20+ AI & SaaS terms defined plainly.
Agent
An autonomous program that can take actions to accomplish goals, often using LLMs to plan and reason.
RAG (Retrieval-Augmented Generation)
A pattern where the LLM retrieves relevant information from a database before answering.
Embedding
A vector representation of text, used for semantic search and retrieval.
Vector Database
A database designed to store and search embedding vectors (e.g. Pinecone, Weaviate, pgvector).
LLM (Large Language Model)
A neural network trained on text that generates human-like language (e.g. GPT-4, Claude, Gemini).
Prompt Engineering
The craft of writing instructions that get the best output from an LLM.
Fine-tuning
Further training a base LLM on your specific data to improve its performance on your tasks.
Tokens
The unit of text LLMs process — roughly 4 characters or ¾ of a word per token.
Context Window
The amount of text an LLM can read at once (e.g. 200K tokens for Claude).
Hallucination
When an LLM generates plausible-sounding but factually incorrect content.
Eval (Evaluation)
Automated tests that score AI output quality on a benchmark dataset.
Guardrails
Rules and filters that prevent AI from producing unsafe or off-brand output.
AEO (Answer Engine Optimization)
SEO for AI assistants — optimising content to be cited by ChatGPT/Gemini/Perplexity.
GEO (Generative Engine Optimization)
Same as AEO — newer term gaining traction.
Multi-modal
AI that handles multiple input types: text, images, audio, video.
Function calling
When an LLM is allowed to invoke external tools/APIs as part of its response.
Workflow automation
Chaining tools and actions (n8n, Zapier, Make) to run business processes automatically.
Agentic workflow
An automation where an LLM decides which steps to take, dynamically.
Knowledge base
The collection of documents your AI is trained or retrieves from.
API (Application Programming Interface)
How two software systems talk to each other programmatically.