The Architect of 2030

What should architecture schools teach when AI can already do the technical work?

By Sandeep Ahuja, Co-Founder & CEO, cove

A principal told me recently that his junior architects were still redrawing trash chute details from scratch. The profession has solved that detail hundreds of times. It gets rebuilt by hand, billed by the hour, and logged toward licensure.

That work is disappearing. On one of our projects at cove, a 17-unit multifamily building in Atlanta’s West End, the design was completed in 15 days. Design timelines dropped 60%, and early cost estimates came in at 95% accuracy. The zoning analysis alone, with overlapping overlays, historic district edges, and height transitions, would normally take weeks. Production is compressing, and it is compressing fastest in exactly the work we use to train people.

The profession feels it. RIBA’s 2026 AI Report, a survey of UK practices, found that three-quarters of practices using AI report higher productivity. Only 17% say their designs are better because of it.

Faster is arriving on its own. Better is not. The distance between the two is a training problem, and it starts in school.

The 41% Problem

To get licensed in the U.S., candidates log 3,740 hours through NCARB’s Architectural Experience Program. The largest category, Project Development and Documentation, requires 1,520 of them. That means 41% of an architect’s formal apprenticeship goes to the work AI is absorbing fastest.

Those hours were never really about drafting. NCARB’s own competency for that area includes communicating a design to clients through written, verbal, and graphic means. The competency is still right. The way we earn it is what’s expiring.

NCARB sees this. In February it wrote that AI is redefining professional competency and that education and regulation have not kept pace. Its 2026 Future Trends Report goes further and questions whether today’s licensure system can assess competency at all in a changing profession. The regulator has named the problem. The hours haven’t moved yet. Schools don’t need to wait for them.

What Gains Value

In conversations with faculty at leading architecture schools in the US, one idea kept coming up. When execution is automated, what’s left is the ability to recognize quality.

I’d add one qualifier. I trained as a building scientist, and I’ve always been the person in the room drawn to the technical side. So when I say taste matters more now, I don’t mean taste instead of rigor. A beautiful building that wastes energy isn’t good. Neither is a perfectly efficient one nobody wants to walk into. Judging both is the job.

For a curriculum, I think that means four things.

Study buildings in person. Students should see how afternoon light enters a room, how a staircase changes the way you move through a floor, and why one plaza fills up while another stays empty. One professor I spoke with is moving case studies and travel earlier in the program, so students encounter great work before they have tools that can imitate it. I grew up moving between Saudi Arabia, Russia, and Tanzania, and much of what I understand about how buildings shape lives came from living in very different ones. But travel is expensive. I’ve argued before that the old apprenticeship shut out people who couldn’t afford the grind, and we shouldn’t swap one financial filter for another. Funded travel counts. So does rigorous study of the buildings in a student’s own city.

Write about space. This means narrative writing, not specifications. Students should be able to describe what a room does to the people inside it, in language a client, a community board, or a contractor can feel. When AI generates twenty options, the architect who can explain why one is right wins the room. Few programs teach this as a core skill.

Sketch first. Several faculty described versions of the same experiment: one semester with AI tools, another with only pencil and paper, and pencil comes first. It isn’t nostalgia. You can’t catch a machine’s mistake in work you’ve never done yourself.

Develop a point of view. AI learns from what already exists, and left alone it drifts toward the average. For decades, many buildings looked alike partly because exploring alternatives was expensive. AI makes exploration cheap. That removes the excuse, and it raises the premium on having a vision worth exploring.

Judgment Without Production

Earlier this year I wrote about AI dismantling the apprenticeship model across professions. Architecture has a sharper version of the problem: 61% of RIBA’s respondents believe AI will make it harder for early-career architects to build essential skills.

It’s tempting to say reviewing AI output replaces the redline, and that is partly true. A junior marking up the machine’s drawing is building pattern recognition. But the evidence cuts both ways. When radiologists were shown wrong AI readings, even veterans’ accuracy collapsed. Review only builds judgment if you know enough to disagree. That’s the case for sequencing: pencil first, then precedent, then tools.

There’s also a reason technical depth can’t be skipped, and it’s the stamp. NCARB’s updated position this year is explicit: only a licensed architect can seal drawings and take responsibility for them, however much AI was involved. The architect of 2030 will sign drawings they didn’t draw. They need enough fluency to be accountable for them.

What Firms Owe

Schools can’t solve this alone. Much of the apprenticeship happens inside firms, including mine. If we automate the work that used to train juniors, we owe them a different path. At cove, juniors review AI output alongside a principal and join client meetings in their first year.

The Architect of 2030

Students starting a five-year program this fall will graduate into a profession where the drawings largely make themselves. They won’t be defined by their software. They’ll be defined by whether they can walk into a room and explain why a building matters: why this option and not that one, what it will cost to run, what it will feel like to live in. And then they’ll sign their name to it.

The tools will keep getting faster. Whether buildings get better depends on what we teach the people choosing between them.