This chapter provides a comprehensive examination of the architectures and challenges associated with the deployment of conversational agents aimed at enhancing user interaction. It addresses core components of such systems, including user intent recognition, dialogue state tracking, and retrieval-augmented generation (RAG) for supporting system transparency and explainability. Additionally, it considers strategies for the systematic collection of user feedback to improve iterative system refinement. A particular focus is placed on the role of Large Language Models (LLMs) in enabling these capabilities, especially in the context of extracting and summarising information from document collections to support user comprehension of underlying methods and algorithms. The chapter further evaluates the suitability and limitations of various intrinsic and extrinsic metrics for the assessment of conversational agent performance. As a practical case study, the chapter illustrates the application of an LLM-powered RAG framework to generate explanations for synthetic image detection tools, alongside mechanisms for capturing user feedback regarding model outputs. The chapter concludes with a discussion of current limitations and identifies promising directions for future research in the domain of retrieval-augmented generation for AI explainability.