Three ways to shape an AI model’s behaviour

When you want an AI model to behave in a specific way for your product, you have three broad options. Prompting. Retrieval-augmented generation, usually shortened to RAG. Fine-tuning.

These three sound like alternatives, but they’re really different tools for different jobs. Each has strengths, weaknesses, and costs. Knowing which one to reach for — and when — can save weeks of work and significant money.

Let’s walk through them in plain language.

Prompting: instruct the model directly

Prompting is the simplest option. You write instructions in the prompt — what the model should do, what tone to use, what format to return — and the model follows them as best it can.

Modern models are good at following prompt instructions. With a thoughtful prompt...

Prompting vs RAG vs fine-tuning (when to use what)