What AI Is Actually Taking from Architects (And What It's Giving Back)

In early 2023, Tim Fu was three years into a job most architects would spend careers chasing — computational design projects inside Zaha Hadid Architects in London. Then he quit. Not because the work was bad, but because he'd been running the numbers on a different kind of practice: one built entirely around generative AI. His peers thought he was reckless. Two years later, Studio Tim Fu's work was in Wallpaper* and the firm was billing on three continents.

A few months later, Eric Reinholdt — a licensed architect running a one-person residential practice in Maine — did something less dramatic but arguably more instructive. He tracked every hour of his workweek for a month. The results were sobering: only 20% of his time involved actual design work. The rest was contracts, emails, invoices, and the hundred small frictions of running any small business. So he handed that 80% to AI. He got back roughly 10 hours a week. He took on an extra project per quarter. He got home for dinner.

Same technology. Two different answers to the same question.

This matters because most architects are still treating AI as a binary: either it's going to upend everything, or it's overblown. Both Fu and Reinholdt made specific decisions — not because they understood AI earlier than everyone else, but because they got clear about what, exactly, was at stake in their particular workweeks. Before the decision comes the diagnosis.

The Architect's Workweek, Dissected

Here's the question most architects have never been asked to answer directly: of everything you did last week, what percentage required a licensed human being?

Reinholdt ran that experiment and found 80% did not. That finding isn't an outlier. Eighty-six percent of architects currently using AI report meaningful time savings, with the average coming in at five to ten hours per week, according to a 2026 Chaos/Architizer survey of more than 1,200 practitioners. The tasks being automated first are consistent across practices of every size.

The first tier — already substantially automatable — covers the tasks that eat the most hours without producing the most value: visualization and rendering, specification first drafts, code-compliance flagging, meeting summaries, client email drafts, and fee proposal templates. This is what Reinholdt handed over. It's also, for most architects, the majority of the day.

The second tier is more complicated. Generative massing studies, energy modeling, client presentation materials — AI tools are genuinely useful here, but dangerous if left uncurated. Ulrich Blum, Head of Analytics Insights at Zaha Hadid Architects, put the problem precisely in a 2026 interview: "AI tools tend to homogenise output rather than reinforce identity. At ZHA, we make sure AI serves the practice's vision, not the other way around." Partial automation with no judgment applied is how a ZHA project starts looking like every other parametric render on Instagram.

AI tools tend to homogenise output rather than reinforce identity. At ZHA, we make sure AI serves the practice's vision, not the other way around.
— Ulrich Blum, Head of Analytics Insights, Zaha Hadid Architects

The third tier is where the licensed architect's moat actually sits. Site visits, stakeholder negotiation, cultural and contextual judgment, professional liability, the stamp on the drawings — these require a human with a license, accountability, and a body that can walk a site. The Bureau of Labor Statistics projects 4% employment growth for architects from 2024 to 2034, roughly in line with the cross-occupation average. That projection reflects the structural reality: the licensed principal's signature cannot be automated away.

No architect's week lives entirely in one tier. But knowing which tier each task belongs to changes how you spend Tuesday — and which skills you invest in over the next five years.

The Jobs That Are Actually Disappearing (And Why)

Before deciding what to do next, it helps to understand what's actually happening to architectural employment right now. The numbers are real, but the cause is misread.

US architecture firms shed a net 4,100 positions between the post-pandemic peak in June 2023 and early 2025, including 1,400 in 2024 alone. AIA chief economist Kermit Baker attributed this directly to firms "right-sizing operations in response to softer market conditions." The AIA/Deltek Architecture Billings Index sat at 45.6 in January 2025 — any reading below 50 signals contracting billings — reflecting eleven consecutive months of declining new design contracts. Higher interest rates killed commercial real estate deals. Institutional project pipelines dried up. The billings followed.

The tell is in the adoption numbers. Only 6% of architects use AI regularly, according to the AIA's March 2025 survey, even though 53% have experimented with it. If AI were driving the layoffs at meaningful scale, those adoption rates would need to be dramatically higher. You can't automate your way to 4,100 job losses when barely one in twenty architects has integrated the technology into daily practice.

