University procurement generates significant but poorly quantified environmental impacts, particularly in Scope 3 emissions from purchased goods and services. This paper presents LCA@TUW, a scalable, AI-enabled pipeline that combines optical character recognition (OCR) with large language models (LLMs) to extract, classify, and map approximately 14,200 invoice line items from TU Wien’s 2024 procurement to ecoinvent 3.11 processes. Impacts are calculated using Brightway2 across 16 Environmental Footprint (EF 3.0) categories. We describe the eight-step methodology, illustrate critical quality challenges, and discuss the reliability boundaries of LLM-assisted environmental assessment. The approach enables granular, product-level environmental footprinting that conventional spend-based methods cannot provide, while maintaining traceability and auditability throughout.