Fine-tuning takes a pre-trained language model and continues training it on your specific data. The model already understands language from billions of tokens. Your job is to specialize it.
Think of it like hiring an experienced writer and teaching them your company's style guide. They already know how to write. You're just adapting their skills to your needs. This is far cheaper than training a writer from scratch.