Jargon Buster / Fine-Tuning
Fine-Tuning.
AI & AutomationFurther training an existing AI model on your own examples so it behaves in a specific, consistent way.
What is fine-tuning?
Fine-tuning is the process of further training an existing AI model on a set of your own examples so that it consistently behaves in a particular way, such as following a format, matching a tone or sorting items into your categories.
You do not build a model from scratch. You take one that already understands language and show it many examples of the input it will see and the output you want back.
Why it matters
Fine-tuning is a spending decision, and often an unnecessary one. It takes preparation, testing and ongoing upkeep, and it ties you to a specific model version that providers eventually retire. For a high-volume, repetitive task it can cut cost and improve consistency. For most SME uses, careful prompting or connecting the model to your documents gets the same result faster, cheaper and with far less to maintain.
How it works
- Choose a base model Start from a model the provider allows you to fine-tune, usually through their developer platform.
- Prepare examples Collect pairs of input and ideal output. Quality and consistency matter far more than volume.
- Run the training The provider adjusts the model on your examples and returns a private version for your account.
- Test against the original Compare the tuned model with a well-prompted standard model on cases it has not seen.
- Maintain it When the base model is updated or retired, you repeat the process.
The honest view
Most businesses asking for fine-tuning want the model to 'know our business': prices, policies, product details. Fine-tuning is poor at that. It shapes behaviour and style; it is an unreliable way to teach facts, and those facts go stale the moment they change. Retrieval-augmented generation suits knowledge far better.
Fine-tuning earns its keep in narrower cases: classifying thousands of documents the same way, producing a strict output format, or running a smaller, cheaper model at volume. As general models improve, that list keeps shrinking.
What to do about it
Before considering fine-tuning, write a thorough prompt with two or three examples of perfect output and test it on real cases. If the problem is missing knowledge, connect the model to your documents instead. Only look at fine-tuning when you have a stable, high-volume task, a large set of consistent examples and a clear way to measure whether the tuned version is better.
Related terms.
Prompt Engineering
The practice of writing clear, structured instructions so an AI model gives accurate, useful output every time.
AI & AutomationSee more RRetrieval-Augmented Generation (RAG)
A method where AI first looks up your trusted documents, then writes its answer from what it finds.
AI & AutomationSee more LLarge Language Model (LLM)
An AI system trained on huge amounts of text that can write, summarise, answer questions and follow instructions.
AI & AutomationSee more MMachine Learning
A branch of AI where software learns patterns from data to make predictions, instead of following hand-written rules.
AI & AutomationSee moreGot a question about your own setup? Let's talk.
Your first call is with our founder, Rhodri Williams.