A workflow for LLM-assisted IFC data enrichment to support circular reuse of large-panel building systems in Poland

This paper examines how artificial intelligence can enhance digital workflows supporting the circular reuse of prefabricated building components. Focusing on Polish large-panel housing systems from the 1970s–1980s and their European counterparts, the study proposes a workflow that integrates Building Information Modelling (BIM), parametric processing and a large language model (LLM) for semantic enrichment of IFC data. The method identifies missing or ambiguous metadata within IFC models and uses LLM-based reasoning, supported by contextual lookup tables and expert oversight, to reconstruct attributes relevant to reuse and circularity assessments. The enriched dataset is then processed through a circularity algorithm and visualised in Grasshopper to evaluate component-level reuse potential. Results show that AI-assisted metadata reconstruction can substantially improve the completeness and interpretability of digital twins for demolition, adaptive reuse and material recovery scenarios. Beyond its technical contribution, the approach emphasises the ethical and cultural importance of stewardship, highlighting how digital tools may support the responsible management of post-war prefabricated heritage and reduce construction waste. This study is among the first to demonstrate how LLMs can semantically enhance IFC models for circularity assessments in prefabricated building systems. • Identifies gaps between computation and material reuse in prefabricated systems • Quantifies technical constraints affecting large-panel concrete reuse • Introduces a workflow linking BIM, IFC and AI for circularity assessment • Uses LLMs to semantically enrich IFC data and improve reuse databases • Supports physics-based and component-level decisions for concrete reuse • Demonstrates AI-assisted metadata reconstruction for legacy prefab buildings