Generative AI has shifted the center of gravity in artificial intelligence from model training toward data orchestration. The past few years have demonstrated that while large models are powerful reasoning engines, the knowledge they can reliably express depends less on their number of parameters and more on the quality and governance of the data pipelines surrounding them. Generative AI systems are built on models that can reason fluently, but what makes them useful in practice is not the model alone—it is the way data is collected, prepared, and continuously managed. This chapter explores the full life cycle of data handling for generative AI applications, tracing how raw information from diverse sources becomes structured, enriched knowledge that large language models can reliably use. We examine the evolution of practices over the past two years, from naive indexing experiments to mature pipelines with semantic chunking, hybrid retrieval, agentic orchestration, and continuous enrichment. Along the way, we cover how design decisions in ingestion, preparation, embeddings, retrieval, and feedback loops directly shape accuracy, trust, and compliance. The result is a framework for thinking about data not as background infrastructure, but as the operating core of generative systems.