An LLM (large language model) is an AI model trained on vast amounts of text that processes and generates natural language: it answers questions, summarises documents, writes and analyses text. ChatGPT, Claude and Gemini all run on LLMs.
An LLM is built on statistics: the model predicts the most probable continuation of a text based on patterns in its training data. It doesn't "know" anything in the human sense and doesn't check facts – which is why it can be wrong while sounding confident (see: AI hallucinations).
In business use an LLM is an engine, not a finished product. Value appears once it's connected to a task: summarising customer tickets, searching documentation, drafting product descriptions. Whether it can answer questions about your company depends on access to your data (the RAG technique), not on the choice of model alone.
Why it matters for business: a conversation about "implementing AI in the company" is, in practice, a conversation about an LLM connected to company data and processes, with quality control along the way. The model is a tool – the result depends on who steers it and how.