Fine-tuning takes an existing AI model and continues its training on your own examples, so it internalises a style, a format or a narrow skill. It changes the model itself — unlike RAG, which changes what the model reads. Fine-tuning requires hundreds to thousands of good examples and must be redone as base models evolve.
For most businesses it is the wrong first tool: good prompting plus retrieval covers the vast majority of cases at a fraction of the cost. It earns its place in narrow, high-volume domains — a question we settle during AI integration scoping.