For two years we wondered how smart these tools were. Maybe it was the wrong question. The right one is another: how vigilant do we remain while we use them?
Meanwhile, artificial intelligence has entered architectural firms almost without warning, and it has done so by trial and error. A tool tried and then abandoned, a prompt copied from a tutorial, an image that surprises but that you don’t know where to place in the project. Almost every week a new tool arrives that promises to change the way we work, and the most widespread feeling is that of chasing novelty.
Because you need a method, not just new tools
The question that interests us, then, is not What it can be achieved with AI – many have already shown that – but When use it, at what stage, with what objective, and how to do it without giving up control of decisions.
It’s a distinction that changes everything. The guides, ready-made prompts, collections of examples tell the result. They rarely talk about the process. Yet it is in the process that a project takes shape: in comparisons, in checks, in second thoughts. A tool that generates twenty alternatives in a minute didn’t do the difficult job. The hard work is choosing one, and knowing why.
Then there is a less obvious aspect, which affects the very nature of these instruments: they are not neutral. A generative model works by statistical approximation: it tends to return what is most probable and most frequent, gravitating towards average and predictable solutions. If you rely uncritically, the risk is not a sensational error, but a silent flattening: correct and impersonal projects, indistinguishable from other products in the same way.
The designer’s first task is to recognize this tendency; the second is to counter it, asking the instrument to question its own hypotheses instead of confirming them – treating it as a critical interlocutor, not as an author.
How the volume is organized
Hence the structure of the book, which follows the project as it unfolds. A first part is dedicated to the method, because criteria are needed before the tools: understanding the role of AI in the different phases and choosing based on the objective, instead of accumulating software that then remains unused. The central part concerns practices – generating images and visual narratives, developing spatial alternatives, writing technical texts, automating work with agents. The last part addresses strategy: how to introduce AI into a practice gradually, starting from a recurring task and building stable procedures, without breaking the workflows that already work.
Along the way we collected case studies from Italian and international studies. From them we acquire observable practices, not examples to imitate: real choices, with the frictions and corrections that each adoption brings with it. Often it is a step gone wrong that reveals how a tool really works.
The responsibility for the choices remains with the designer, and must be exercised at each step. Especially in the intermediate moments, the less visible ones, where relying on the tool would be more comfortable and where instead the quality of a project is decided. It is in those choices, not in the final outcome, that a project becomes recognizable as one’s own.
