AI, Architecture & Environment: a new thesis for PCA-STREAM

Promises and paradoxes

On the agency side, the strategies adopted are varied: resistance in the name of defending “architectural intelligence” that remains in the hands of architects, total integration into an “AI-native” perspective, or other forms of circumstantial adoption. Some delegate their writing tasks—reports, architectural notes, and even competition documents—to LLMs, others boost their ideation phases with image generation, while some focus their efforts on new possibilities for data analysis, whether internal (to capitalize on a better understanding of their projects) or external (to make sense of unstructured urban and territorial data sets), in order to inform design choices.

Among the capabilities attributed to these technologies, one that often comes to the fore is their ability to contribute to a better understanding and management of environmental issues (carbon and material footprint, energy performance, and biodiversity). However, this remains largely unexplored and poorly documented, while evidence is mounting of the insane (and exponential) amounts of energy and resources—in water, rare earths, but also in space and precarious human labor—required for their development and that of the infrastructure on which they rely (graphics cards, data centers, submarine cables, etc.).

From research to practice…

At PCA-STREAM, our thinking on AI is part of the dialogue we have been engaged in for fifteen years with the worlds of research and creation. The latest issue of our journal Stream, published in 2022, was dedicated to “new intelligences,” and “artificial” intelligences were compared and discussed with those of living beings and human collectives. In 2023, we explored the question of their role in creative processes in greater depth by organizing a series of conferences entitled “AI and Creation^{The content of these discussions has been published in our new “Stream Studies” format, available in open access at the following link: https://vimeo.com/showcase/11019589. See also our Stream Voices content dedicated to this topic: https://www.pca-stream.com/wp-content/uploads/2025/07/IACREA_mise-en-ligne.pdf}“.

As with all topics that challenge our professions, we provide 360° monitoring of AI technologies. This is shared with the agency’s design teams as part of a cross-functional working group led by our research and innovation team. It covers technological advances, their applicability, and an understanding of their environmental impacts, but also includes critical reflection on the political and societal issues related to these tools, informed by social science research that analyzes the interactions between science, technology, and society.

Deployed across the entire “THINK-BUILD” continuum, our strategy starts with theoretical research and ends with technical solutions developed by our technology manager to meet the challenges posed by the uniqueness of each project (project data analysis platform, REX archiving and agency memory preservation, image generation tools trained for our specific needs, etc.). These innovations are primarily aimed at guaranteeing the highest quality of our productions, from sketch to construction site, while controlling risks (technical, legal, financial). They push us to produce ever more striking visuals by reducing calculation times, without giving in to the ease of all-out generation and with a strong command of our artistic direction. Finally, on what we believe to be the greatest challenge posed to AI by architecture (and never the other way around), namely supporting a design that is more resource- and energy-efficient and more welcoming to a diversity of living beings, we acknowledge the paradox between promise and lack of evidence, and we are mobilizing to overcome it.

… to applied research: a thesis exploring the possibilities of AI for the environmental performance of projects

Faced with what still appears to be a “technological barrier” for the sector, our response involves scientific collaboration and applied research, convinced that our agency size and market position oblige us to do so. PCA-STREAM will soon welcome a new Cifre thesis with the “AI4ARCHI” (AI for Architectural and Urban Generative Design) Chair and the Methods and History of Architecture (MHA) laboratory at ENSA Grenoble. This thesis will aim to explore and prototype devices based on different artificial intelligence technologies (machine learning, language models, generation models, genetic models, etc.) to enable more iterations guided by environmental objectives in the early stages of the design project. It will document the conditions for the validity of these approaches and the environmental footprint specific to the technology in order to put the promise of greening into perspective, while analyzing their effects on design practices, work organization, and, more generally, on projects.

Through all these initiatives, we intend to contribute, at our level, to ensuring that the AIs we adopt today help us build desirable urban worlds for tomorrow.