A short history of the end of architecture.
AI is likely to make many aspects of architectural practice faster and more efficient. But previous technological changes show that productivity gains do not automatically improve profitability. Architecture practices need to decide deliberately how AI-created capacity is used, valued and priced before increased efficiency simply becomes increased client expectation.
For most of my working life, architecture has apparently been on the verge of being transformed beyond recognition. Sometimes the predictions have been optimistic: new technology would make architects dramatically more productive, eliminate repetitive work and improve the economics of practice. At other times they have been darker: technology would diminish the architect’s role, commoditise professional expertise or render parts of the profession redundant.
Neither prediction has ever quite come true. Architecture has certainly changed, and in some respects beyond recognition. But in other ways, it remains surprisingly familiar. That long history of change may contain one of the more useful lessons for architecture practices as they consider what artificial intelligence will mean.
Architecture has always adapted
It is tempting to imagine that architecture once existed in a relatively stable state and has only recently become subject to repeated disruption. Of course, that was never really the case. Even Vitruvius described architecture as requiring a broad combination of technical understanding, judgement and practical ability. The context was entirely different, but the underlying proposition remains recognisable: architects have always needed to interpret human needs, reconcile competing interests, understand materials and construction, communicate ideas and make decisions where there is no single obvious answer.
What has changed continually is the means by which those things are done.
When I began my own career in the late 1960s, architectural practice operated at a very different speed. Drawings were produced by hand, information was largely physical, correspondence was typed, posted, filed and retrieved manually, and changes took time to communicate and incorporate. Tools such as the Rotring pen and IBM Golf Ball typewriter now seem almost quaint, but at the time they represented meaningful improvements in the speed, accuracy and consistency of work.
Then came photocopiers, increasingly sophisticated office equipment and eventually computers. Each promised to remove labour from tasks that had previously consumed substantial amounts of time, and largely they did. Yet a recurring pattern also began to emerge: the time saved did not simply remain saved.
When efficiency becomes expectation
Computer-aided drafting was perhaps the most obvious transformation of my own working life. It changed how drawings were produced, amended, coordinated, stored and distributed. Compared with manual drafting, the productivity gains could be extraordinary. A change that once involved hours of erasing, redrawing and coordination could increasingly be made quickly.
It would have been reasonable to assume that these efficiencies would flow directly through to stronger practice economics. Sometimes they did. But the market adapted as well. Once revisions became easier, more revisions became possible. As drawings became easier to reproduce and distribute, more information could be expected. As technology made architects more responsive, responsiveness itself became part of the service expectation.
The productivity gain was real, but much of its value was eventually absorbed into the normal cost of doing business.
Email followed a similar path. Before email, communication contained natural friction. Letters took time to prepare, send, receive and answer. Drawings were physically issued. Information moved more slowly because the systems available to move it were slower. Email removed much of that friction, which was unquestionably useful, but once communication became instantaneous, expectations shifted. A response that might once reasonably have taken days gradually became expected within hours. Smartphones then extended that expectation further, making the office effectively portable.
Building Information Modelling brought another wave of optimism. Better coordination, richer information, reduced duplication and improved integration between disciplines all promised genuine benefits, and many of those benefits have been realised. But BIM also required substantial investment in software, hardware, systems, training and people. At the same time, the quantity and sophistication of information clients and contractors expected from architects increased.
The recurring lesson is not that technology fails. Quite the opposite. New tools often work very well. The problem is that improved productivity does not automatically translate into improved profitability. Some of the benefit becomes better service, some becomes additional output, some becomes a new competitive baseline, and only some remains with the practice.
Not every disruption has been technological. The Global Financial Crisis changed markets, fees, procurement and commercial confidence. COVID changed something else entirely by demonstrating, almost overnight, that practices could operate in ways many had previously assumed were impractical. Teams worked remotely, client meetings moved online, interstate collaboration became easier and geography mattered less.
Some of those changes endured because they created genuine efficiencies. But they did not produce a simple productivity dividend either. Savings in travel and physical proximity were accompanied by new expectations of availability, different management challenges and new demands on communication and culture.
What makes AI different?
Much of the discussion about AI in architecture has focused, understandably, on creativity. Can it generate ideas, images, forms, options or concepts? Can it support design development or visualisation? These are important questions, but I suspect architecture practices have paid less attention to the commercial implications.
AI may prove more consequential than many earlier technologies because it does not only affect the production or movement of information. It can participate in research, synthesis, comparison, drafting, analysis, visualisation and elements of decision support. In other words, it reaches into work that practices have traditionally regarded as requiring significant professional time.
That creates a familiar but potentially more powerful version of the same economic problem.
Who gets the productivity gain?
Imagine that a task that currently takes an architect ten hours can, with AI, be completed to the same or better standard in five. The immediate reaction is to call that a productivity gain, and it is. But the more important commercial question is: who gets the value of those five hours?
The practice might retain some of the benefit as additional margin. It might reinvest the saved time in deeper design exploration, more rigorous checking, stronger client engagement or less pressure on staff. It might increase capacity without increasing headcount at the same rate.
But another outcome is equally possible. Clients learn that the work can now be done faster. Competitors begin pricing on the same assumption. Programmes tighten. Response times shorten. What begins as a productivity advantage becomes the new minimum expectation. Eventually, the five hours disappear.
That is why I think the immediate risk from AI is less about whether it will replace architects and more about whether architecture practices allow another major improvement in productivity to flow disproportionately away from them.
Architecture has spent decades becoming more technologically capable, yet few practice leaders would argue that the profession has enjoyed a corresponding transformation in profitability. That should make us cautious about assuming that efficiency and commercial value are the same thing.
The choices architecture practices still control
A practice only captures the value of increased productivity if it makes conscious decisions about what that productivity is for. Where should the time saved be reinvested? Should greater efficiency reduce the fee, improve the margin, create additional client value or some combination of the three? How should the practice explain that value to clients? And do existing fee structures still make sense if hours consumed become a less useful measure of value delivered?
Those are not really technology questions. They are business model questions.
Architects cannot control the development of artificial intelligence, and individual practices cannot control how quickly clients, competitors, contractors or consultants adopt it. But that is different from having no control at all. Practice leaders can decide where AI should and should not be used, what still requires human judgement and review, how saved time should be reinvested, and whether existing pricing approaches reward efficiency or unintentionally give it away.
Those choices matter because, if practices do not decide how the benefits of AI should flow, the market will eventually make that decision for them.
I have lived through enough supposed revolutions in architecture to be sceptical of predictions at either extreme. Technology rarely destroys a profession overnight, and nor does it magically solve its underlying business problems. Instead, it changes what is possible. Then clients change, competitors change, expectations change, and eventually the profession changes around them.
Architecture today would look extraordinary to the young practitioner I was in the late 1960s. Yet I would still recognise much of the work. Someone must still interpret what a client needs, reconcile competing demands, exercise judgement where the answer is uncertain, and accept responsibility for the consequences.
AI may change how much of that work is done, how it is done and who participates in doing it. But perhaps the most immediate challenge for architecture practices is not existential at all. It is commercial.
AI may create one of the greatest productivity gains the profession has seen. The real question is whether architects capture enough of that value to create better architecture and stronger practices, or whether another remarkable technological advance simply becomes another thing clients learn to expect.
About the author: Ross Clark is an Australian business coach, mentor and adviser to architecture practices and their leaders, drawing on more than 40 years' experience across architectural practice, ownership, leadership and advisory work.
If you’re thinking about enhancing the next stage of your practice, or want an experienced perspective on how things are currently set up, please reach out.
