AI glossary: 15 terms in plain English

Agent, prompt, token, RAG, fine-tuning and more, each explained in a sentence or two.

The terms

• Prompt: the instructions you give an AI tool.
• Token: a chunk of text (roughly ¾ of a word) models read and write.
• Context window: how much text a model can consider at once.
• LLM: large language model, the engine behind chat assistants.
• Agent: an AI system that plans and takes actions with tools.
• RAG: retrieval-augmented generation; the model looks up your documents before answering.
• Fine-tuning: extra training on your own examples.
• Hallucination: a confident answer that is wrong.
• Multimodal: works with text, images, audio, or video.
• Embedding: a numeric fingerprint of text used for search.
• Automation: rules that move work between apps without you.
• Guardrails: limits on what an AI system may do.
• Human in the loop: a person approves key steps.
• Open-weight model: a model you can download and run yourself.
• Inference: the moment a model produces an answer.

Want help applying this? See how we work or browse AI tools.