01 October 2026 | StageOne, Zurich

AI increasingly comes in flavours defined by what it can do: Spatial AI perceives and reconstructs the world, Physical AI helps robots act in it, and Generative AI's language and world models dream in text and pixels. All matter to architecture, engineering and construction (AEC), yet none was built around its particular challenges.

Buildings and cities form the largest physical system humans make. They must satisfy physics and regulations, combine material and construction knowledge with design judgement, and evolve over decades. Much of our work is not creating new buildings but understanding, maintaining and transforming existing ones. Yet knowledge about them is incomplete and fragmented across drawings, models, point clouds, photographs, sensors and documents, produced by many different actors.

What would an AI native to this domain need to do? It would need to infer what cannot be seen, combine spatial cognition with reasoning across representations and regulations, generate alternatives and predict their long-term consequences, learn from precedent, coordinate many actors, and connect design reasoning to fabrication, robotics and construction. Such capabilities could help us use less material, adapt existing buildings, make construction safer, and understand and preserve places that have never been fully documented.

This workshop, organised and facilitated by Design++, ETH Zurich, asks whether AEC needs its own flavour of AI, or whether these challenges can be addressed by combining existing paradigms. Through short inputs and discussion, we will identify the capabilities, research problems and opportunities that should define AI for the built environment.