Automated building component alterations driven by LLM-formalized human strategies to achieve code compliance

Ensuring code compliance remains a critical challenge in building design, particularly when navigating interdependent requirements. While Automated Compliance Checking (ACC) systems identify violations, they offer limited guidance for issue resolution, leaving post-checking design adaptation as a laborious process. To support the search for feasible design solutions, this paper proposes an approach that automates component-level alterations toward compliant alternatives. This approach leverages Large Language Models (LLMs) to formalize human-authored improvement strategies into building component operations, ensuring that the alteration scope remains governed by designer input. Topologically related components associated with the violations are queried to instantiate executable alterations. The method is demonstrated through a case study focusing on International Building Code (IBC) spatial requirements. The approach identifies feasible alterations and resolution clusters to support code-compliant design exploration across interdependent compliance issues. With its potential to resolve cross-system computational challenges, this framework offers a foundation for more integrated design-authoring and compliance-checking workflows.