Missing data fields and lack of metadata standards for low-resource digital collections limit both discoverability and accessibility in large-scale repositories. In this talk, I will present a case study using dissertation data from the HathiTrust Digital Library to demonstrate how incomplete metadata records can be semantically enriched using an LLM-based agent. This case study will demonstrate how metadata enrichment using LLM-based agents is particularly valuable for large-scale digital repositories, as it creates additional metadata access points that may not have been originally anticipated, especially for low-resource content types.