Fine-tuning
Fine-tuning is further training a pre-trained model on a specific dataset so it specializes in a particular task, tone, or domain.
Analogy
Like sending a skilled generalist employee to a focused training course so they nail your company's specific way of doing things.
Why it matters
It is an alternative to prompting when you need consistent specialized behavior, though most automation needs are met more cheaply with good prompts and grounding.
In practice
Fine-tuning a model on past support tickets so it replies in your brand's voice.