Foundation models (FMs) are at the core of generative AI (genAI), but baseline FMs have a range of limitations in terms of accuracy, relevance, coherence, and domain expertise that require fine-tuning. Retrieval-augmented generation (RAG) provides an integrated approach to optimize the generative output and address these challenges, paving the way to AI agents for intelligent automation. This report, the first in a series, defines RAG, describes its business value and representative use cases, analyzes the architecture of RAG engines, and provides strategic recommendations for adoption.