The architectural design process spans a complex, iterative value chain, requiring diverse expertise across conceptualisation, technical development, and building use. While recent advances in machine learning (ML) have sparked growing interest in its application to architectural design, most research remains narrowly focused on task-specific automation in early-stage design. This paper addresses a critical gap by systematically mapping ML methodologies—alongside computational design techniques—onto architectural processes and principles, with a focus on the RIBA Plan of Work. Emphasising the conceptual design phase (Stage 2), the study reveals both a disproportionate research emphasis on early-stage applications and a lack of integration across later lifecycle stages such as construction and post-occupancy. Through a synthesis of quantitative and qualitative data, the paper identifies methodological imbalances, sustainability blind spots, and risks posed by opaque ML systems. It argues for a more holistic, transparent, and lifecycle-oriented approach to ML in architecture—one that supports creativity, decision-making, and design agency throughout the built environment’s full trajectory.
Keywords Machine learning, Computational design, Design value chain, Architectural design process, Spatial planning