An LLM is a model trained on large text corpora to generate and transform language — answers, summaries, code, and dialogue. It is useful as a flexible interface over knowledge and tasks when paired with clear inputs and, where needed, retrieved sources.
Whilst LLMs sound like they “know” your business, in practice they only know training data plus what you put in the prompt. For example, asking for last quarter’s policy without retrieval invents a plausible answer. We often recommend treating the LLM as the reasoning layer — and grounding it with retrieval or tools when facts must be current.