Glossary

Every term a beginner might not know, defined in plain words. Terms are listed alphabetically. Pages across this site link here the first time they use a term.

Agent

A single AI worker with a specific job. In an AI team, each agent is described by a text file that says who they are and what they do. Examples: an HR agent, a research agent, a writing agent.

Agentic AI

An AI that can do things inside files on your computer — create folders, write and edit files, follow procedures — not just answer questions in a chat. This whole approach only works with agentic AI. Examples: Claude Cowork and ChatGPT Codex. A regular chatbot can describe a team but can't actually build and run one in your folder.

Append

To add to the end of a file without erasing what's already there. For example, an end-of-day journal appends each new entry below the last, so nothing is lost.

BKM

Short for Business Knowledge Management — your team's playbook of proven ways to do things. In the setup, the BKM folder holds important procedures, like the rule for how hiring is done.

Build prompt

The single block of instructions you paste into your AI tool to create your whole team. You'll find it, ready to copy, on the Build Your Team page.

Claude Cowork

One of the agentic AI tools you can use for this setup. It can work inside a folder on your computer.

ChatGPT Codex

Another agentic AI tool you can use for this setup. Like Claude Cowork, it can work inside a folder on your computer.

Context file

A file you add to a project so the AI can see it every time — not just for one chat. You add one in your project's Context, Knowledge, or Files section.

HR

The team member responsible for "hiring" new AI specialists. HR uses what the Research member finds out to decide what a good new team member should look like — and asks for your approval before adding anyone.

Inbox

A designated place for passing information. Your team uses two:

• Team Inbox — where you drop information for the team.
• Owners Inbox — where the team puts finished work for you to review.

Keeping them separate prevents incoming and outgoing work from getting mixed up.

Journal

A running log where the team writes a short end-of-day summary of what you did. When you start a new session, the team reads the journal to get its bearings and catch you up on anything unfinished. This is how your team "remembers" across days.

Orchestrator

The manager of your AI team. Its only job is to route each task to the right specialist — it never does the work itself. This is the core idea of an AI team: one part decides who should handle a task, and specialists do the actual work. You can rename it anything you like (for example, Dash).

Persona

An optional name, personality, and identity given to a team member so it's easier to address them directly. Adding personas is the optional part of the build prompt.

Prompt

The instructions you type to an AI. A short prompt might be a single question; the build prompt on this site is a longer set of instructions that creates your whole team.

Research

The team's senior researcher. Research finds out what skills a real human expert in a field has, so HR knows what a strong new hire should look like. The setup also has Research define what elite work looks like, so the team only aims to hire at the very best level.

SOP

Short for Standard Operating Procedure — a short written rule that tells your team how to do something the same way every time (for example, "always get the owner's approval before hiring"). Your team stores SOPs as files so the rules are visible and repeatable.

SOP index

A simple list of all your team's SOPs, kept up to date so everyone can find the right procedure.

Team index

A simple list of all your team members and their roles, kept up to date as the team grows.

Tokens

The small chunks of text an AI reads and generates — roughly a word or piece of a word at a time. Think of tokens as the AI's "fuel": every request uses some, and fewer tokens means lower cost and less effort to run. Because an AI team is stored as plain text, it runs on relatively few tokens.

Don't see a term you ran into? Email us at wwo@byui.edu and we'll add it.