How will AI change architecture? A sweeping transformation or something more subtle?

The RIBA Horizons 2034 future-scanning programme provides a 10-year outlook on emerging global developments, aiding decision-makers in anticipating future trends and challenges. One of the themes of these scans, Technological Innovation, explores how architecture is being affected by machine learning and artificial intelligence (AI). Moving beyond fears of AI replacement, this theme provides a broad insight into how architects might leverage technological innovation to enhance both their practice and the built environment. This article delves into some insights from these Horizons scans, considering how they might affect architectural practice in the coming decade.

While the scans discuss several emerging technologies, AI is the dominant theme. All the scans view AI as a transformative technology, likely to touch almost every aspect of the profession. They acknowledge the important role that industry-specific data will play as AI integrates. They also envision an imminent future where we will need to address how AI coexists with issues such as professional standards, authorship, liability and expert knowledge.

However, despite the consensus on AI’s significance, the scans offer subtly different visions of the future for the coming decade.

But before we look at what effect AI will have on the profession, it is useful to examine where the profession is currently at. 

Architecture has always been a field of contradictions. On one hand, it holds on to ancient construction techniques and responds to timeless issues such as shelter, public and private spaces, finance and power. On the other, it often pioneers the future, imagining and creating change. In sci-fi movies, futuristic architecture frequently serves as an indicator of the world to come.

The architectural world we live in is simultaneously primal, raw, handmade, relational, local and inefficient, while also being industrialised, abstracted, process-driven and efficient. This duality enriches the profession, reflecting the complex, contradictory nature of society itself.

Architects have often been ahead of their time, addressing topics such as climate change, green building, passive design and energy efficiency long before they became mainstream. However, the industry as a whole can be slow to adopt necessary changes, continuing to use high embodied carbon materials such as concrete and steel, and designing buildings dependent on air conditioning and fossil fuel heating.

Architecture might also serve as a bellwether for societal trends. The criticism of mass-produced housing blocks from the 1960s foreshadowed societal concerns about mass production on an industrial scale. Similarly, the Grenfell tragedy highlighted the dangers of a society overly reliant on complex processes with diffuse responsibility.

Architects are full of contradictions. We pride ourselves on originality, problem-solving, creativity, and ingenuity, yet we can be stuck in our ways, bound by stylistic trends and outdated processes. These contradictions mirror a society full of contradictions.

So, how does this complex profession meet the coming wave of AI technology, which all the essays suggest will radically transform the field?

Early adopters of AI in architecture will undoubtedly gain a competitive edge. By integrating AI tools, firms can enhance efficiency, win more projects, increase profitability or invest more time in design refinement. Examples include AI-driven software that can generate design options based on verbal descriptions or analyse historical data to improve project outcomes. These tools provide immediate benefits and set early adopters apart from their peers.

However, as AI becomes more widespread, this initial competitive advantage may diminish. The risk of commoditisation will loom as efficiency gains become the industry norm, potentially driving down fees. To avoid this pitfall, architects must differentiate themselves by offering unique products and services. This requires a strategic approach to AI adoption, ensuring that the technology aligns with a practice's core values and enhances its distinctive strengths.

From the Horizons essays, two visions emerge. One of these anticipates an automated and industrialised industry, where AI, VR, robotics, products, factory production and digital twins align in an industrialised process. Relationships and disciplines merge, communications are honed, and efficiency drives the industry. This vision could tackle many challenges and be supercharged with AI, making the architecture industry resemble the car industry, leveraging similar technologies and efficiencies.

The scenario imagines the perfect architecture machine where everything works seamlessly and autonomously. A brief is supplied and a building is delivered, perfectly responding and adapting to its users. The vision involves organising data for AI, setting up manufacturing facilities and creating standards and products that can be reused. However, it also requires industry compliance and a degree of commitment to an idealised vision.

While some practices may pursue this vision successfully, I think it is unlikely to see industry-wide adoption. Architecture may, in fact, have more in common with the music industry than car manufacturing, with diversity and individual creativity celebrated and a diverse set of tools and styles enjoyed.

