Retrieval-augmented generation (RAG) plays a key role in enterprise adoption of generative and agentic AI by augmenting the model’s context at inference time with authoritative knowledge. Leading vendors have released the first generation of commercial RAG offerings — but given the inherent complexity of RAG architecture, getting RAG right requires conceptual alignment on terminologies and extensive engineering efforts. This report analyzes the key challenges hindering RAG practices and provides key RAG terminologies to facilitate cross-team collaboration.