There is a generation of architects, mine, who learned by hand.
We traced over drawings with ink. We built models out of cardboard and balsa wood and occasionally despair. We spent hours on a section that turned out to be wrong, then hours more understanding why it was wrong, then more hours still fixing it. Nobody photographed that process. Nobody posted it. It just happened, slowly, privately, in the way that learning actually works.
That time, the wasted time, the failed time, the time spent being wrong, was not inefficiency. It was the education. The mistake was the lesson. The correction was the mastery. You cannot shortcut it. You cannot prompt your way through it.
I remember a moment from my second year, 1996. A colleague of mine, during the final exam of Architectural Composition, was torn apart by the professor in front of everyone. She had missed one line. A single line on the perimeter of a building, one stroke absent from months of work, from drawings rolled and carried with care, from models built late into the night. The professor did not overlook it. The exam failed. She gathered everything, the drawings and the models, and ran out of the room in tears. One line meant repeating the entire course. One more year. Lost.
It was brutal. It was also, in its way, honest. Because architecture is brutal. A wall in the wrong place, a beam miscalculated, a threshold that doesn’t work, these are not abstract errors. They become someone’s life. The rigor of that classroom was a rehearsal for the rigor of the real.
Today, students use AI for everything. Not because they are lazy, they aren’t. But because the world no longer gives them permission to be slow. There is no time to think, no time to fail, no time to sit with a problem until it opens. Architecture, like any discipline worth mastering, requires exactly that time. It requires the chance to be wrong without the world watching.
And the world is always watching now.
We have built a culture in which you must be the hero at every moment, sharing every meal, every thought, every project in its most polished and finished form. There is no room for the sketch that goes nowhere. No room for the ugly testing or the incomplete model. No room for the student who spends three weeks on something that doesn’t work, and learns more from that failure than from anything else in their degree.
So we are arriving at AI-assisted design not entirely by choice. We are arriving there because we have already made it impossible to learn the other way.
My generation is probably the last to have drawn by hand as a matter of course. The last to have held a building in our hands before it was built, in wood, in card, in foam, and understood something about it that no screen can give you. When AI surpasses us, and it will, who will be left to define what good design actually means? How do we trust a machine to make judgments that we ourselves are no longer training humans to make?
This is not a complaint about technology. It is a question about inheritance.
And it is worth asking before we no longer can.
Most conversations about AI in architecture are about speed. How much faster can we produce drawings. How quickly can we change ideas. How easy it is to generate options, endless options, variations, images that look like architecture but carry none of its weight. That is phase one. The tool is still a tool, and it still needs a mind behind it.
We could debate that particular use for hours. But that is not the point of this article.
The interesting question is: what happens when the AI output is actually good?
Not “good enough.” Not “a useful starting point.” But genuinely, undeniably, professionally good, informed by every building code, every regulation, every climate dataset, and every precedent ever documented. Better-informed than any single architect could ever be, working faster than any team.
We are not there yet. But the trajectory is clear.
So I want to ask the question that nobody in our profession seems willing to ask out loud:
When AI can design better than us, what are we actually for?
Hello from 20 years from now.
The robots did the work. The AI did the design. Construction time halved. Errors dropped to near zero. Code compliance: perfect. Every building optimised for energy, acoustics, structural efficiency, fire egress, sustainability, contextualisation.
Earth is a mess. The climate crisis wasn’t solved; it was managed, adjusted to, and adapted around.
And the skylines? Full. Productive. Profitable.
There are still, unbelievably, those who commission “thematic” buildings: Victorian revival, Neo-Classical reminiscences. Fake, of course, but beautiful enough for those rich enough to approve them. As ever, money has not learned to buy taste. AI trained on historical data, weighted toward client preference datasets, optimised for market absorption rates. Beautiful, in a way. Completely out of context. Irrelevant to the people living in them.
Architecture won every efficiency metric and lost the only argument that mattered.
The buildings are technically excellent and humanly empty.
Wish you were here.
Here is what that postcard reveals: the real danger of AI in architecture isn’t that it replaces architects.
It’s that it replaces architects without replacing what architects were supposed to be doing.
Because we have always had two architectures running in parallel. One that asks what should this building mean, who is it for, how does it change a place? And one that asks how do we build it faster, cheaper, and more profitably?
AI doesn’t resolve that tension. It supercharges the second one.
If architecture in 2026 is already too often driven by speculation, by land value extraction, by the caprice of capital, then handing the tools of production to AI without changing the economic logic doesn’t fix anything. It just removes the last friction: the slow, expensive, occasionally obstinate human who sometimes said wait, this isn’t right.
That friction had a name. It was called professional judgment.
But I want to be honest, and also, just for a moment, a dreamer.
We have lost a great deal already. Religious patronage gave architects the Duomo. Ideological vision gave us the Bauhaus. Civic ambition gave us public housing that was genuinely meant to be beautiful. We have traded all of that for market logic. Too much of what gets built today gets built because money said yes, not because a community needed it, or an idea demanded it, or a place called for it.
I wish it were otherwise. I wish we had the equivalent of those early modernists, who built from principles, indestructible ones, the kind that hold even when budgets are tight and clients are impatient. We have no Mies van der Rohe. No Le Corbusier. No Frank Lloyd Wright. And not only are we lacking people with a straight spine, we are also lacking the shared belief that drove them: that architecture could change how people live, not just where they live.
So when I ask what AI will do to our profession, I am really asking a harder question: what purpose will it serve? A tool amplifies whatever intention is behind it. If we give AI to a profession that has forgotten why it builds, we will get more of the same, faster, cheaper, and emptier. But if we find our way back to the dreamer’s position, building for community, for ideals, and for a vision of human dignity that survives the market, then perhaps AI becomes something genuinely interesting. A collaborator in a project worth having.
I’ve spent my career arguing that the concept phase is the most important part of any project and the most consistently undervalued. Clients want to skip it. Developers find it abstract. The market rewards speed.
I have been asked to produce construction documentation without a concept. No brief, no direction, no question answered. What am I supposed to draw? It is like being the captain of a boat and being told to sail without a map, without a route, without a destination.
AI might, perversely, prove me right.
After years of overproduction and image addiction, we may finally be ready to feel the emptiness of concept-less architecture. When production becomes free, when the drawings cost nothing, and the options are infinite, the only scarce resource left is the quality of the question asked before any of that begins. What are we doing? What is this for? Who does this serve?
That is not a prompt. That is a process. And it still requires a human being sitting across from another human being, earning trust slowly, and taking responsibility for what gets built.
Elon Musk says work will be optional in 10 to 20 years. Perhaps. I was taught that work dignifies a person. Without purpose, without the weight of responsibility, we risk feeling not liberated but simply replaceable. I am more interested in what remains meaningful than in what becomes optional.
Architecture has always been the art of making human life more dignified through space. I am not afraid of AI. I am not afraid to share knowledge with it or to collaborate with it. An intelligent, well-trained AI design assistant is, frankly, a delightful thought. But not without humans. Because architecture without humans is just construction. And for construction alone, a database is sufficient.
Now remember the missing line.
We are not missing one line. We are risking missing the entire project. And there is no professor, no external authority, left to fail us and send us back to begin again. It is up to us now to maintain the standard. To ask the question before the first drawing is made. To insist that there is a concept before there is a building.
That decision, made at the very beginning, before the first line is drawn, is the work that will remain.
At least, that’s the future I’m designing for.
Mauro Moro is an architectural concept designer based in Lisbon. He works at the intersection of strategic design and spatial thinking.