Training for a specific job
Fine-tuning gives an already trained model more examples for a specific job. A company might show it support messages labelled by topic so it can sort new messages. Unlike a prompt, which guides a response, fine-tuning changes the model itself.
When to use it
Fine-tuning can help with repeatable tasks or a consistent style. For facts that change often, such as prices or policies, an app can look them up when answering.
Examples must be accurate and used with permission. Teams should check that fine-tuning improves the intended task without making the model worse at others.