The distinction that matters is between teaching a model facts and teaching it behaviour. Fine-tuning is much better at the second. If you feed a model hundreds of examples of the tone you want, the structure you want, or the way you want it to refuse certain requests, it will start producing that pattern by default. What fine-tuning does not do well is keep a model up to date on information that changes, because every change means preparing new examples and running the training process again.
That is why most business chat tools do not rely on fine-tuning for knowledge. Retrieving the right passage from a live source at the moment of the question is faster to update, cheaper to run and easier to audit. If your opening hours change, you edit the page. With fine-tuning you would be retraining a model to remember a fact that was true last month.
Clerkzo takes the retrieval route. It trains by crawling your website, so the answers a visitor gets come from your own published content rather than from a model that memorised something during training. When an answer is not quite right, the corrections loop lets you fix it directly instead of commissioning a training run, and guardrails keep responses inside what your site actually says.
Related terms
Browse all 142 terms- TokensTokens are the small pieces of text, usually a word or part of a word, that an AI language model reads and generates. Models measure everything they process in tokens rather than characters or words.
- Context WindowA context window is the maximum amount of text, measured in tokens, that an AI model can take into account when producing a single response. It covers the instructions, the retrieved source material and the conversation so far.
- Model TemperatureModel temperature is a setting that controls how much randomness an AI model applies when choosing its next word. Lower values produce consistent, predictable output; higher values produce more varied and creative output.
- InferenceInference is the act of running a trained AI model to produce an output from a new input. It is the moment a visitor asks a question and the model generates an answer, as opposed to the earlier training phase that built the model.