Years of studio, debt and a licence that takes over a decade. Now AI makes renders in seconds. What that means for your place in architecture.
This in-depth guide covers everything you need to know about will ai replace architects? what firms and data say. Based on verified income data and real-world case studies from our database of 138 side hustle tactics.
AI is unlikely to replace licensed architects, because every US state requires a licensed person to take responsibility for a building. It is already taking over parts of the job that juniors and drafters used to do: concept images, massing options, early code checks and documentation. Most architects in the UK now use it on most projects, and the profession's own surveys say the people most at risk are the ones just starting out.
You spent five years in studio. All-nighters before crits, a model you carried across the city in the rain, tutors who tore your work apart in front of everyone. You probably finished with debt, and with parents who were quietly proud that you would be an architect, a professional with a title. Then the first job paid less than your friends in finance, and the path to licensure turned out to be more than a decade long.
Now you watch a partner type a sentence into an image generator and get forty concept renders before lunch. The renders still need an architect to become buildings. They used to be your week.
What AI can already do in architecture
Image generators reached the top of the profession early. In April 2023, Patrik Schumacher, principal of Zaha Hadid Architects, said the studio was using DALL-E 2 and Midjourney to come up with design ideas, and when asked how many projects, he said: "Not every single project is using it but let's say most". That was three years ago, with tools that are primitive by today's standard.
The bigger shift is AI moving from pictures into the design model itself. In September 2025, Autodesk announced "neural CAD" models for its Forma and Fusion software, including one aimed at buildings. Autodesk said the models could automate ["80 to 90% of what you [designers] typically do"](https://develop3d.com/cad/autodesk-unleashes-neural-cad-3d-generative-ai-foundation-models/). Mike Haley, Autodesk's senior vice president of research, described the building model this way: "If I was to change the shape of a building, it can instantly recompute all the internal walls". That is a launch-day claim for a product announced as coming later. It also describes, almost exactly, the layout rework a junior architect spends nights on when a client moves the envelope.
Code compliance is the next layer. In July 2025, the International Code Council, which publishes the model building codes most US cities adopt, became a reseller of Archistar's eCheck after a pilot with 11 US jurisdictions that tested AI plan review against the I-Codes. The same month, Los Angeles began beta testing eCheck so people rebuilding homes after the Palisades fire could check their plans against building and zoning codes before submitting them. Mayor Karen Bass called it "an important step in the implementation of this AI tool to further expedite the rebuilding timeline."
Code review used to be a training ground. A junior would sit with the code book, check egress widths and stair dimensions, and learn the rules by applying them. When software flags those issues first, the lesson moves somewhere else, or nowhere.
A question to sit with: which part of your current week would you struggle to explain to a client as something only a person could do?
How fast architects have adopted it
The two big professional surveys tell different stories, and the gap between them matters.
In the UK, RIBA has surveyed members for three years running. In 2024, 41% of practices used AI. In 2025 that rose to 59%. In July 2026, RIBA reported that 74% of practices now use AI for most projects. Three quarters of practices said it had improved their productivity, and 57% reported a positive return on investment.
In the US, the American Institute of Architects found much lower use. Its March 2025 report found that only 6% of the profession regularly use AI for their jobs, and only 8% of firms had implemented AI, with another 20% working on it. The most common uses were chatbots, image generators and text tools. Nearly all (90%) were concerned about inaccuracies, security and transparency.
The surveys measure different things, a year apart, in different countries. The UK numbers show how quickly it can go once tools mature: from four in ten practices to three in four in two years (the 2026 question asked about AI use on most projects, a slightly stricter test). US firms tend to follow the same software vendors, so the AIA figure is best read as a starting point.
The quieter threat: clients who skip the architect
Most of the fear inside firms is about headcount. There is a second worry that gets less attention. In RIBA's 2025 survey, 44% of respondents were concerned that AI will let people without sufficient professional knowledge design buildings, 35% worried AI will threaten the profession, and 67% thought it increases the risk of their work being imitated.
Think about what that means for the small jobs that keep many practices alive: extensions, loft conversions, small housing schemes. A homeowner who can generate a convincing render and run an automated code pre-check may decide they only need someone to sign and submit. That squeezes fees at the bottom of the market, which is also where many young architects get their first projects of their own.
The wider design industry shares the unease. In Autodesk's April 2025 survey of people across architecture, engineering, construction and manufacturing, 48% said AI will destabilize their industry, up seven points in a year, and 46% of leaders said AI skills will be a top hiring priority. Both numbers point the same way for a graduate: the bar to get hired is moving, and it is moving toward people who already work with these tools. For the broader picture of which roles are changing fastest, see our guide to jobs AI will replace.
Who is losing work first
Ask the profession who gets hurt, and it points at the bottom of the ladder.
In RIBA's 2026 survey, 59% of practices believe AI adoption will lead to staff reductions across the architecture profession. A year earlier, only 18% expected AI to eventually lead to job losses. 61% agree that AI will make it more difficult for early-career professionals to acquire essential skills and experience. RIBA's write-up warns that AI "increasingly automates tasks that have been traditionally used as training and development opportunities", and that practices could become less inclined to recruit junior staff if they believe AI can do most of that work.
RIBA's president, Chris Williamson, said it directly: "AI adoption is likely to have its greatest (and potentially most detrimental) impact on entry-level and early-career roles." In RIBA's own release he asked: "how will practices support the next generation of architects?"
