Back to blog
Editorial draft AI RAG LLM

RAG That Actually Works in Production

Vector search alone isn't enough. Retrieval, reranking, and grounded prompts — the pattern I ship for enterprise RAG.

May 28, 2026 11 min read

Most 'RAG demos' fall apart the moment they meet real documents. Here is the stack I've had success with at scale.

Chunk with intent

Semantic chunking beats fixed-window every time. Split on structure — headings, sections, list boundaries — then attach breadcrumbs as metadata.

Hybrid retrieval

Dense embeddings for meaning, BM25 for exact matches. Fuse with reciprocal rank fusion. Rerank the top 50 with a cross-encoder to get the top 5 that actually matter.

results = rrf([dense.search(q, k=50), bm25.search(q, k=50)])
top = reranker.rerank(q, results)[:5]

Ground the prompt

Cite passages inline, force the model to answer 'I don't know' when confidence is low, and log every retrieval for evaluation.

A RAG system without evaluation is a chatbot with a database.