agent
agents · AI agent · AI agents · Personalized AI Agent · Personalized AI Agents
In this bootcamp, an AI tool you guide with instructions and background about your work, starting with your Agent Profile.
Glossary
Short definitions for the words these guides use. Every term here can also be opened right where it appears in a guide.
agents · AI agent · AI agents · Personalized AI Agent · Personalized AI Agents
In this bootcamp, an AI tool you guide with instructions and background about your work, starting with your Agent Profile.
profile · profiles · Agent Profiles · Personalized AI Profile · Personalized AI Profiles
The portable document used in this bootcamp to describe your role, responsibilities and preferences. You give it to your AI tool so it has useful background for your requests.
Information an AI tool saves to help with later conversations. What it saves and how you manage it depend on the tool and your account settings.
AI tools
An app you use to ask AI for help. This library covers ChatGPT, Claude, Microsoft 365 Copilot Chat and Gemini for Google Workspace.
class-safe example · class safe
A classroom example designed to show a situation without identifying or restricted details.
code blocks
A separate box used to display text or code while preserving spacing. AI tools often provide a Copy button for the box.
Communication Style Guides · My Communication Style Guide
The file you build in Week 2 that describes how you usually write, so your agent can draft in your voice. Any examples in it are ones you're allowed to use. Later guides add sections to it, such as your Voice Dial.
connectors · connected app · connected apps · MCP server · MCP servers
A link that lets your AI tool reach another app, such as your email or files, and sometimes act in it. Connect one only if your organization allows it. Rules can't fence in a connector. Once it's connected, assume your agent can reach everything the connector can. The real limits are your tool's controls: block the actions you don't want (send, change, delete), turn the connector off in chats that don't need it where your tool lets you, and disconnect it when you're done.
The information available to an AI model while it answers, including your request and relevant conversation, files or retrieved sources. It helps the answer fit the task.
context windows · context length · context limit
The limit on the information a model can work with at one time. A longer conversation may leave less room for new material or require earlier content to be summarized.
custom agents · custom GPT · custom GPTs
A saved AI setup that reuses instructions and may include reference material. Its features and controls depend on the tool.
dictate · dictating
A feature that turns your spoken words into text you can review and edit. It lets you speak a message instead of typing it.
setup check
A section at the end of the Agent Profile template that asks your agent to summarize what it read and name anything unclear. Run it in a chat where your agent has your profile, then compare the answer with your file, so you catch a misreading before you rely on the agent.
An answer connected to identifiable sources, such as files or web pages. Check that those sources actually support the answer.
hallucinations · hallucinate · hallucinating
An AI-generated statement that is false or unsupported but presented as if it were reliable.
handoffs · handoff prompt
A summary you carry into a new chat so it can continue the work. It includes the goal, key decisions, relevant facts and next step.
HTML files · HTML page
A file a browser displays as a web page. It can contain interactive features, and whether it works offline depends on what the file needs.
large language models · LLM · LLMs · language model · language models
A type of AI model trained on large amounts of language data to generate text. Often called an LLM, it can help AI tools write, summarize and answer questions.
master copies · master file · master files
The file you keep yourself, outside your AI tool, as the current version of your Agent Profile, style guide or rules. Make lasting changes there, then replace the saved copy in your agent, or attach the new file to each chat. A correction in a chat doesn't change the file.
models · AI model · AI models
The underlying AI system that processes information and generates a response. An AI app may use more than one model.
personal accounts
An account or workspace you use as an individual, separate from your organization's managed work setup.
placeholders
Stand-in text, such as [client name]. Templates use placeholders to mark what you fill in, and you use them to keep a real detail out of an AI tool. In that case your agent should keep each one as written, without guessing what it replaced, and you put the real detail back yourself.
plain-text · text file · text files
Text stored as characters, without formatting such as bold type, fonts or page layout. A .txt file is a common plain-text format.
Notebook instructions · Project or Notebook instructions
The directions saved in your agent's Project or Notebook. Your Agent Profile describes your work, and the instructions tell your agent how to use it and your other files. Every chat you start there begins with them.
Projects or Notebooks · notebook · notebooks · ChatGPT Project · ChatGPT Projects · Claude Project · Claude Projects · Copilot Notebook · Gemini Notebook · My Agent Project · My Agent Notebook
Your agent's saved home in your AI tool: a Project in ChatGPT or Claude, or a Notebook in Copilot Chat or Gemini. It keeps your files and instructions together, so every chat you start inside it begins with them.
prompts
A message that asks an AI tool to do something. It can include instructions, background and examples.
Able to open and read something without changing, sending or deleting it. Asking your agent to stay read-only doesn't make a connector read-only. Only your tool's controls can block a connector's other actions, and some tools have none.
reasoning setting · thinking setting · thinking mode · extended thinking
A setting that lets an AI model spend more time working through a request before replying. It may help with complex tasks and can make replies take longer.
redacts · redacted · redacting · redaction · scrub · scrubs · scrubbed · scrubbing
To remove or replace identifying or restricted details in a document or example, often with placeholders, before you give it to an AI tool. Also called scrubbing. Redaction reduces exposure. It does not create permission.
Safety and Security Rules · rules file · rules files · My Safety & Security Rules
The file you build in Week 3 that sets the boundaries, checks and human review your agent works within when it handles your information. It works alongside the data rules and your organization's policies, and never replaces them.
tokens
A piece of text a model processes, such as a word, part of a word or punctuation. Many model limits are measured in tokens.
trained on
The examples an AI model learns from during training. They help shape its responses but do not guarantee accurate or current answers.
version histories
A record of a file's versions and what changed in each. In this bootcamp, it's the VERSION HISTORY section that versioning adds to your Agent Profile, style guide or rules file, with a one-line summary for each new version.
Voice Dials · My Voice Dial
A section you can add to your Communication Style Guide, with a named setting, or audience style, for each audience you write to, such as your director or a client. When you name a setting, your agent adjusts length, opening and detail the way you defined it, without changing the facts.
work accounts · enterprise account · business account · company account
An account or workspace your organization provides or manages for work.