Prompting Is the Door, Not the Destination

Audience

Adults with mixed technical experience, including people who have never used a chatbot, casual users, small-business owners, employees, community members, older adults, and interested but skeptical learners. No technical background is required. Learners may write their own request, revise a supplied request, work with a partner, read the examples, or observe a facilitator demonstration.

What Do You Need

Level: beginner · Format: individual, pair, observation, or facilitated demonstration

No learner account is required. Learners may use an approved general-purpose chatbot, watch a facilitator use one on a shared screen, or work from the supplied examples in One Task, Three Conversations. A paid account is not required.

Useful materials:

• the printable or digital Learner Notes;
• the conversation worksheet, if the learner wants to record the comparison; and
• one fictional or ordinary low-risk task from the Activity, or another task that does not involve consequential decisions or sensitive information.

Learning Outcomes

By the end of this module, learners should be able to:

• State what they are trying to accomplish.
• Provide relevant background without oversharing.
• Describe the type of response they need.
• Add useful constraints or boundaries.
• Inspect the response rather than automatically accepting it.
• Improve the response through follow-up.
• Recognize when more context or outside verification is needed.
• Pause before sharing private or sensitive information.
• Keep responsibility for consequential decisions with an appropriate human or process.

Four Useful Handles

A useful first request may include:

Goal + relevant context + desired output + boundaries

Treat these as handles, not a rigid formula:

• What are you trying to do? State the actual goal rather than a few disconnected keywords.
• What does the chatbot need to know? Add only the background or audience information that matters.
• What should the result look like? Describe the form, length, tone, structure, or intended audience.
• What should it be careful about? Name exclusions, uncertainty, privacy concerns, or practical constraints.

You do not need to place everything into the first message. Start naturally, inspect what happens, and add what matters.

The Repair Loop

The goal is not to engineer a perfect prompt before pressing Enter. Stay engaged and steer:

Ask → Inspect → Notice → Adjust → Check again

Useful follow-ups include:

• What assumptions did you make?
• What information is missing?
• Explain that term in plain language.
• Give me three different options.
• Shorten the second paragraph.
• What could be wrong with this answer?
• Ask me what you need to know.

A weak first response is information. It may reveal a missing fact, an unclear goal, or an assumption that needs correction.

Two Safety Gates

Keep information going into the chatbot separate from information coming out of it.

Before You Put Information In

Pause and ask:

• Is this private, confidential, or someone else’s information?
• Am I authorized to share it?
• Can I remove names, identifiers, or unnecessary details?
• Does my organization have rules about which tools or information may be used?
• Can I use a fictional or sanitized substitute instead?

Before You Use What Comes Out

Pause and ask:

• Does this contain factual claims that need an outside source?
• Could the information be outdated, incomplete, or based on an assumption?
• Am I treating this as a draft, an idea, a recommendation, or a final decision?
• Could acting on it affect someone’s health, money, employment, reputation, rights, or safety?
• Does a qualified human or approved process need to review it?

Be thoughtful about what goes in. Be responsible for what comes out.

Progressive Demonstration

Watch how relevant information changes a response without assuming a longer prompt is always better.

Start broad:

Help me plan a community event.

Then add only what the task needs:

Help me plan a beginner-friendly workshop about generative AI. It will be held at a neighborhood coffeehouse for about 15 adults with mixed technical experience. Create a 60-minute agenda with a welcome, short demonstration, hands-on activity, and closing discussion. Avoid technical jargon. Do not require attendees to purchase an account or share private information.

Ask:

• What changed?
• Which added detail mattered most?
• What did the chatbot assume before the detail was added?
• What still needs human review?

Practice: One Task, Three Conversations

One Task, Three Conversations is the primary practice for this module. Shorter and extended routes are available.

Use one low-risk task for three moves:

1. Ask naturally.
2. Add the goal, relevant context, desired output, and one useful boundary.
3. Continue the second conversation with one focused follow-up.

Compare what became more useful, what the chatbot assumed or invented, what remained missing, what you chose not to share, and what still needs human review or outside verification. Use the conversation worksheet if you want a place to keep the comparison.

Reflection

• Which detail or boundary changed the response most?
• What did the chatbot assume?
• What did you deliberately keep out of the conversation?
• What would need checking before you used the result?
• Would another prompt help, or is it time to use another source, ask a person, restart, or stop?

Keep the four useful handles, the repair loop, and the two safety gates. A completed worksheet is useful when you do the Activity, but the point is the judgment you practice—not the paper you retain.

Where to Go Next

Continued participation is optional.

• Conversation Engineering develops follow-up, correction, comparison, checkpoints, and knowing when to continue, steer, restart, or stop.
• Context Engineering explores what information the chatbot can see, what is missing or irrelevant, and how sources and boundaries shape the work.

Tool Boundary

An approved general-purpose chatbot may support the demonstration and Activity, but no learner account or paid tool is required. The four handles and repair loop are guides for judgment, not promises that a particular wording will produce the same result across tools, models, or conversations.

Privacy Boundary

Use fictional, public, learner-owned, or sanitized low-risk material. Do not enter personal, client, student, employee, health, legal, financial, account, credential, safety, or another person’s information. Stop or switch tasks when ownership, permission, consequence, or tool approval is unclear.