This distinction matters because conflating two different problems leads to the wrong response. Panicking about AI disruption when the real culprit is a credit cycle means making career decisions calibrated to the wrong threat. Ignoring real AI adoption pressure because "the layoffs are just the economy" means arriving late to a shift that won't reverse when rates come down.

The business cycle will turn. AI adoption won't. Which means the window for getting ahead of the curve is open — but it isn't going to stay open indefinitely.

Two Paths, Both Documented

Architects who've already navigated this decision tend to have taken one of two routes. Neither requires blowing up an existing career. Both have documented outcomes.

The first path is integration — using AI to reclaim time and expand capacity within your current practice. Reinholdt's toolkit is deliberately unglamorous. He is not using AI to generate wild parametric facades. He uses Claude to draft fee proposals in his own voice, NotebookLM to summarize consultant reports and zoning documents he used to read cover to cover, and a custom GPT trained on his own contract templates to turn post-call notes into a first-draft invoice in fifteen minutes. The outcome is arithmetic, not aesthetic: ten hours a week returned to design work and client relationships. He used that time to take on one additional project per quarter and, in his own words, to finally write the book he'd been planning for three years.

This path is for the architect who wants to stay in their current role, current firm type, current specialty — and simply work better. The entry cost is one free tool trial and one tracked workweek.

The second path is repositioning — using AI fluency to build an entirely new professional identity. Hassan Ragab is a trained architect who recognized, around 2022, that Midjourney gave him access to a visual vocabulary he couldn't produce through conventional tools, and that his architectural background gave him something most AI image-makers lacked: judgment about scale, material, light, and cultural context. He didn't abandon architecture. He expanded what architecture could mean for him. The moment that crystallized it: in 2024, his AI-generated imagery was projected onto the Pyramids of Giza during the Forever is Now exhibition — the first artist to show AI-generated work at that venue. His practice now spans exhibitions, commissions, and workshops across three continents.

We want to explore in depth how we can use the technology of generative AI, of neural networks, deep learning, and large language models as well, in an effort to facilitate an accelerated way of designing and building, but also thinking.
— Tim Fu, Founder, Studio Tim Fu

This path is for the architect who suspects their most interesting work doesn't fit neatly inside a conventional practice model.

Patrick Chopson, co-founder of the AI-native architecture firm Cove, has framed the underlying choice this way: the question is not whether to use AI, but whether to "design the firm around the technology, or keep bolting it onto a 1990s workflow."

The adoption window is still open, but it is closing. Fifty-nine percent of UK architectural practices now use AI, up from 41% in 2024, according to the RIBA AI Report 2025. In at least one major market, AI adoption has crossed the majority threshold. Early-mover advantage is still available; early-adopter status is not.

Three things you can do this week, regardless of which path fits your situation. Time-track your workweek for five days using the three-tier framework above — write down every task and assign it a tier. The numbers will tell you where your actual exposure is. Then pick one Tier 1 task and run it through a free AI tool for thirty days; track the time before and after. Finally, read the RIBA AI Report 2025 or the AIA's March 2025 AI adoption survey — both are free and publicly available, and both will tell you where your firm sits relative to the broader profession.

The Decision Is the Thing

Tim Fu left a prestigious firm and built a studio around AI tools. Eric Reinholdt stayed in his one-person practice in Maine and used AI to get home for dinner. These are not the same story. But they share one thing: both architects decided. They didn't wait for the profession to settle the question of what AI would become. They made a bet on what it already was.

The architects who will look back on this period as a turning point aren't the ones who understood AI earliest. They're the ones who started smallest — one task, one tool, one tracked workweek — and let the data from their own practice tell them what to do next.

Pick one task from your Tier 1 this week — a specification section, a client email, a compliance check — and run it through a free AI tool. Track the time before and after. That's the whole assignment. Not a transformation. A measurement. Everything else follows from that.

The window is open. It is not going to stay open.

Recommended Tools & Resources