How can architects tailor AI to improve their practice?

Whether it’s for generating early stage visuals, creating floor plans, or managing documents, four out of 10 UK architects use artificial intelligence (AI) on their projects. This perhaps surprising figure was revealed by a survey of members carried out by the RIBA earlier this year.

It was against this backdrop that the RIBA Journal assembled a group of architects and design professionals – some at the cutting edge of AI, others looking to dip their toes in the water – to talk about how they can harness the technology to maximum advantage. The roundtable discussion in association with Autodesk and chaired by RIBAJ editor Eleanor Young, also touched on the risks and knowledge gaps in current AI development.

Adrian Malleson, head of economic research at RIBA, who commissioned the AI survey of practices, pointed out that although 41% were using AI on projects, only 2% used it on all projects. This suggested relatively immature use of AI within the profession, he felt.

Most survey respondents said they would adopt AI within the next two years, citing perceptions that it would boost productivity, promote collaboration beyond the profession and help with tasks like environmental calculations. Malleson said the survey also highlighted worries about the impact on both fees and job losses. This area of concern became strongly apparent in the discussion.

Below we highlight five of the key messages to emerge in the wide-ranging discussion.

It’s all about experimentation

Experimentation emerged as a watchword, with many attendees saying that students and practices were both investigating possibilities. Phil Bernstein, associate dean and a professor adjunct at the Yale School of Architecture, who teaches on the AI theory course there, said: ‘There are two parallel things going on. One is a lot of experimentation. And the other is a lot of anxiety. There is lot of student experimentation that’s somewhat random. Our students are fearless, so they’ll do pretty much anything up to the limits of academic integrity. But it’s still extremely early days.’

Samuel Omans, senior manager, architecture industry and strategy at Autodesk, picked up on the same theme. He pointed out that Autodesk sees AI as a set of technologies which it is looking to use in a number of ways – to change the way users can search existing data, by running analysis and automation in a more effective, rapid way, to ‘help their human imagination and human practices’. Generative AI was also being explored, he said.

Think about the business case and integrate with BIM

While there was a focus on using AI for early design stage visualisations – a key use that came out both in the discussion and the RIBA survey – others like Foster + Partners were not stopping there. Martha Tsigkari, the practice’s head of applied research and development, said: ‘What we’re trying to do is identify actual business cases, and use AI so that it becomes part of our workflow and creative process.’

Fosters has, for example, developed an app called Ask F+P. Through that, Foster’s staff can interrogate the practice’s design guides by simply asking a question in natural language, rather than having to scroll through the guides to find the answer. So, if an architect wanted to find the best insulation for a commercial building, the app goes through the guides and will provide – within a level of confidence – the best answers.

Another app in development at the firm creates ideation images rather than use Midjourney, a popular AI programme that generates images from natural language descriptions. Tsigkari said she had two issues with Midjourney. First, the images generated are not related to a design created by the practice; and secondly, in order to associate anything an architect has uploaded with images he/she wants, the IP must be relinquished. The IP then stays with Midjourney, releasing any confidentiality you may have had on the images, explained Tsigkari.

‘We’re not asking the machine to imagine things for us but to help us with the visualisation of things we’ve already designed, like adding photo-realistic lighting effects or texture of materials’, she added.

Pablo Zamorano, head of geometry and computational design at Heatherwick Studios, also noted a number of benefits – from translation to image generation to rendering, and to the production of models and animations. ‘For instance, we tend to visualise ideas quickly through renderings,’ he said. I think AI already helps us get to those conversations more quickly, and with a larger pool of options, we can leverage the right one.’

Jaina Valji, architect and founder of Copy and Space, thought the future of AI was in its integration in BIM, ‘and how architects can use that information to create models and extract data from it and create parameters.’

As a case in point, Zamorano noted that Heatherwick has incorporated a GPT model into BIM workflow so it could extract information using natural language.

AI does free up time

‘The thing we’re finding most, which might seem really mundane, is just the ability for AI to free up time for people,’ said Darren Russell, AI development director, Mott MacDonald.

