Short answer
Fine-tuning is further training of an existing language model on a set of example inputs and outputs, so it adopts a specific style, format or classification behaviour by default. Train one on 500 of your own past support replies and it writes in your house tone. Fine-tuning teaches patterns, not facts.
Also called: Model fine-tuning, Supervised fine-tuning
The distinction that saves money is between behaviour and knowledge. Fine-tuning shifts behaviour: output that always parses as the same structure, a fixed reply format, reliable sorting of messages into eight categories, a voice matching the rest of your writing. It is a poor way to install facts, because a fact learned this way cannot be edited without training again and cannot be cited when the model uses it.
The prerequisite is data that most small businesses have not collected yet: several hundred clean, consistent examples of the input and the ideal output, prepared by someone who knows what good looks like. Assembling that set is usually the largest part of the work. Prompting and retrieval are cheaper first attempts, and both are worth exhausting before paying to train, since a well-written prompt closes most of the gap for style alone.
Common questions
Do we need to fine-tune a model to use AI in our business?
Rarely, and almost never first. Clear prompting handles tone and format for most cases, and retrieval handles company facts. Fine-tuning becomes worthwhile when a task repeats at volume and needs output identical in shape every time, or when a smaller, cheaper model has to match a larger one on a narrow job.
Related terms
Large Language Model (LLM)
A large language model is a neural network trained on large volumes of text to predict likely continuations, which lets it write, summarise, classify and translate without being programmed for each task.
Retrieval-Augmented Generation (RAG)
Retrieval-augmented generation is a technique that searches a private collection of documents for passages relevant to a question, then passes those passages to a language model as context for its answer.
Prompt Engineering
Prompt engineering is the practice of writing model instructions that produce the required output reliably: stating the role, the task, the constraints, the output format, and one or two worked examples.
Inference Cost
Inference cost is what a business pays each time a model produces an output, billed by most providers per token of input and output, where a token is roughly three-quarters of a word.
Reading definitions because you are scoping a project? Skip ahead and just ask.