Conversational, agentic AI-enhanced architectural design process: three approaches to multimodal AI-enhanced early-stage performative design exploration

Abstract

Most AI-enhanced architecture design tasks are confined to individual use and linear design flows, with a steep learning curve for mastering complex AI tools. Although advanced AI systems have shown enormous potential for improving human-AI in creative collaboration, their use in architectural design has yet to be fully explored. This paper proposes a conversational design framework for early-stage architectural design ideation and exploration, featuring a WeChat-integrated, architecture-centric multimodal agentic AI system. Leveraging the common messaging application facilitates the medium for design collaboration. A case study of a design workshop for a mixed-use renovation development in Suzhou old town is presented. Participants included undergraduate architecture students, and the instructors joined a WeChat group with the developed AI bot for design conversation and collaboration. Each student developed a design proposal, and three modes of conversational design were observed: performance-based, evaluation-based, and designerly-based. The challenges, opportunities and implications of research, education, and real-world practice are discussed through observation, count of interactions, questionnaires and semi-structured interviews with the students.