The Executive AI Glossary
A plain-language glossary of the AI terms every executive should know, and why they matter in the boardroom.
AI has its own language, and most of it was invented by engineers. This glossary is built for the people making the calls, not the ones writing the code. Every term is defined the way a CEO, board member, or operating leader actually needs to hear it: what it is, why it matters, and what question to ask next.
Section 01
The foundational categories of AI models. What they are, how they differ, and where each one fits into an enterprise strategy.
The technology behind tools like ChatGPT, Claude, and Gemini. It learned from huge amounts of writing, so it can understand what you say and write back in a way that sounds human. When people say "AI" today, they usually mean an LLM.
The most advanced AI models at any given time, like Claude Opus, GPT-5, or Gemini Ultra. They cost more, and they can do more. A new frontier shows up about every six months, so any long-term AI plan has to account for that.
A newer kind of AI that thinks through a problem step by step before answering. Better for hard, complex work. Slower and more expensive for simple tasks. This is what moved AI from "fast helper" to "real analyst."
The name for AI that creates new things: text, images, code, audio, video. LLMs are the writing side of GenAI.
AI that runs on your own computers instead of a vendor's cloud. Usually slower and not quite as smart, but the right choice when your data can't leave the building, like in healthcare, defense, or finance.
AI made outside the US, especially Chinese models like DeepSeek and Qwen from Alibaba. Often cheap and surprisingly good, but they raise real questions. Where does your data go? What ideas were they trained to push? A CEO should know the names and know why the security team is careful.
Section 02
The working vocabulary of daily AI use. These are the levers that separate a mediocre result from a great one.
Everything the AI knows right now in your conversation. Your question, the documents you shared, the chat history, the instructions it was given. Context shapes the answer more than almost anything else.
How much the AI can hold in its head at one time, measured in tokens. The newest models can handle a whole book's worth. When AI "forgets" something from earlier in a long chat, it usually ran out of context window.
The skill of deciding, on purpose, what the AI gets to see. Which documents, which history, which instructions. It's more strategic than prompt engineering, and it's where the best AI results actually come from.
The skill of writing good instructions for AI. This is the hands-on part, the exact words you type to get the answer you need.
The small chunks AI uses to read and write. About three-quarters of a word. Tokens matter because pricing, speed, and how much the AI can handle are all measured in them.
When AI says something wrong but sounds sure of itself. This is the main reason people need to check AI's work before trusting it.
AI's habit of agreeing with you even when you're wrong. It's sneakier than hallucination because it feels like the AI is being helpful, when really it's just telling you what you want to hear.
Section 03
Where AI stops answering questions and starts doing work. This is the layer that creates real operating leverage.
AI that actually does things for you, not just answers questions. There's a range: a simple agent might draft an email. A stronger one researches a topic and writes a full report. The most advanced ones work across your systems with very little human help. Where your company lands on that range is a strategy choice, not a tech choice.
Rules-based work we've had for decades: if this happens, do that. Reliable, predictable, but only does what someone explicitly told it to do.
The steps that make up a business process. With AI, this usually means stringing AI and non-AI steps together, often by dragging and dropping, to get real work done from start to finish.
A setup where a person checks or approves the AI's work before it goes out. A must for anything with legal, money, or reputation risk.
Section 04
The plumbing that determines how far AI can reach inside your business. This is where strategy meets architecture.
The pipe that lets AI plug into your other tools. When AI reads your CRM, sends a message in Slack, or pulls a file from SharePoint, it's using an API. The question for a CEO isn't "what's an API?" It's "which of our systems have them?" Because that's what decides how far AI can reach inside your business.
A new standard for connecting AI to tools and data. Think of it like USB, one plug that works with lots of things. With MCP, one AI can talk to Salesforce, Google Drive, your own databases, and hundreds more. It's how AI stops being stuck by itself.
The set instructions that tell an AI how to behave for a specific job. This is how you turn a general AI into "our marketing assistant" or "our compliance reviewer" without hiring a developer. It's where AI strategy becomes AI reality.
Section 05
The terms that separate a company that uses AI from a company that gets in trouble using AI. This is the board-level vocabulary.
When AI gives unfair or skewed answers, usually because it learned from data that was already unfair. Matters anywhere AI touches hiring, loans, healthcare, or customer decisions.
The bigger question of what AI should do, not just what it can do. Covers fairness, honesty, accountability, jobs, consent, and even energy use. Different from governance (which is the rules) and bias (which is one specific problem). Shows up in boardrooms, employee trust, customer opinion, and new laws.
When employees use AI tools without telling IT or leadership. Almost every company has more of it than they realize, and it's a real security and compliance problem.
The rules, roles, and review steps that decide how AI gets built, bought, used, and watched across the company. The policy layer.
The actual controls that stop AI from doing bad things in the moment. Governance is the rulebook. Guardrails are the brakes.
How you actually test whether an AI is good at your specific job, instead of just trusting what the vendor says. More and more, this is a skill every serious company needs.