AI in Architecture: Is Your Degree Already Outdated?
6 days ago
9 min read
Design students across countries are already experimenting with artificial intelligence — or rejecting it outright — while universities struggle to decide what AI belongs in the studio, and what doesn't.
One student feeds her own 3D models and material photographs into an image generator, then asks it to refine the render. Another has never opened one, and has no plans to. Both study at Toronto Metropolitan University, where artificial intelligence remains a standalone elective within an otherwise traditional design curriculum, rather than being integrated into the program.
That patchwork isn't unique to Toronto. Across interviews conducted for this piece in 2026, spanning Toronto Metropolitan University, University of Toronto, and Politecnico di Torino, the same gap shows up again and again: a global survey cited by UNSW Sydney's Business School found that 60% of architecture firms, currently using AI, are teaching themselves how to use this new technology, with almost no formal guidance from schools or employers.
The real story of Artificial Intelligence in design education right now isn't the technology. It's the vacuum around it.
Left Off the Syllabus
Natalie Yuen, a recent graduate of Toronto Metropolitan University's interior design program, remembers exactly one elective focused on AI-assisted rendering, and not much else. She used the tools anyway — for ideation, for early-stage proposals, for tracking down OBC (Ontario Building Code) sections faster than she could by hand. It sped up her process. It also left her uncertain whether the resulting work still felt fully hers, a tension she's still sitting with.
Esther Bailey, another TMU alumna now working at Toronto's Mason Studio, describes the same absence. AI was never built into her coursework. It was simply a tool some classmates picked up, and others ignored, with no instructor steering the decision either way.
Neither experience is unusual.
The Royal Institute of British Architects puts practice-wide AI adoption at 41%. The American Institute of Architects, measuring something narrower — regular, integrated use rather than occasional experimentation — found just 6%. Somewhere in that gap between "has tried it" and "actually relies on it" is where most new graduates currently land: aware the tools exist, undertrained in them, and left to close the distance on their own.
Not every school has handled that gap the same way.
The University of Miami now formally introduces AI in first-year Visual Representation courses, rather than leaving it to student initiative.
The University of Maryland is going further still: a new AI-in-Architecture minor launching this fall pairs foundational AI concepts with the practical and ethical side of using them, spread across five required, studio-linked courses — the kind of structural commitment a single elective can't match.
Parsons School of Design has gone a different direction again, treating AI less as a skill to certify and more as a habit to police. Associate professor Mark Gardner has described the school's approach as teaching AI as a tool for interaction, not something meant to "do the work for you."
The differences matter because students themselves don't agree on what the right response should be.
Natalie Yuen wants AI folded into the core curriculum, paired explicitly with a class on ethics — where it's appropriate to use, and where it isn't.
Olivia Condo, a TMU classmate, wants the opposite emphasis: less time at a screen, more time on fabrication, more real clients and fewer simulated ones.
E.L., a master's student from Politecnico di Torino, wants a course entirely on AI, on the grounds that her current training only gave her general foundations she's had to expand on the job, unsupervised.
G.S., a recent graduate from Politecnico di Torino, puts it more bluntly: none of this existed when he was a student, and universities need to move faster.
Turin's own architecture department has, in fact, already started moving. This July, Politecnico di Torino ran a dedicated AI-Enhanced Architecture summer school at Castello del Valentino, built in collaboration with the Rotterdam firm MVRDV, simulating a full design competition under AI-assisted conditions. Neither Politecnico student interviewed for this piece had taken part.
One Tool, Two Verdicts
There is no student consensus on what “catching up” should actually look like.
Olivia Condo uses AI narrowly, mostly to polish grammar or tighten a script, and stays wary of it everywhere else. She's seen AI-generated floor plans up close, and calls them undercooked: mistakes repeated, details overlooked, the kind of errors that only surface once someone with actual training reads the drawing properly. Whether prompting a bot compromises her claim to being a designer, she says, depends entirely on what it's being asked to do. "Decorating, maybe. Planning, ABSOLUTELY NO."
A.S. student from Toronto Metropolitan University goes further. She has never used an AI design tool and doesn't intend to, and finds the suggestion that it might improve her work "a bit insulting." Her program hasn't pushed the technology on her, and she's relieved. Given the choice, she'd rather her school taught SketchUp properly than any AI plugin. Struggle, in her account, isn't an inefficiency design school should engineer away. It's the actual curriculum. "You need to struggle to learn," she said. "If all you know how to do is ask an AI chatbot for help, you are a useless, useless employee."
Esther Bailey sits closer to the middle and draws her own, quieter line. She doesn't ask a generator to invent a form from nothing. She feeds it her own 3D models and material photographs, and asks it to refine what's already hers. It's a small distinction with a large implication: the difference between directing a tool and handing it a decision.
The Image Was Never the Idea
That distinction has a name in professional circles, even if none of the students interviewed used it themselves.
Professor M. Hank Haeusler of UNSW Sydney's School of Built Environment divides architectural work into "action tasks", standardized and executable, traditionally handed to junior staff, and "decision tasks", which require judgment, context, and an understanding of who a space is actually for. Rendering sits in the first category. Deciding whether a floor plan will hold up under real use sits in the second.
It's the same line Esther Bailey draws instinctively, and the same one Olivia Condo draws when she separates decorating from planning.
