Autodesk just gave every Revit subscriber its new Forma design tools.

That includes daylight, sun hour and carbon checks at the earliest sketch stage (Engineering.com).

And practices were already moving: 59% of UK architecture practices used AI in 2025, up from 41% a year earlier (RIBA).

So AI tools for architecture students aren't a side quest anymore. Here's what's worth learning, and where each one can quietly mislead you.

How AI is changing architecture and construction work

Three things worth knowing:

  1. The big software is going AI-first. Forma Building Design, launched in April, lets you set up a real geolocated site, generate building options with facades, floor plans and units, then send the result into Revit as a native model. Autodesk also put an AI assistant into Revit as a tech preview (Engineering.com).
  2. Architects are keen but wary. 65% of RIBA respondents expect AI to improve construction productivity. But 67% worry it raises the risk of their work being imitated, and 44% worry it lets people without enough professional knowledge design buildings (RIBA).
  3. Trust is falling as use rises. Across design and make industries, 65% of professionals said they trust AI in their field, an 11-point drop on the year before. Yet 46% expect AI skills to be a top hiring priority within three years (Autodesk).

More use, less blind trust. That's exactly the gap a good graduate fills.

What uni won't teach you about AI in architecture

Architecture school is brilliant at concept, crit and craft. Where it falls short:

→ analysis arrives late. Students often add daylight and carbon studies after the design is fixed. Forma-style tools put them in at the sketch stage → pretty images win crits. An AI render can look finished whilst having no structure, no services and no budget behind it → authorship gets fuzzy. If a client asks which parts were yours, can you answer? (We dug into the legal side in AI copyright in creative work.)

The architecture student's AI toolkit

  1. Early-stage analysis: Autodesk Forma. Use it to test building options for sun hours, daylight and carbon on a real site. Check whether your uni's Autodesk education access covers it. → misleads when: you treat early estimates as final numbers. They're for comparing options, not signing off.
  2. Concept images: an AI image generator. Good for mood boards and testing an atmosphere in minutes. → misleads when: the image becomes the design. Nothing in it has to stand up, meet regulations or be buildable.
  3. Writing and research: a chatbot. Handy for summarising a brief, drafting a design statement or explaining a regulation in plain English. → misleads when: it quotes building regulations or planning policy. Always check the official document. Rules differ by country and change.
  4. BIM with an assistant: Revit plus AI features. Learn the tool properly first, then let AI speed up repetitive tasks. → misleads when: you can't tell whether what it changed in the model is correct.

The critical-thinking move: 4 questions for any AI design output

  1. What's it assuming about the site? Orientation, context, ground conditions. Wrong inputs give confident wrong answers.
  2. Could I defend this at a crit without the tool? If you can't explain why the form works, it isn't your design yet.
  3. Is the carbon number per option or absolute? Comparing two massings is useful. Quoting one figure as fact is risky. (Same "per unit vs total" trap as in AI and climate change.)
  4. Who's liable if it's wrong? The tool won't sign the drawings. You, or your future practice, will.

What to do this month

  1. Run one Forma study on a past studio project. Compare its daylight or carbon results with what you assumed at the time.
  2. Do a "render vs reality" exercise. Generate an AI concept image, then try to draw a section through it. Note everything that doesn't work.
  3. Fact-check a chatbot on regs. Ask it about fire escape or accessibility rules where you study, then check against the official guidance. Count the errors.
  4. Read the RIBA AI report summary. Quote one stat in your next interview. Most applicants won't.

More on the field in our architecture and construction hub.

so TLDR: learn the analysis tools early, treat AI images as sketches, and always be able to explain the design without the software.

Which part of your design process would you never hand to an AI tool?

Sources

Get the next issue by email

theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.