Mott is using Microsoft 365 Copilot – a chatbot introduced in 2023. ‘That is releasing quite a lot of time for people. And it’s really nice that they can take that time and use it however they want. So we find it improves quality, and it improves wellbeing for people and enhances their sense of satisfaction.’

That said, it wasn’t necessarily a total safe pair of hands. ‘It’s like having a really, really clever sort of intern that sometimes makes stuff up.’

AI can bring fundamental change

For Dale Sinclair, head of digital innovation, property and buildings at WSP, AI brings with it the opportunity to start designing and constructing differently, rather than making current processes more efficient. ‘I see some amazing things with generative design and so on. But I still feel that everyone’s playing to their traditional project roles and making the buildings the same way. I don’t know a single client, even those that want radical one-off buildings, that don’t want them faster, greener and cheaper. Shifting to industrialised construction and the move to net zero are really important contextual pieces because they’re happening at the same time. We have this amazing tool in our armoury now that allows us to say, wow, we can do this.’

Poor data will produce poor designs

Among the opportunities and benefits discussed, concerns were inevitably raised about the march of AI – in particular the impact machine-aided design would have on quality.

One widespread fear was that the quality of data machines learned from was not good enough. As one participant pointed out, if design information is substandard, it potentially leads to a self-fulfilling prophecy of bad buildings.

It was considered vital that the industry collectively assess how good the data is, and how it can be used appropriately, but it wasn’t clear who would resolve such issues.

Another worry voiced by the group was the issue of who questions the machines. It was felt that if people no longer learned certain skills because machines do the work for them, a knowledge gap could emerge. How to protect against this was not clear.

Fosters’ Tsigkari, commented: ‘Generative AI is going to come in architecture. That’s a given. And when it happens, how do we want to use it appropriately? That is the question.’

WHAT SHOULD ARCHITECTS PRIORITISE TO MAKE THE BEST USE OF AI?

Darren Russell, AI development director, Mott MacDonald

If you are going to take one action, it should be raising awareness across the practice.

Jaina Valji, architect and founder, Copy and Space

Everyone should switch to BIM. That’s the best way to access AI in the most useful way outside generative AI.

Adrian Malleson, head of economic research, RIBA

Put the time aside to learn about AI and play with it. Because if the profession isn’t ahead of it, then someone else will be.

Samuel Woodford, conservation and design officer, Cumberland Council

Ultimately, we’re all engaged in finding ways to improve the human experience of place. So I would just start from that and say use whatever tools that assist you in getting closer to that.

Tom Holbertson, lecturer, The Bartlett School of Architecture, University College London

Engage with AI and play with it and understand how it might be changing knowledge and what people do.

Martha Tsigkari, head of applied research and development, Foster + Partners

Think how you can you use AI as a business case for your practice. What are the problems you’re trying to solve, and how they might evolve in the next five to 10 years.

Samuel Omans, senior manager, architecture industry and strategy, Autodesk, and lecturer, Yale School of Architecture

Get ready for continued rapid change in technology. So get your house in order when it comes to data management, get things on to the cloud and structure your data in a standardised way so that everybody can leverage the opportunities.

Pablo Zamorano, head of geometry and computational design, Heatherwick Studios

Understand your shortcomings and your values, and try to see if AI can actually help. Also avoid data silos in your organisation.

Phil Bernstein, associate dean and professor adjunct, Yale University

I would find a young person in the practice and let them experiment freely with AI, identify where AI technologies can enhance your competitive advantage and collaborate with fellow professionals so we can collectively figure out how we can move this thing forward.

Dale Sinclair, head of digital innovation, property and buildings, WSP

I would encourage practices not to think about how they can use AI for design as an end, but to solve the challenges that clients are trying to solve. I think that will help us develop the business models of the future and show where AI can drive real value.

---

This RIBAJ roundtable was produced in association with Autodesk, autodesk.co.uk