Generative digital tools applied to architecture: workflows put to the test by the emergence of Artificial Intelligence

« Generative digital tools applied to architecture: workflows put to the test by the emergence of Artificial Intelligence » par Arpi Mangasaryan

École Doctorale : n° 602
Sciences de l’ingéniérie et des systèmes
Architecture et études urbaines

composition du jury

Bernd DAHLGRÜN, Professeur de construction architecturale / Professor für Baukonstruktion, HafenCity Universität Hamburg – HCU – rapporteur
Andrea URLBERGER, Professeur des ensa, Toulouse, Chercheure permanent LRA, Docteur en Esthétique sciences et technologies des images, HDR (université Paris 8) (école doctorale TESC) – rapportrice
Anetta KĘPCZYŃSKA-WALCZAK, Prof. dr hab. inž. arch., Full Professor (Professor of Lodz University of Technology, Politechnika Łódzka) – examinatrice
Lernik MIRZAKHANYAN, PhD in Economics, Co-founder and CPO, BeeGraphy, Yerevan, Armenia – examinateur
Myriam SERVIÈRES, Professeure, HDR, École Centrale Nantes, AAU-CRENAU, UMR CNRS 1563 – examinatrice
Laurent LESCOP, Professeur, HDR, ensa Nantes, AAU-CRENAU, UMR CNRS 1563 – directeur de thèse
Anne PHILIPPE, ensa Nantes, AAU-CRENAU – invitée
Xavier POIRIER, Architecte diplômé d’État – Architecte DE, founder/director, Spectrum Immersive Architecture, Saint-Nazaire – invité

résumé

Architectural design develops through successive translations between drawings, models, images, simulations, datasets, and specialist knowledge. As these elements multiply, the design process becomes increasingly fragmented. This thesis investigates how Artificial Intelligence can help architects navigate this fragmentation while maintaining control over the project’s direction.

The research proposes a methodology based on project-specific AI modules, each assigned a clearly bounded task. These modules identify, process, and translate information, returning their outputs for architectural validation and reintegration. The architect remains responsible for defining intentions, selecting sources, controlling transformations, and evaluating results.

Four experiments examine human–AI interaction, conceptual exploration, workflow transitions, and territorial analysis. Together, they show how AI can support movement between representations and disciplines while preserving architectural authorship and specialist validation. A professional proof of concept with SNKH Studio further tests the methodology within active architectural visualisation workflows.

The thesis contributes a reusable approach to integrating AI through modular, architect-directed operations. It positions AI as a mapping and translation tool that supports the design process while keeping decisions and responsibility with the architect. It also establishes a foundation for future architectural AI connecting visual representations with geometry, materials, structure, environmental performance, and physical behaviour.

Mercredi 14 octobre 2026, 14h
ensa Nantes, amphi 150
Lien vers la visioconférence