Someone in a meeting says “we should use RAG with an open-weight model and connect it over MCP,” and everyone nods.
You don’t have to nod anymore. Here are the AI terms explained in plain English, each with why it matters to you as an owner or operator. Keep this page open next to the how to use AI guide and you’ll never be lost in an AI conversation again.
The basics
Artificial intelligence (AI)
Software that learns patterns from examples instead of following only hand-written rules. Why it matters: it handles fuzzy, language-heavy work that old software couldn’t touch, like drafting, summarizing and sorting messages.
Generative AI
AI that creates new content, such as text, images, audio, video or code. Why it matters: this is the kind of AI most business owners actually use day to day.
Large language model (LLM)
The engine behind chat assistants, trained on massive amounts of text to predict what comes next. Why it matters: knowing it predicts rather than “knows” explains both its strengths and its mistakes.
Model
A specific trained AI system, often released in versions or families. Why it matters: different models suit different jobs, so test a few on your real work instead of assuming the newest is always best for you.
Chatbot or AI assistant
The app you type or talk to, built on top of a model. Why it matters: the assistant adds features like file upload, memory and tool connections that the raw model doesn’t have.
Multimodal
A model that works with more than one kind of input or output, such as text, images, audio and video. Why it matters: you can show it a photo of a whiteboard, a receipt or a chart instead of typing it all out.
Talking to AI
Prompt
The instruction you give an AI. Why it matters: prompt quality is the biggest factor in output quality; my prompt templates show the structure.
System prompt or custom instructions
Standing instructions that apply to every message, like who you are and how you want answers. Why it matters: set them once and every conversation starts smarter.
Token
A small chunk of text, roughly a piece of a word, that models read and write. Why it matters: limits and usage are counted in tokens, which affects how much material you can feed in at once.
Context window
How much text a model can consider at one time, including your instructions, documents and the conversation so far. Why it matters: in very long chats, early details can fall out of view, so restate what’s important or start a fresh, well-briefed chat.
Hallucination
When AI produces confident information that’s false, like an invented fact, source or figure. Why it matters: it’s the main reason a human verifies anything that matters before it’s used.
Temperature
A setting that controls how predictable or varied a model’s output is. Why it matters: lower is better for consistent, factual tasks; higher is better for brainstorming. Most assistants handle this for you.
Making AI know your business
Training
The process of teaching a model patterns from large amounts of data before you ever use it. Why it matters: a model’s built-in knowledge stops at its training data, so it won’t know recent events or your business unless you supply them.
Fine-tuning
Further training a model on a narrower set of examples to specialize it. Why it matters: most small businesses don’t need it; good instructions and your own documents usually get you most of the way.
RAG (retrieval-augmented generation)
The AI looks up relevant passages from a source you control, like your documents, and uses them to answer. Why it matters: it grounds answers in your real policies and processes instead of the model’s general knowledge.
Embeddings
A way of turning text into numbers that capture meaning, so a system can find similar content. Why it matters: it’s the plumbing behind “search my documents by meaning,” which powers most RAG setups.
Knowledge base or second brain
An organized store of your processes, decisions and reference material that AI can draw from. Why it matters: my own agents run on one, and it’s what makes them work the way my business works.
When AI does the work
Workflow automation
Connecting apps so a trigger starts a set of steps automatically, like a new form creating a task and sending an email. Why it matters: many “AI” wins are really good automation with one AI step in the middle. More in AI workflow automation.
AI agent
AI that pursues a goal by planning steps, using tools and checking its own work, handing off to a person when needed. Why it matters: this is the jump from AI that answers to AI that does. Start at AI agents explained.
Agentic AI
A description of AI systems that act with some independence across multiple steps. Why it matters: “agentic” is a spectrum, and knowing where a tool sits on it tells you how much oversight it needs.
Orchestrator
A coordinating agent that breaks work into pieces and hands them to specialist agents. Why it matters: it’s how one AI chief of staff can manage several specialists, the way mine does.
Human in the loop
A design where a person reviews or approves AI output at key points. Why it matters: it’s how you get speed without giving up control, especially for anything customer-facing.
Tool use or function calling
An AI’s ability to take actions through other software, like searching, updating a spreadsheet or sending a draft. Why it matters: this is what lets agents act, and why permissions matter.
MCP (Model Context Protocol)
An open standard for connecting AI assistants and agents to tools and data, created by Anthropic and, as of October 2026, stewarded by the Agentic AI Foundation under the Linux Foundation. Why it matters: it’s becoming a common plug for AI, so connecting your tools gets simpler. See what is MCP.
API
A way for software to talk to other software. Why it matters: APIs are how AI gets wired into your business systems, and they’re usually what a developer means by “integration.”
The landscape
Open-weight model
A model whose trained settings are published so others can download and run it. Why it matters: it can offer more control over where data lives, with more setup responsibility on you.
Closed model
A model you access only through the maker’s app or service. Why it matters: usually easier to start with; you rely on the provider’s data and privacy settings.
Vibe coding
Building software by describing what you want to an AI and iterating on what it produces. Why it matters: non-coders can now build working tools, though security and data still need expert review. See build an app with AI.
Your next step
Bookmark this page. Next time you hear a term you don’t know in a meeting or video, look it up here, then try it on real work.
If you’re just starting, read AI for beginners next. And if you want to see these ideas in a working business, I show my agents and builds on YouTube, or you can read about me and why I run my companies AI-first.
Questions people ask
What is an LLM?
An LLM, or large language model, is the AI behind chat assistants. It’s trained on huge amounts of text to predict what comes next, which lets it write, summarize, answer questions and follow instructions.
What are tokens in AI?
Tokens are the small chunks of text an AI model reads and writes, roughly pieces of words. Models measure their limits and usage in tokens, which is why very long documents can get cut off or summarized.
What is RAG in AI?
RAG stands for retrieval-augmented generation. The AI looks up relevant information from a source you control, like your documents, and uses it to answer, which makes answers more grounded in your actual material.