Fine-tuning
Definition
Fine-tuning is extra training that adjusts an existing AI model on your own examples so it learns a specific style, format or task. It changes the model itself, while a prompt or a document search only changes what the model sees for one request.
By Ronan Liedmeier, fractional head of growth and course author · Published 17 October 2026 · Last updated 17 October 2026
Also called: fine tuning, model fine-tuning
Most teams should try cheaper fixes first. A better prompt with a few examples, or a retrieval step, will fix most problems for less money and effort. Fine-tuning pays off when you need the same narrow behavior thousands of times and prompting alone keeps slipping.
How it differs from RAG
Retrieval-augmented generation hands the model relevant documents at the moment it answers. Fine-tuning bakes patterns into the model ahead of time, through a training run on pairs of example inputs and approved outputs.
So the two suit different jobs. Retrieval fits facts that change or need a citation, such as price lists and policies. Fine-tuning fits behavior: a fixed output format, a house tone, or a classification task with your own labels.
Example
An ecommerce operator has a few thousand past support tickets, each tagged by a person with one of the company's 12 contact reasons. A general model with a good prompt tags most of them right but keeps inventing new categories. Fine-tuning a smaller model on the tagged tickets teaches it the exact list. The team scores it on a held-out batch of tickets before switching over.
Common mistake
Fine-tuning a model on the company FAQ or product catalog so it "knows" the business. The catalog changes next quarter, the model keeps answering from the old version, and there is no source to check the answer against. Put that knowledge in the prompt or behind a retrieval step.
Learn it in AI Academy
- RAG: When It's the Right AnswerExpert · 9 min
- The Adoption PlaybookExpert · 9 min
Related terms
- Retrieval-augmented generation (RAG)
- Retrieval-augmented generation (RAG) is a way of getting an AI model to answer from your own documents. A search step first finds the passages that match the question, then the model writes its answer using those passages.
- Few-shot prompting
- Few-shot prompting is giving an AI model a few examples of the input and output you want, right inside the prompt, so it follows the pattern. The model picks up the format and style from the examples on the spot, with no retraining.
- Prompt engineering
- Prompt engineering is the practice of writing and testing the instructions you give an AI model so it produces reliable, useful output. For everyday users it means writing clear briefs; for teams running AI in production it means treating prompts like versioned, tested code.
- AI eval (evaluation)
- An AI eval (evaluation) is a repeatable test that checks whether an AI prompt or workflow gives good answers, by running it on a fixed set of inputs and scoring the outputs against agreed criteria. It turns "seems fine" into a score you can compare.