In plain English
Instead of training a model from scratch, fine-tuning starts with a model that already knows a lot and gives it more examples aimed at a particular job. This can shape how the model responds, what patterns it emphasizes, or how well it performs on a specialized task.
Example
A company could fine-tune a model on examples of its preferred support responses so the model learns to follow that style and format more consistently.