A BIM-Based AI Agent for Early-Stage Seismic Eccentricity Assessment and Design Decision Support: EccenAI

Architectural decisions made during the early design stage can substantially influence the mass–stiffness distribution of a building, yet eccentricity-related seismic performance is typically assessed only after major design decisions have been established. This study presents EccenAI, a Building Information Modeling (BIM)-based artificial intelligence (AI) agent developed as an Autodesk Revit plug-in to support early-stage eccentricity assessment and design decision-making. The framework integrates deterministic rule-based calculations with large language model (LLM)-assisted interpretation and design guidance. Geometric, physical, and structural data are extracted through the Revit API to calculate the center of mass (CM), center of rigidity (CR), and normalized eccentricity. The framework was evaluated as a proof-of-concept using square, rectangular, and L-shaped floor-plan configurations. The square plan yielded eccentricity ratios of 1.93% and 0.61% in the x- and y-directions, respectively. Following AI-assisted design revisions, the rectangular plan reduced its critical x-direction eccentricity from 23.32% to 9.70%, while the L-shaped plan reduced its critical y-direction eccentricity from 23.07% to 15.80%. The results demonstrate the feasibility of integrating deterministic BIM-based assessment with LLM-assisted reasoning and validated model modifications as an early-stage decision-support approach rather than a substitute for detailed structural analysis.