Curating value through AI

In the studio we’re navigating our way through AI with a simple aim: to give our clients better value. That means clearer communication, faster iteration when it helps, and high-quality outputs delivered to realistic timescales, without losing accuracy or judgement. Like most shifts in the profession, it’s not a question of adopting technology for its own sake, but working out where it genuinely improves outcomes, and where it can quietly undermine understanding.

A memory keeps coming back to me from my student days.

In a crit, Charlie MacCallum said something I’ve never really shaken:

“You don’t draw drawings, you discover them.”

At the time I heard it as advice about process: don’t force a scheme, let the work teach you. Now it feels relevant in a new way, because we’re in a moment when drawings, images and videos can be produced faster than they can be understood.

Charlie’s method was disarmingly simple. He would pick on three scales: the site plan, a portion of the plan, and then one detail. It wasn’t performance. It was a test of coherence, and a reminder that if you draw it, you should understand it. Not just what it looks like, but why it is the way it is, how it works, and what it asks of the people who will use it.

That feels like a useful lens for the AI era, because AI makes it possible to generate an enormous amount of material quickly. The interesting question is not whether this is “good” or “bad”, but what it does to the balance between making and understanding, and how we protect accountability as we move faster.

Curating visuals

In the early stages of a project, AI can be genuinely helpful. When the brief is still moving and the site is still being read, rapid visuals can help clients respond, help teams test options, and help everyone align around intent. In that sense, AI can support the earliest, most productive conversations, and it can compress time without compressing ambition.

But there’s a subtle shift that comes with it. Instead of drawing your way into an idea, you can find yourself choosing between outputs. The work becomes curatorial: steering, editing, selecting, and iterating. That isn’t lesser, but it is different. Curation is not the same as discovery, and it doesn’t always produce the same depth of understanding.

For clients and collaborators, that distinction matters. The most persuasive image is not always the most robust solution. The value is in knowing which is which, and being able to explain why.

An evolving process

As projects progress, our approach becomes more disciplined. When information is fixed and decisions carry cost, programme and delivery consequences, we treat AI less as a generator of ideas and more as a way to strengthen communication and representation.

In practice that means keeping the underlying project work grounded in accountable information: coordinated models, clear assumptions, and a transparent chain of decisions. We will use AI-assisted tools where they genuinely help, for example to refine or enhance outputs, to speed up iteration, or to communicate a proposal more clearly. But we are careful about the line between interpretive imagery and something presented as “real”. That line matters as much to a developer as it does to an architect.

This is also why understanding remains central. The ecosystem around AI is still evolving: questions of verification, traceability, and responsibility are becoming clearer, but they’re not fully standardised across design, procurement and delivery. In that gap, the obligation to understand what is being shown, and what it implies, becomes even more important.

What would Charlie think now?

I don’t think Charlie would simply reject AI, and I don’t think he’d be seduced by it either. I suspect he would do what he always did: insist on coherence across the three scales, and keep asking the same unglamorous question.

Why is that line there?

That question doesn’t forbid new tools. It just refuses to let speed replace understanding. As AI becomes part of everyday practice, we may find ourselves acting a little more like curators than makers. If so, the challenge is to make that curation intelligent: grounded, accountable, and connected to real experience.

That is the approach we’re taking at Coffey Architects; using AI to move faster where it genuinely helps, while keeping the work anchored in the clarity, judgement and understanding that clients ultimately rely on.