To be fair to the data, RIBA also says there is no clear evidence yet of sector-wide AI-induced staff cuts. These are expectations. But they come from the people who do the hiring.
The US graduate numbers already look weak. Using 2024 Census data, the New York Fed found that recent architecture graduates aged 22 to 27 had an unemployment rate of 6.8%, against 4.2% for all majors. Their early-career median wage was $60,000, and about one in four (25.7%) were underemployed, working in jobs that do not need a degree. That predates most of the tools above, so AI is not the main cause. It does mean the starting position was already fragile when the tools arrived.
| Measure | Figure | Source |
|---|
| Architects, US jobs (2025) | 124,600 | BLS |
| Architects, projected growth 2025 to 2035 | 4% | BLS |
| Architects, median pay (May 2025) | $99,280 | BLS |
| Drafters, US jobs (2025) | 180,200 | BLS |
| Drafters, projected growth 2025 to 2035 | 1% | BLS |
| Recent architecture grads, unemployment (2024) | 6.8% | New York Fed |
| All recent grads, unemployment (2024) | 4.2% | New York Fed |
| UK practices using AI (2024, 2025, 2026) | 41%, 59%, 74% | RIBA 2025, RIBA 2026 |
| UK practices expecting AI staff reductions (2026) | 59% | RIBA |
| US architects who regularly use AI (2025) | 6% | AIA |
| Average years to earn a US licence (2025) | 12.3 | NCARB |
A question to sit with: if the firm you work for expects to need fewer people, are you the person it trains, or the person it does not replace when you leave?
Drafters: the warning from the last wave
If you want to see what software does to an architecture job, look at drafting. BLS projects drafters overall to grow just 1% from 2025 to 2035, with the median drafter earning $68,090. Its explanation: "The increasing use of computer-aided design (CAD) and building information modeling (BIM) technologies is expected to lead to a decline in employment for many types of drafters." Those tools "allow engineers and architects to perform many tasks that used to be done by drafters."
Architectural and civil drafters are the exception, projected to grow 5%, while mechanical drafters are expected to shrink 6%. That growth comes from construction demand. It also shows the pattern: each new generation of software moves work from the drafter to the architect, and from the junior architect to the senior one with better tools. AI is the next step in the same direction, and probably a bigger one. Engineers are seeing the same shift in their own drafting teams, which we cover in will AI replace engineers.
Licensing protects the stamp, and the road to it is long
Architecture has one of the strongest legal moats of any profession. BLS states that "All states and the District of Columbia require architects to be licensed." Licensing typically means a degree, paid experience and passing the Architect Registration Examination. Software cannot do any of that. RIBA's president made the same point in 2026: "Professional oversight of design remains essential if tomorrow's buildings are going to be safe and sustainable."
That protection is real. It is also slow to reach. According to NCARB, the person who earned a US architecture licence in 2025 took an average of 12.3 years to do it. In 2025, nearly 3,500 candidates became architects, a 6% drop, out of just over 39,000 active candidates. The US has over 123,000 licensed architects.
So the licence protects people who already have it. The 39,000 people working toward it are doing the exact tasks AI is learning first: production drawings, renders, code checks, revisions. And their required experience hours come from that work. If firms need fewer juniors to do it, fewer people reach the stamp, and the ones who do take longer.
A question to sit with: if it takes twelve years to get licensed, what does the work in years two to six look like by the time you get there?
The business pressure behind it
AI is arriving in a profession already under strain. The AIA/Deltek Architecture Billings Index was 46.6 in July 2026, meaning more firms reported falling billings than rising ones, in a downturn AIA calls "nearly three and a half years, the longest in the ABI's history." US employment in architectural services was about 204,500 in August 2026, down from about 207,500 in July 2023.
When fees are tight and the work is slow, a tool that turns one junior's week into an afternoon is very tempting. That is how AI changes hiring even before it is good enough to change design. Firms that would have hired three graduates this year hire one, and nobody announces a layoff.
The counter-evidence
Demand for buildings has not gone away. BLS still projects architects to grow 4% from 2025 to 2035, with about 6,900 openings a year on average over the decade. The number of licensed US architects grew 6% in 2025. And the profession itself is not convinced AI improves the work: only 17% of UK respondents agree their designs are better because of AI, and 77% agree AI can never replace human creativity. Clients, planners and building officials still want a person to answer for the building.
There will be architects in 2035. The open question is how many people get trained along the way, and who pays for that training.
Where this leaves you this month
If you are licensed, your position is solid. Learn the tools well enough to direct them and to catch their mistakes, because judgment on what the model gets wrong is the skill firms will pay for. Your relationships with clients and building officials are hard for software to copy.
If you are on the path to licensure, guard your experience hours. Ask for site visits, consultant coordination and client meetings, the tasks where a person is still needed in the room. Sit your ARE divisions on schedule. Get good at the AI tools your firm uses, because RIBA's numbers suggest the junior who uses them well is the one who stays.
If your income needs a buffer, your training has value outside the office. Architecture students pay for help with software, portfolios and structures, which makes online tutoring or a focused course on a tool you know well realistic side income. If you are thinking about leaving the field, read career change at 40 first, and compare the field against our list of AI-proof jobs. For the visual side of the industry, our post on whether AI will replace graphic designers covers what image tools are doing to creative work more broadly.