Algorithms & Aesthetics: Using Generative Design & AI in Architecture

In this Architecture Social conversation (around 48 minutes), Stephen Drew talks with Keir Regan-Alexander, architect and founder of the London consultancy Arka Works, about how architects can actually use generative design and AI tools in day-to-day practice, rather than just experimenting with a bit of ChatGPT.

Who this is for

Architects, architectural assistants, technologists and students who want a grounded view of where generative design and AI genuinely add value in practice, plus practice leaders weighing how to adopt these tools responsibly. Job seekers will also find practical thoughts on where AI does and does not belong in a CV, portfolio and covering letter.

Learning outcomes

By the end of this session you will be able to:

1. Distinguish generative design from generative AI, and explain where each fits in an architectural workflow.
2. Identify the categories of AI tools emerging for architecture and judge which are genuinely useful versus hype.
3. Describe how live environmental analysis, such as daylight and sunlight testing, can speed up early-stage option testing.
4. Recognise the risks of unmanaged or covert AI use within a practice and outline a more deliberate approach.
5. Apply a considered view of AI to your own career materials, using it to support rather than replace your CV and portfolio.
6. Plan first steps for introducing AI into a practice through workshops, applied research and clear ownership.

From practice to applied AI

Keir trained and built his career in architecture, spending around a decade at Morris+Company in London, the last five years as a director delivering offices, large housing schemes, health centres and education buildings. In late 2023 he went deep on AI and set up Arka Works to bring these tools into professional work.

The three legs of the stool

Arka Works sits between three worlds: architectural practice, software development, and early-stage venture and client work. Keir transfers learnings across all three through transformation planning, workshops with practices, and applied research and development, combining less familiar new tools with real project experience to judge what is and is not useful. He frames it as a creative consultancy, noting that this space is computer science but also genuinely creative.

Carbon as a live design driver

One collaboration is with Preoptima, a start-up using generative tools for early-stage carbon estimation. Keir points to growing client appetite to pick these tools up for early business planning and viability, bringing performance data into the design conversation sooner.

Bridging craft and computation

Keir describes himself as a haptic, model-making designer who values handmade craft. His interest is in bridging that tactile design culture with computational and AI methods, rather than replacing one with the other.

A live look at option testing

Walking through a real research project, Keir shows design options being tested quickly, then runs live daylight and sunlight assessments in Autodesk Forma to get fast feedback on how massing choices perform. The aim is better-informed iteration at the stage when decisions matter most.

Reading the tool landscape: real versus hype

Keir shares a map of the many AI and generative design tools now competing in the sector. He cautions against chasing every brand and draws a comparison with earlier hype cycles such as NFTs and the metaverse, encouraging architects to look past the noise and focus on tools that solve real problems.

Design is not just prompting

For Keir, generating an image from a prompt is not design. He argues for thoughtful deployment: understanding physical constraints, keeping professional judgement central, and using AI to support a design process rather than to shortcut it.

Avoiding covert AI use

Keir's hunch is that there is already a lot of informal, covert AI use inside practices. His view is that firms should know what is being used and how, so that adoption is deliberate and managed rather than hidden and risky.

AI, your CV and getting hired

On careers, Keir is clear that a CV and portfolio are the most important documents an architect has, and AI will not do that job for you. Where it can help is the covering letter: still written by you, but a good place to use AI to sharpen a fluent summary of your experience. He draws on his own time reviewing applications in practice to underline how quickly readers form an impression.

Key terms

• Generative design: computational generation and evaluation of many design options against set goals and constraints.
• Generative AI: AI models that produce new content such as images or text from prompts and training data.
• Autodesk Forma: an early-stage planning and design tool, used here for live daylight and sunlight analysis.
• Large language model (LLM): an AI model trained on text, the basis for tools such as OmniChat.uk.
• Applied research and development: practical, project-led experimentation to test what new tools can do in real work.
• Early-stage carbon estimation: predicting a design's carbon impact early enough to influence decisions, the focus of Preoptima.

Reflective prompts for your CPD record

1. Where in your current workflow could rapid option testing or live environmental analysis change a design decision?
2. Which AI tools in your practice are being used informally, and what would a deliberate, managed approach look like?
3. How would you use AI to support, not replace, your CV, portfolio and covering letter?

About the guest

Keir Regan-Alexander is an architect and founder of Arka Works, a London consultancy applying generative design and AI to the built environment. Following a directorship at Morris+Company, he founded Arka Works and went on to co-found the LLM platform OmniChat.uk and launch the learning platform Omni Accelerator. Find out more at arka.works.