Fine-tuning continues training a pretrained model on a narrower dataset so it adopts a style, domain, or task. It is more invasive than prompting alone, cheaper than training from scratch, and a common way enterprises specialize foundation models without owning pretraining.
Fine-tuning starts from a general model and nudges its weights toward your examples — support tickets, code style, medical phrasing.
Done poorly, it memorizes noise or erases useful generality; done well, it reduces prompting gymnastics for repetitive workflows.
Latent stories on enterprise AI often hinge on whether vendors are selling prompts, retrieval, or true fine-tunes with governance around training data.
Fine-tune for style, format, or stable skill; use retrieval (RAG) when facts change often and must stay citeable from source documents.
LoRA is a popular parameter-efficient fine-tuning method that updates small adapter matrices instead of every weight, cutting cost and storage.