IT term explained

Large language model (LLM)

A large language model (LLM) is a type of artificial-intelligence system that processes and generates language. It learns patterns from a very large amount of text and uses those patterns to produce a response based on instructions and information supplied by a user or application.

People can interact with an LLM through a chat application. Software can also send instructions and data to a model through an API and use the response as part of a larger business process.

For example, an application might ask an LLM to summarize project notes, draft a response to a customer, classify support requests, or extract information from a document.

What should a business know about LLMs?

LLMs can be useful when a task involves language and does not have one simple, fixed answer. They can help people work with large amounts of unstructured information without requiring every rule to be programmed in advance.

Their responses are not guaranteed to be correct. An LLM can misunderstand instructions, omit important details, or confidently generate inaccurate information. Its output should be reviewed in proportion to the business risk of the task.

Businesses should also understand what information is being sent to the model provider, how that information may be stored or used, and whether the service is appropriate for confidential, personal, or regulated data.

An LLM is usually most valuable as one component of a well-defined process, with software and people responsible for validating its output and deciding what happens next.

Mentioned in

Choosing (and then re-choosing) a database for my work app