Definition
Large language models, or LLMs, generate text by predicting likely sequences of words or tokens. They can summarize, draft, classify, translate, reason over context, and interact with tools when connected to external systems.
How it works
LLMs transform text into tokens, process those tokens through neural network layers, and generate the next likely token based on patterns learned during training and context supplied at runtime.
Why it matters at work
LLMs power many workplace AI tools, including chat assistants, copilots, summarizers, research tools, and agentic systems. Understanding their strengths and limits is the foundation of safe use.
Workplace example
A legal team uses an LLM to summarize a contract, but keeps a lawyer responsible for checking clause references, missing obligations, and jurisdiction-specific nuance.
Frequently Asked Questions
Can large language models know whether they are correct?
Not reliably. LLMs can sound confident while being wrong, so important outputs need verification against sources, calculations, policies, or expert review.
Are LLMs databases?
No. LLMs encode patterns from training data, but they are not reliable databases. For current or factual work, connect them to trusted sources and verify the answer.