Imagine you're a customer of some software product, asking the company's AI support chatbot on their website: "We're still on the 2023 version — if we upgrade to the new release, will our historical data and custom settings carry over?"

The answer to a question like this is scattered across the company's release notes, migration guides, and support manuals — material a general-purpose large model was never trained on. So in practice, the approach is: organize and store these documents ahead of time; when a user asks a question, first "look up" the most relevant passages, then hand those passages to the model along with the question, so the model can compose an answer "based on the material."

This approach — "look up the material first, then have the model answer based on it" — is known in the industry as RAG (Retrieval-Augmented Generation). It's currently the most mainstream, and the easiest starting point for understanding how LLMs get applied in practice, so we'll use it as the throughline for this guide.

This guide starts with the RAG pipeline, works from the surface inward, then expands outward to Agents, compliance, and where RAG sits within the broader landscape of LLM applications.

If you're someone who can't avoid LLMs at work but doesn't need to write code by hand — a product manager, project manager, pre-sales or solutions engineer, founder, or a manager who needs to work with AI projects — what you need usually isn't runnable code, but the kind of understanding that lets you talk with engineering teams, evaluate approaches, and explain things clearly to clients. This guide is written for exactly that: by the end, you should be able to follow how a real LLM system actually works, keep up with the relevant technical discussions, and judge whether RAG or some other approach fits a given need.

If you're just starting out technically, it will help you first build a complete conceptual map, so that later, when you read framework documentation or start writing code, you won't be stuck seeing the trees but missing the forest.

No math or programming background is needed anywhere in this guide. The content moves from simple to deep: the early chapters start with everyday examples, and later chapters move closer to how the system actually works internally under the hood. You can read straight through, or just take the parts you need.

Everything here is free to read, and will stay that way. If you'd rather read it off the browser, there's a typeset PDF + EPUB edition.