Glossary
A large language model (LLM) is a machine-learning model trained on large amounts of language data to recognise patterns and generate or process text. An LLM can perform tasks such as summarisation, classification and drafting. Its output is not automatically a verified database fact, even when the answer is detailed and confident.
An LLM may be combined with retrieval, tools, validation and human review. Those surrounding components determine whether it can access current supplier documents, retrieve catalogue records or write proposed changes. The model and the complete application are therefore different things.
For example, a product-data application can give an LLM a manufacturer datasheet and ask for a structured set of candidate attributes. The workflow should retain the document, identify the relevant product and test the returned fields before they become approved catalogue values.
A model can explain a technical term without knowing the correct value for a particular model or variant. Training data may be incomplete, outdated or unrelated to the current record. Even a well-written answer can therefore attach a real specification to the wrong product.
Evaluate the specific task with representative products and known outcomes. Check extraction, product matching and formatting separately, rather than relying on a model's general reputation. Use traceable evidence and explicit uncertainty handling when missing or conflicting facts would otherwise be turned into plausible text.
Discuss the product information and controls your AI use case needs.