Nothing in this guide requires a technical background. It is written for leaders who are excellent at what they do and who did not get there by memorizing software manuals. Every term is defined the way you would explain it to a smart colleague who has been off the grid for two years. If something is still unclear after reading it, bring the question to Session 1. That is exactly what Session 1 is for.
How to Access AI
The four ways you get there
AI tools are not software you install once and find on your desktop forever. They live in multiple places and you can reach them however is most convenient for the moment. Here are the four primary ways to access the tools we will use in this program.
Web Browser Version
Open Chrome, Safari, Firefox, or Edge and go to the tool's web address. For Claude that is claude.ai. For ChatGPT that is chatgpt.com. For Gemini that is gemini.google.com. Log in, see a text box, type in it. That is the whole thing. Nothing to install.
Desktop Application
A downloadable app that lives on your computer. Slightly smoother than the browser version because it does not compete with your other tabs. For most executives the browser and desktop app feel nearly identical. An option, not a requirement.
Mobile App
Same tool, smaller screen. The reason to use it is voice input — tap the microphone, speak naturally, and the AI receives your spoken words as a prompt. For executives who think faster than they type, this is often the highest-leverage access point in the program.
Browser Extension
A small add-on that attaches to your browser and lets you access AI without navigating away from what you are doing. Highlight an email, open the AI sidebar, draft a response right there. A workflow efficiency tool for when you are already comfortable with the basics.
Where to Start
Start with the web browser version. It works on every device, requires no installation, and is exactly what we will use in our sessions together. Once you are comfortable, explore the mobile app for voice input and on-the-go use. Everything else is optional and can wait.
The Parts of an AI Interaction
A map of what you are looking at
When you open an AI tool for the first time the interface is usually simpler than you expect. Here is a map of what you are looking at and what each part does.
The Conversation Window
The large area that takes up most of the screen. Your messages appear on one side, typically the right or bottom, and the AI's responses appear on the other. As the conversation continues the window scrolls down to show the history of the exchange.
Think of it exactly like a text message thread. You send something. You get something back. You respond. It goes back and forth. The conversation window is just a text message thread with a very capable correspondent.
The Input Box and Send Button
The rectangular box at the bottom of the screen is where your prompt goes. You click into it, type your message, and press Enter or click the send button. Some input boxes also have a microphone icon for voice input.
Pressing Shift and Enter creates a new line within your message without sending it — useful when you want to write a longer, multi-paragraph prompt.
The Response Area
The space where the AI's reply appears. Responses stream in word by word in real time rather than appearing all at once. This is normal. You do not need to wait for the full response before you start reading.
After the response finishes you can respond in the input box, ask a follow-up, tell it what to change, or start a completely new direction. The conversation continues as long as you want it to.
The New Conversation Button and History
Usually labeled New Chat or found as a plus icon in a sidebar. Clicking this starts a fresh conversation with no history from previous exchanges. Use this when you are starting a genuinely new topic.
Within a single conversation the AI remembers everything that has been said. Across different conversations it does not. If you start a new conversation and want the AI to have context from a previous one, you need to provide that context in your new prompt.
A list of your previous conversations is usually visible in a sidebar on the left side of the screen. You can click any previous conversation to return to it and continue where you left off.
The Core Terminology
Words we will use and what they mean
A prompt is what you type into the input box to start or continue a conversation with AI. It is your instruction, your question, your request. It can be one sentence or several paragraphs. It can be casual or highly structured. The quality of what you get back is directly related to the clarity of what you put in.
The word comes from the theater. A prompter whispers lines to an actor who has forgotten them. You are prompting the AI to produce something. That is the whole concept.
Every interaction with AI begins with a prompt. Learning to write better prompts is the primary skill we build in Session 2.
What the AI produces in reply to your prompt. Also sometimes called a completion, a generation, or the AI's answer. We will use the word response or output throughout the program.
Responses are not retrieved from a database of correct answers. They are generated fresh each time based on the patterns the AI learned during training and the specific context you provided in your prompt. This is why two slightly different prompts can produce notably different responses.
The information the AI has available when generating a response. Context includes everything said in the current conversation plus whatever you include in your current prompt.
Context is one of the most important concepts in using AI effectively. The more relevant context you provide, the better the response. If you ask AI to help you draft a board memo without telling it anything about your organization, your board, or the situation, you will get a generic memo. If you tell it who you are, what your organization does, who the audience is, and what tone fits the relationship, you will get something that sounds like it came from your desk.
Providing good context is the difference between AI that feels like a powerful tool and AI that feels like a disappointing parlor trick.
When AI generates something that is factually incorrect, fabricated, or misleading while presenting it with complete confidence. The term comes from the psychological phenomenon of perceiving something that is not there.
Hallucinations happen because AI generates plausible responses based on patterns rather than retrieving verified facts. It does not know what it does not know and it does not always flag uncertainty clearly.
