This thesis addresses this challenge by investigating the quality of a virtual assistant built upon Retrieval Augmented Generation (RAG). RAG is an optimization technique for Large Language Models (LLMs) that enhances the quality of generated texts by incorporating knowledge from external databases. The research goal of this thesis is to investigate the quality of the retrieval and generation mechanism of a RAG system tailored to the specific domain of answering corporate process-related questions by using information contained in already existing process diagrams. RAG emerged as a promising technique when it comes to knowledge-intensive tasks such as factual question- answering.