Architecture doesn't need AI to transform it; it needs AI to make it calmer

At almost every industry event, work meeting, or doom scroll of social media, someone seems to be asking the same question: "So… is AI going to take our jobs?" 

The response is often a nervous laugh. The repetition, though, tells you the anxiety is real. Architecture is already under pressure: shrinking fees, compressed programmes, complex code requirements, a deepening housing crisis, and growing client expectations for faster delivery at lower cost. 

Into that mix arrives a technology that promises to be faster, cheaper and tireless. Of course, people are worried. 

But look past the headlines and a considerably calmer picture emerges than the conversation would suggest. The most interesting thing about AI in practice right now is not what it can generate. It's what it can remove. 

The invisible work 

If you have only scanned social media, you could be forgiven for thinking AI in architecture is limited to text-to-image tools: type in a prompt, receive a glossy render. Those tools are improving quickly, and they have a role. But they are the front cover, not the book. AI's main role is not to do our creative thinking for us – humans do that better than AI can. 

Inside a live practice, the real opportunity sits in the slow, repetitive work that consumes hours: sorting and renaming files so a team can find things; copy-pasting the same paragraphs into planning reports and design statements; manually checking models for inconsistent door ratings or missing fire compartments; writing the first draft of a specification clause that everyone knows will be standard; summarising 200 RFIs to work out where coordination really went wrong. 

These tasks are necessary. Though they are not what most of us went to architecture school to do. AI is well suited to this terrain – reading, sorting, drafting, checking – not because it is creative, but because it is tireless and indifferent to boredom. 

To let AI take the lead on what we do best would be misguided. Across creative industries, AI's value lies in freeing up time for more of what humans do best: think, question and create. 

In our own work at Warren and Mahoney, we have started using narrow AI tools in exactly this unglamorous territory. The noticeable change has not been a dramatic drop in headcount. It has been a calmer tempo to projects: more of the day spent in real design conversation, and less lost to hunting, sorting and re-typing. That may not make for a dramatic keynote slide, but it makes a very real difference to the quality of working life in a practice. 

The amplifier, not the author

AI behaves less like a genius and more like an amplifier. If your systems are fragmented and opaque, it will magnify that. If they are collaborative and reasonably disciplined, it can help you go further. 

This is where architecture and AI intersect in a way that matters. Most practices do not suffer from a lack of intelligence. They suffer from friction: re-entering the same information across multiple systems; fighting inconsistent models; losing hours to searching for "final_final_v6.dwg". When those systems are messy, AI has very little to work with. 

The practices that will benefit most from AI are not necessarily the ones with the biggest technology budgets. They are the ones with the clearest data, the most coherent processes, and the most honest relationship with how their information flows. As an AI Lead, I've made deliberate choices to direct AI first to the places where it can raise the quality and comprehensiveness of our work, the operational and administrative layers that consume professional attention without contributing to design outcomes. 

The ambition is not speed or cost. It is accuracy, thoroughness, and ensuring that the highest and best purpose of our people's expertise, the design intelligence, the spatial judgement, the client relationships, has the most room to be expressed. When our architects are not chasing 

inconsistencies, re-entering data, or searching for precedents that should be at their fingertips, they are doing what they trained to do. That is where their attention belongs. 

None of this replaces an architect. All of it allows one to do the work with greater care. 

The human stakes 

There is a risk, of course, that AI is used purely to do more of the same, only faster. That would simply accelerate burnout and impose a kind of mechanical rigidity on the work. The harder and more interesting task is to be intentional about the time dividend. 

Because there are things AI will not do. It will not decide what kind of cities we want to live in. It will not determine whose voices belong in a brief, or how a building should make you feel when you walk into it for the first time. It will not navigate the relationship between a grieving client and a heritage building, or find the design move that resolves a twenty-year neighbourhood dispute. Those choices remain stubbornly, wonderfully human. 

What we are building 

Warren and Mahoney is continuing to invest in our AI programme with care and intent. We are not doing this because AI is fashionable. We are doing it because the evidence, across our own experience and across the sector, is clear that the practices approaching AI with genuine strategic purpose now will be best positioned to serve their clients and their people through the decade ahead. 

We, and the businesses best adopting AI, are not asking our people to become technologists. We are asking them to remain what they are. Curious, rigorous, client-focused, and to help us direct AI to the places where it can genuinely support that. The technology is not difficult. The clarity about what matters is.

That is, in the end, what good architecture has always required.