The expanding integration of AI techniques into daily organizational processes, has transformed the way business-related questions are addressed, significantly reducing the need for technical expertise. In this paper, we evaluate the extent to which responses to stakeholder-relevant questions can be addressed using Retrieval-Augmented Generation (RAG) and its variant, GraphRAG, for information extraction from financial documents such as invoices. This study assesses the performance of these approaches at varying levels of complexity and question categories, identifying how factors such as structure, complexity, and the processes involved in the query pipeline impact their effectiveness. Particular attention is given to how Graph-RAG broadens the scope of answerable questions by structuring documents through graph representations. Guidelines are also included on how each procedure can be integrated into financial workflows, highlighting their respective strengths and limitations compared to a SPARQL benchmark approach.