As two of the essays suggest, the experience with building information modelling (BIM) offers a valuable lesson for AI adoption in architecture. BIM was introduced with a grand vision of revolutionising the industry but often fell short due to the complexity of implementation. Many firms found the transition challenging, and the expected seamless integration and efficiency gains have arguably never been fully realised.

The second vision suggests that when approaching AI, architects might do well to start with small, manageable steps that build on what works rather than attempting a massive overhaul. This incremental approach allows for practical, real-world testing and refinement of AI tools. By integrating AI in specific areas where it can provide immediate benefits, architects can gradually expand its use based on proven success. For this approach to work, we must address real-world problems, identifying current pain points and potential solutions. This is a complex task that requires real-world experience and cannot be solved by technologists alone. The best solutions will come from those who encounter these problems daily.

This process can be thrilling as you experiment with what AI is capable of and push the boundaries of the technology. Generating concept images can be fun but I have found the real excitement to be in text and data analysis. You can now use large language models, such as Claude and ChatGPT, feeding them with hundreds of pages of text and spreadsheets as prompts. This allows for complex data analysis and manipulation, using natural language as your instructions. 

My own breakthrough moment occurred when I was analysing hundreds of appeal documents for specific information about self-build. When AI came back with the results it almost felt like magic. Working in this experimental way aligns with the day-to-day realities of architectural practice. Instead of developing a single piece of ‘killer software’ that attempts to do everything, architects can use their expertise to create and adopt AI tools that address specific challenges. This tailored, step-by-step integration ensures the technology truly enhances the work of architects. 

You begin to realise that the main thing holding you back is the limits of what you can conceive of using it for. Problems that I have spent years wishing I could solve suddenly seem like they are more within reach, for example:

• A collective details library with contributions from architects and variants or combinations inferred by AI.  
• Building simply, enabled by cutting-edge technology (what many of us crave), I have been building my own house and, every step of the way, I have been thinking: how can I make this simpler and strip the process back to direct relationships with the craftspeople? Might AI help bridge a gap between formal and informal relationships? Between digital design and craft?
• New approaches to measuring sustainability. Might AI go beyond the accountancy of calculations to allow for a more intuitive and nuanced way of assessing buildings and recommending sustainability improvements?

These challenges are just a few prompts away from the start of solutions: ‘Use these 10 pages of our sustainability rules of thumb and detailing knowledge plus this outline specification for a new building. Give me the three top things to improve the building’s sustainability.’

Both the industrialised vision and the more bespoke human-centred approaches, despite their subtle differences, share an opportunity for architects to reclaim a central role in construction. As architects, we are trained to bring visions together and make sure we serve the end-users as best we can.  

Several of the Horizons scans also emphasise the potential of borrowing from other industries. As industries become specialised they tend to become niches that are difficult for people outside the industry to engage with. When a revolutionary technology comes along, boundaries between industries become more porous and innovations in one industry are more easily applied to another. Recently architecture has borrowed technologies from the computer games industry including VR headsets and real-time games engines. Construction, meanwhile, is beginning to borrow from more robotic assembly systems. 

One of the thrilling aspects of AI is the ease with which it can be instructed and the way it can translate from one form of communication to another. This means that innovations in one industry are going to rapidly spread to other industries. Not since the invention of the personal computer or the internet have we had a technological innovation that is so ubiquitous.  The challenge with ubiquitous technologies is that they are so general-use it is sometimes easy to miss how useful they might be for extremely specific tasks.  So here is a challenge: think of a task that is highly specific to what you do – it could concern historic conservation or building close to trees – and see if AI can provide even a small amount of help or value. We would love to hear what you discover or create, even if your experiments fail.        

Tomas Millar is director of the Millar Howard Workshop

For foresight on technological innovation and other subjects search RIBA Horizon 2034

Horizons 2034 sponsored by Autodesk