The No-BS AI Glossary

An agent fleet is a system where multiple AI agents run in parallel with distinct roles, coordinate with each other, and have their combined output verified — ideally by something automated rather than a human reading every piece.

Agent-washing is the practice of rebranding existing software, chatbots, and plain automation as "agentic AI" because agents are what buyers want to hear.

AI, in working terms, is software that produces probabilistic output — answers that shift with the input, the model, and a little built-in randomness, in ways you can't fully predict.

An agent is an AI you hand a goal to that then decides the steps and takes the actions itself — sends the email, updates the record, moves the money — without stopping to ask permission at each one.

AI-washing is sticking the "AI" label on conventional, deterministic software because AI is what buyers and investors want to hear.

Automation is software that runs without you — triggered by a schedule or an event, like a file landing in a folder — and does exactly the same thing every time.

The chat is the conversation surface of AI — Claude.ai, ChatGPT, Gemini in a browser tab.

Chunking is breaking AI work into pieces small enough that each piece's output can be verified in under five minutes — the five-minute rule.

A coding agent — Claude Code, Cursor, and their kin — is an AI workplace that lives in a folder of files on your computer: it reads and writes those files, can run commands, keeps its state between sessions, and produces durable artifacts instead of text trapped in a conversation.

The context window is a model's working memory — the maximum amount of text, measured in tokens, it can consider at once: your instructions, the conversation so far, whatever documents you've handed it.

An enterprise suite — Microsoft 365 Copilot, Google Workspace with Gemini, Salesforce's AI layer — isn't really one workplace; it's a wrapper containing several.

A hallucination is AI output that's wrong and delivered with exactly the confidence of output that's right — an invented citation, a policy that doesn't exist, a total that was never in the document.

Human in the loop describes a workflow deliberately split between AI and a person — some steps handled by the machine, some by human judgment.

Human-washing is marketing an AI system as having "human oversight" when the human's actual role is decorative — providing the appearance of oversight with none of the substance.

An in-context assistant is the AI button inside a tool you're already using — Copilot in Word, Gemini in Docs, the AI in your CRM.

MCP — Model Context Protocol — is a shared standard for connecting AI systems to your other software: the calendar, the inbox, the CRM.

Production-ready means an AI tool is reliable, predictable, and safe enough to depend on for real work — not just impressive in a one-off demo.

A system prompt is the standing instruction a model receives before your message ever arrives — who it should be, what rules it follows, what it's for.

The 100-times rule is the fast tiering for whether recurring work deserves automation: 100+ times, build the workflow; 10–100 times, write a checklist or a template; 1–10 times, just do it by hand.

The four questions are what to ask before opening any AI tool, because most "AI didn't work" verdicts are really wrong-workplace verdicts.

The six questions are the decision discipline to run — in order — before committing real money or real weeks to any AI build.

The two-strike rule governs what to do when an AI task fails: you get one rerun with a sharper prompt, and if it fails again, you stop rerunning and refactor.

"Then what" is the two-word question that stops a doomed AI project cold: the AI wrote the thing — then what?

A token is the unit AI models actually read and write — a chunk of a word, roughly three-quarters of an English word on average.

In AI, a tool is something you pick up, use, and put back down — a spell-checker, a calculator, ChatGPT opened to write one email.

Tool/Toy/Trash is a three-second triage for any AI product, workflow, or subscription.

A workflow is something you do over and over that produces a result someone actually needs.

A workflow platform — n8n, Zapier, Make, Power Automate — is where you build automations that run on triggers: a schedule fires or an email lands, a chain of connected steps executes, and the same work happens the same way every time.