Blurring that line has a cost, and it shows up before graduation.
At a recent HOK Up Next panel, University of Toronto professor Humbi Song described what she calls the "illusion of knowledge": students turning in polished, AI-assisted work, then struggling to explain the reasoning behind it the moment someone asks a direct question. The gap only surfaces in conversation, after the render is already finished. Song has started bringing pencil and paper back into her classroom on purpose, as a corrective.
It's an old worry in new clothing. Long before generative tools existed, design critic Eric Cesal, a designer, educator, and author, argued that architecture had drifted into a fetishization of the image, a culture where a striking rendering could stand in for a genuinely tested idea, sometimes for an entire career. AI didn't invent that habit. It just made it available to anyone with a laptop and an internet connection, which is exactly why Cesal argues the image can no longer prove anything on its own. If a convincing render is one prompt away, producing one is no longer the achievement.
Already in the Pipeline
While schools debate whether AI belongs in the studio, the tools keep shipping regardless. Researching this piece turned up three still-emerging examples worth naming.
One, still in development, would connect Building Information Modelling (BIM) software like Revit directly to jurisdiction-specific building codes — Canadian, European, Italian, and others — automatically cross-referencing a model against the applicable code, rather than leaving that lookup to a junior architect with a PDF and an afternoon to spare. It's close to the exact task Natalie Yuen described doing manually, at the Ontario Building Code level, section by section.
A second tool promises to turn a simple sketch into a high-fidelity visual concept render, with its developers claiming time savings of up to 75% compared to a traditional workflow. That figure comes from the company, not from independent testing, and is worth treating as a claim rather than a fact for now.
A third, a recently launched product called Roomagic, lets a designer upload a single photo of a room along with a text prompt and get back several styled versions of that same room within seconds — a fast way to show a client what a space could look like before committing to one direction. General-purpose tools like ChatGPT and Gemini can attempt something similar, but with less precision, since neither was purpose-built for the task.
None of these tools replace the judgment Olivia Condo or Humbi Song are describing. They automate the action tasks Haeusler named, not the decision tasks. But they're arriving fast, and mostly outside the classroom — which is exactly the pattern the interviews above describe: the tool ships, the profession adapts around it, and the curriculum catches up last, if it catches up at all.
None of the students interviewed for this piece think their own program has caught up to that shift yet. Esther Bailey, for her part, has already found herself on the other side of it, in professional practice, where the calculus keeps changing. "People in the workplace expect people like me to be better at AI than them," she said. "Which I wasn't ready for."
A Pattern, Not a Rupture: Our Thought
Forma's position is straightforward: the AI era in design has already begun. The question is not whether studios and universities will encounter these tools, but how deliberately they choose to incorporate them. For Forma, AI is a tool a designer directs, questions, and improves upon — not a substitute for the judgment that makes a design meaningful. New tools are arriving by the week; not all of them will matter.
Giuseppe, Forma's founder and a graduate of Politecnico di Torino's Master's program in Architecture, sees the current moment as part of a much longer cycle of technological change in design.
"Looking at AI today is exactly like looking at BIM in the early 2000s, or CAD and CAM in the early 1980s. Designers were scared. Draftsmen were terrified of losing their jobs. Big firms sat lost in thought about the future. And today? CAD is the foundation, and BIM is the standard. Software like Revit, Archicad and AutoCAD are used every single day to design the world we see around us. Look around you: many of the houses, high-rises, even sheds, were designed using one of the three systems people once feared. What will happen with AI, we cannot know. But we can learn from the past: technology always finds its place in society, whether we want it to or not, and with AI, that has already begun. It's our duty to decide whether we run with it or fall behind without it." - Giuseppe, Forma's founder.
The comparison is not a prediction that AI will follow exactly the same path as CAD or BIM. It is a reminder that technological change in design rarely arrives with a finished set of rules. The tools emerge first; education, professional practice, and standards adapt around them.
That process is already underway. Forma's ongoing coverage will continue to examine which emerging tools are genuinely useful to studios and which remain largely claims, because, as this reporting suggests, students are already making those decisions for themselves.
Sourcing & Credits
This piece combines Forma's own reporting with direct interviews conducted by the team, along with data and commentary drawn from external coverage of the field.
Interviewed, named with permission: Natalie Yuen (Toronto Metropolitan University), Olivia Condo (Toronto Metropolitan University), and Esther Bailey (Toronto Metropolitan University, now at Mason Studio).
Interviewed, but asked not to be named: additional students and recent graduates from Toronto Metropolitan University, the University of Toronto, and Politecnico di Torino were interviewed for this article and requested anonymity. While their insights informed the broader scope of this piece, specific accounts appear under pseudonymized reference codes assigned by Forma — A.S., E.L., and G.S. — to fully protect their identities, as requested.
Institutional and statistical context is credited by source where it appears in the piece: Royal Institute of British Architects, American Institute of Architects, UNSW Sydney's Business School, University of Miami, University of Maryland, Politecnico di Torino, and HOK's Up Next panel series, Parsons School of Design, among others.
In support of this article, we also want to highlight the RIBA AI Report 2025 & Architectural Research: Architectural Design in the Age of Generative AI: Toward a Balance Between Efficiency and Responsibility
