In light of rising demands for ecological, economic, and social sustainability in the construction sector, data-driven decision support is now essential. This article introduces a GraphRAG (Graph-based Retrieval-Augmented Generation)-based pipeline that leverages Large Language Models (LLMs) to extract and structure unstructured product information from datasheets and Environmental Product Declarations (EPDs). We then map this data into a Neo4j property graph. A custom in-house Retrieval-Augmented Generation (RAG) framework supports context-aware queries and analyses of key sustainability indicators for building materials. Through an intuitive user interface, engineers, architects, and project owners can quickly integrate economic data - such as detailed cost evaluations - alongside technical and environmental metrics. This approach widens the basis for decision making. The pipeline increases transparency in material selection and life cycle assessment. It enables more informed insights into resource efficiency and environmental impacts. Our results demonstrate that combining LLMs with a knowledge graph environment improves data quality and decision making by delivering context-relevant information in real time. Overall, the GraphRAG pipeline offers a novel contribution to data-centric sustainability assessment in the built environment and provides a robust foundation for future-proof material decisions.