AI Glossary
Clear, useful definitions of the AI concepts that matter for business decisions. Not a dictionary of every AI term — just the ones that affect your choices.
Core Concept
RAG (Retrieval-Augmented Generation)
An approach where an AI system retrieves relevant information from a knowledge source before generating an answer. This grounds the response in approved content, reduces hallucination risk, and provides citations to sources.
Embeddings
Numerical representations of text that capture semantic meaning. Similar content has similar embeddings, enabling semantic search and retrieval. Embeddings are fundamental to RAG and intelligent search systems.
LLM (Large Language Model)
A large AI model trained on vast amounts of text that can generate human-like responses, answer questions, summarize content, and perform various language tasks. LLMs are the reasoning engine behind most AI applications.
AI Agent
An AI system that can reason about a task, use tools, call APIs, and perform actions to achieve a goal. Unlike a chatbot that only responds, an agent takes actions within defined permission boundaries.
Context Window
The amount of text an LLM can process in a single request. Context window sizes vary by model and affect how much information can be included in a prompt. Managing context windows is essential for RAG and conversation systems.
Tool Calling
The ability of an LLM to invoke external tools, APIs, or functions to perform actions or retrieve information. Tool calling enables AI agents to interact with external systems, but requires permission boundaries and validation.
Semantic Search
Search that understands the meaning of a query rather than just matching keywords. It uses embeddings to find content that is semantically similar, even if the exact words do not match.
Hybrid Search
Search that combines semantic (meaning-based) and keyword (exact match) retrieval. Hybrid search leverages the strengths of both approaches — semantic understanding and precise keyword matching.
Hallucination
When an AI system produces information that is incorrect or fabricated but presented as factual. Hallucinations cannot be fully eliminated, but RAG, citations, validation, and human review significantly reduce the risk.
AI Evaluation
The systematic testing of AI system outputs for quality dimensions including correctness, relevance, groundedness, factuality, and task completion. Evaluation is essential before deploying AI to production.