This is why the rule you verify before it leaves your desk exists. It is not a flaw that will be entirely eliminated. It is a characteristic of how these systems work that requires your judgment as the human in the loop. We cover real-world hallucination examples in detail during the program.
The process of refining an AI response through follow-up prompts. Rather than expecting the first response to be perfect, you treat it as a draft and iterate toward what you actually need.
Iteration is not a sign that AI is failing. It is the intended workflow. The executives who get the best results from AI are almost always the ones who iterate most naturally, treating the exchange like a working conversation rather than a one-shot query.
The practice of keeping a human involved in reviewing, judging, and taking responsibility for AI outputs before they are used or acted upon. The human in the loop reads what AI produces, evaluates whether it is accurate and appropriate, makes any necessary changes, and takes ownership of the final output.
In the context of this program, you are always the human in the loop. AI assists. You decide. That relationship does not change regardless of how capable the tools become.
The technical category that most AI tools fall into. A large language model is an AI system trained on enormous amounts of text that has learned to generate language in response to prompts. Claude, ChatGPT, Gemini, and Grok are all large language models.
The practical implication: these tools are fundamentally language tools. They are exceptionally good at anything involving reading, writing, summarizing, explaining, and generating text. They are less reliable for tasks requiring precise calculation, real-time information, or guaranteed factual accuracy.
The underlying AI system that powers the tool you are using. Different models have different capabilities, strengths, and behaviors. Claude is a model made by Anthropic. GPT-4 is a model made by OpenAI and powers ChatGPT. Gemini is a model made by Google.
When people talk about AI getting smarter or a new version being released they are usually talking about a new model. You do not need to understand how models work technically. Think of the model as the engine under the hood, and different engines have different personalities and capabilities, which we cover in the program.
A newer category of AI that can take a goal, break it into steps, and execute those steps autonomously without a human prompting each individual action.
Basic AI responds to one prompt at a time. Agentic AI can chain actions together, use tools, browse the web, write and run code, and complete multi-step tasks with minimal human intervention between steps.
Most of what we use in this program is basic conversational AI. Agentic AI is where the field is heading, and understanding that it exists prepares you for the governance conversations coming in your organization. When AI can act autonomously, the human-in-the-loop question becomes significantly more complex and significantly more important.
Consumer AI tools are the free or personal subscription versions. They are accessible to anyone, require only a personal account, and are designed for individual use.
Enterprise AI tools are versions of the same products with additional features and protections for organizational use. Enterprise agreements typically include explicit commitments about how your data is handled, guarantees that your inputs will not be used to train the model, administrative controls, and compliance documentation.
The practical implication: consumer tools are fine for personal productivity use with non-sensitive information. Organizational deployment, especially with information that touches clients, donors, employees, or legal matters, should use enterprise-grade tools with appropriate agreements in place. We cover this in detail in Session 3.
The unit of text that AI models process. A token is roughly three quarters of a word on average. You do not need to think about tokens directly, but the concept explains some AI behaviors.
The most relevant thing to know: AI tools have a context window limit, meaning there is a maximum amount of text they can process in a single conversation. For most current tools this limit is large enough that you will never hit it in normal use. For very long documents you may need to work in sections.
A technical setting that controls how creative or conservative an AI's responses are. High temperature produces more varied, unexpected, creative outputs. Low temperature produces more predictable, focused, conservative outputs.
Most consumer AI tools manage this automatically and you never need to adjust it directly. It is worth knowing the concept exists because it explains why asking for creative brainstorming produces different-feeling output than asking for a precise factual summary. The tool is adjusting its behavior based on the type of task.
Instructions given to the AI before a conversation begins that shape its behavior throughout the session. In most consumer tools the system prompt is set by the tool developer and you do not see or interact with it directly.
In enterprise deployments, organizations can customize the system prompt to give the AI specific instructions relevant to their context — for example, always respond in our organization's brand voice, or never discuss competitor products. This is why the same underlying model can behave differently in different contexts and applications.
The Three Rules
The only things you need to remember before Session 1
If you remember nothing else from this guide, remember these three things. They apply to every interaction with AI in every context for the rest of your career.
You are always the author.
Anything AI produces that leaves your desk has your name on it. The thinking, the judgment, and the accountability are yours. AI helped you get there. The output reflects you.
Verify before it goes anywhere.
AI can be confidently wrong. Read what it produces like an editor reviewing a draft, not like a passive recipient accepting a final document. If something sounds off, push back on it.
The first response is a draft, not a decision.
Your instinct to refine, redirect, and push back on AI output is correct. Use it. The conversation continues until you have what you actually need.
This guide is a living reference. Bring it to your sessions. Add to it as you learn. The best version of this document is the one you have marked up, questioned, and made your own.
If something here raised a question, bring it to Session 1. That is exactly what Session 1 is for. Nothing in this program requires you to have figured anything out on your own first.