In plain language
An LLM predicts useful sequences of tokens, enabling tasks such as drafting, extraction, classification, and question answering.
Example
A workflow asks an LLM to extract a customer’s requested product and urgency from an email into defined JSON fields.
Related terms
artificial intelligence · Prompt · Hallucination
Common misconception
An LLM is not a database or a guaranteed source of current facts, even when it produces detailed and authoritative-sounding prose.
Implementation guidance
Use the smallest suitable model, constrain the task and output schema, supply trusted context, evaluate representative cases, and plan for refusals and invalid output.