AI glossary
Plain English definitions for the terms you keep hearing.
LLM (Large Language Model)
The text “engine” behind tools like ChatGPT/Claude. It predicts the next word based on patterns from training data.
Prompt
Instructions + context you give the model. Good prompts include constraints, examples, and the desired format.
RAG (Retrieval-Augmented Generation)
A method where the system fetches relevant internal docs before generating an answer. Helps keep outputs grounded in your own info.
Hallucination
When a model produces plausible-sounding but incorrect content. Guardrails, validation, and source retrieval reduce this risk.
Agent
A workflow where the model can take actions (send a message, update a CRM, schedule). Safe agents have strict permissions and logs.
Guardrails
Controls that make AI safe and useful: approvals, templates, validation, limited actions, and auditing.
Fine-tuning
Training a model on your examples. Often unnecessary; many businesses get better results with good prompts + RAG + workflows.
Automation
Rules-based steps (if/then). AI is helpful when the input is messy (a conversation) but the output should be consistent.