The increasing use of digital learning resources in vocational education and training (VET) has created a need for effective management and discovery of these resources. Learning resource metadata, which provide descriptive information about digital resources, have the potential to improve the discovery, reuse, and sharing of learning resources. However, the use of metadata in VET poses a number of challenges. This paper presents the opportunities and challenges of using metadata for learning resources in the context of the German funding program "Innovationswettbewerb INVITE". It discusses selecting metadata and metadata standards for learning opportunities, learners, and digital credentials within the INVITE projects and how, in contrast, international experts select metadata and metadata standards. One of the challenges frequently mentioned is the effort needed to assign large amounts of metadata. This paper thus explores the use of Large Language Models (LLMs) as a tool to address this challenge.