How AI is being used to design cities and infrastructure

The vast majority of professionals (91 per cent) believe it is important to have ethical guidelines for AI in the built environment. Around half (49 per cent) also reported apprehension about the dominance of global tech companies in AI development.

Arup commissioned the research to explore how AI is already changing the way cities and infrastructure are designed. The firm is calling for increased focus on developing AI tools that can help decarbonise the sector and restore nature, while delivering prosperity and resilient infrastructure for a growing global population. The consultancy reports that in 2024 there was $252bn of corporate investment in AI in total and its AI experts contend if just 10 per cent of this investment was used to design AI for major challenges in the built environment, it could help radically transform cities and infrastructure to improve people’s lives.

“This survey shows the scale and pace of AI adoption in the design of cities and infrastructure, cutting through the noise around large language models like ChatGPT to show how advanced AI systems can improve people’s lives,” said Will Cavendish, global digital services Leader at Arup and ex-DeepMind strategist. “AI is already making a significant impact – empowering what we call ‘Total Design’ and helping deliver everything from renewable energy infrastructure to transport systems.

“While global investment in AI is enormous, it often misses the most impactful areas. We need AI that delivers real-world benefits – from new sustainable materials to renewed global biodiversity. Our industry should focus much more resources on developing these types of AI systems.”

AI in action

Arup is using AI-powered tools to model nature-based solutions that protect people from heatwaves and floods, and to extend the life of critical infrastructure like offshore wind turbines and bridges for decades – significantly reducing cost and carbon emissions.

For example, it is using AI to interpret satellite mapping and help cities understand how their urban design can aggravate or mitigate the effects of extreme weather. Arup’s UHeat tool uses AI-powered land mapping for rapid modelling of the urban heat island effect at a local level in cities. Meanwhile, its Terrain’s tool uses machine learning to help understand the natural ability of cities to absorb rainfall by mapping of the amount of “green” and “blue” natural infrastructure in urban centres, going right down to the level of soil types and the effect of vegetation on rainfall runoff. 

“We need AI that delivers real-world benefits – from new sustainable materials to renewed global biodiversity. Our industry should focus much more resources on developing these types of AI systems”

Arup also uses AI and digital workflows as part of Total Design to make this process more efficient, impactful and sustainable. It explains that traditional design is characterised by a linear flow of siloed activities, often involving time-consuming back and forth iterations.

For PwC’s Sanya campus in China, nine Arup disciplines worked together, underpinned by automated workflows. Designers and engineers are using generative algorithms to test design alternatives and embed sustainability across processes. At the same time, chaining together digital tools allow different disciplines to work collaboratively and simultaneously. Arup’s Shanghai teams have been pioneering these approaches, seeing 25 per cent carbon reductions in carbon emissions in concept design and saving up to 35 days in the design phase.   

The research was conducted by Opinion Matters, among a sample of 5,000 (500 per country: US, UK, Australia, Nigeria, Germany, China, Singapore, Brazil, India, and Indonesia) professionals (manager-level and above) in the built environment sector into the use of AI technologies. Respondents split evenly across four main job categories: City/urban planners; Architects; Engineers; Digital officers in the built environment sector. The data was collected between 12.02.25 and 21.02.25.