What is a large language model?
A large language model, or LLM, is an AI system trained on very large amounts of text to predict the next word in a sequence, which lets it write, summarize, translate, classify and answer questions in ordinary language.
The LLM is the engine; ChatGPT, Claude, Gemini and Copilot are the cars built around it. Under the hood, the model has learned statistical patterns of how words follow each other across a huge body of text. That sounds modest, but at scale it produces something that can follow instructions, reason through a problem step by step and adapt its tone to an audience.
An LLM does not look things up unless it is connected to a source. What it knows is frozen at its training date, and it fills gaps with plausible-sounding text. That is why the useful business setups feed it your own documents, records or call transcripts and ask it to work only from those, rather than trusting what it remembers.
For buying decisions, the model matters less than how it is deployed. Ask where your data goes, whether it is used for training, how long it is kept, and whether the vendor offers an agreement that covers your obligations, such as a HIPAA business associate agreement for a medical practice. Those answers differ more between plans than between models.
Example
For example, a CPA firm could use an LLM to read a client's year-end email thread and produce a short list of open items: missing 1099s, an unanswered question about a vehicle purchase, a pending K-1. A staff accountant checks the list against the client file in Karbon before sending a document request to the client.