes.

Even during the operational phase of the building, AI-powered analytics are helping deliver significant performance gains by simply analysing building systems and identifying inefficiencies.

However, at the end of the day, AI outcomes depend on the quality of data. Thanks to BIM systems, Sutherland says it’s possible to access information about the building, underlining the importance of having the right, consistent, data for decision-making.

“A lot of the optimisation work is to know genuinely the constraints and opportunities of the materialisation strategy,” says Pigram. By embedding fabrication constraints directly into optimisation processes, designers can ensure that generated forms are buildable, avoiding the waste and compromise of post-rationalisation.

Using genetic algorithms, for instance, thousands of design iterations can be evaluated against multiple criteria, with embodied carbon increasingly the primary objective. The ability to place material only where it is structurally needed underpins the radical gains being achieved.

Pigram also suggests that such open-ended, customisable materialisation systems, where one can link design intent to built outcomes, also enable richer engagement with culture, connecting with Country, storytelling and place.

Unlike the controlled environment, clean data from early AI, large language models (LLMs) – such as ChatGPT – which are increasingly being used in design practices, operate on messy data, making verification and reliability key concerns, says Dr Donna Wheatley from Gray Puksand.

While AI tools are proving highly efficient for research, communication, writing and image generation, Wheatley stresses the need for making the right judgement call on outputs based on domain knowledge, given the constraints of NDAs and proprietary information that limit what data can be shared.

“It is getting better and better, but we're always going to be challenged by some of these barriers to getting the information,” she adds.

Architecture & Design Article