The Problem & Solution
AEC projects often struggle to calculate embodied carbon before construction because reconciling data across BIM models, spec books, and quantity takeoffs is a time-consuming, manual process. To solve this, a two-agent AI system was built using CrewAI and Anthropic Claude Sonnet. It automatically ingests IFC models and project documents to instantly produce a defensible sustainability report, shifting carbon analysis from a post-design compliance checkbox to an active design driver.

How It Works
The architecture relies on sequential AI agents equipped with five custom tools. First, a BIM Quantity Analyst agent extracts element volumes from the IFC model and cross-references the spec book to map materials. Then, an Embodied Carbon Specialist agent applies ICE Database v3 coefficients to calculate total emissions, identify material hotspots, and generate a final DOCX report complete with prioritized mitigation actions, ROI, and estimated costs.

Outcomes & Deployment
When run end-to-end on a 186 m² residential model, the system successfully generated a comprehensive 14-page report identifying rebar steel as the primary carbon hotspot and proposed actionable mitigations to reduce total embodied carbon by 17.7%. This auditable pipeline eliminates manual data reconciliation and can be retargeted to any AEC firm's standard IFC models and project documentation within a single working week.