What you will be able to do
By the end of this learning path, you should be able to turn a vague request into a useful prompt, choose the right amount of context, control the format of an answer, evaluate whether the output meets your goal, and improve weak results through focused follow-up.
Module 1
Define the task and success standard
Before writing a prompt, decide what the AI should help you produce. “Help me with marketing” leaves the task, audience, and desired result undefined. “Draft three Facebook post ideas for first-time homeowners who need a fall maintenance checklist” gives the model a concrete job.
Write the success standard first
A success standard is one sentence describing what a usable answer must contain. It helps you judge the result instead of accepting the first fluent response.
“Write something about cybersecurity.”
“Create a one-page phishing checklist that a nontechnical employee can use before clicking an email link.”
Quick exercise
Rewrite one task you perform regularly as an outcome. Begin with an action verb such as compare, draft, summarize, explain, or organize. Then add one sentence explaining how you will know the result is useful.
Knowledge check: Which request is easier to evaluate?
“Create a five-step checklist for preparing a 20-minute team meeting.” It defines the deliverable, length, and situation. “Help me run meetings” does not provide a testable output.
Module 2
Use the five-part prompt framework
Most practical prompts can be built from five parts. You do not need every part for a simple question, but this framework helps when the result matters.
- Task: State exactly what the AI should do.
- Context: Provide the audience, situation, and facts it needs.
- Constraints: Define limits, protected facts, tone, or exclusions.
- Output: Specify the structure, length, or format.
- Quality check: Tell it to identify assumptions, uncertainties, or missing information.
Task: [What should the AI produce?]
Context: [Who is it for, and what facts matter?]
Constraints: [What must it preserve, avoid, or limit?]
Output: [What structure and length do you need?]
Quality check: [What should it flag before finishing?]
Worked example
“Draft a courteous appointment-rescheduling email. The recipient is a long-term customer, and the original meeting is Tuesday at 2:00 p.m. Offer Wednesday at 10:00 a.m. or Thursday at 3:00 p.m. Do not invent a reason for the change. Use a subject line and a message under 120 words. Before the draft, list any information you still need.”
Knowledge check: Why include “Do not invent a reason”?
It protects an important factual boundary. Without that constraint, the model may add a plausible explanation that the user never supplied.
Module 3
Control context, constraints, and format
Useful context is information that changes the answer. Include the intended reader, relevant background, source material you are authorized to use, and decisions already made. Exclude private or irrelevant information.
Choose context deliberately
- For an explanation, provide the learner’s experience level.
- For a rewrite, identify facts and phrases that must remain unchanged.
- For a comparison, name the decision criteria—not only the options.
- For a plan, state time, budget, tools, and other real limits.
Use output formats that support the task
Ask for a table when comparing the same criteria across several choices, numbered steps for a process, bullets for a quick checklist, or short paragraphs for an explanation. Do not request a table merely because it looks organized.
Quick exercise
Take the prompt from Module 1. Add only the context that would materially change the response, two useful constraints, and the most practical output format.
Knowledge check: Is “Be very smart and professional” a useful constraint?
Usually not. It is subjective and difficult to test. “Use plain English, define technical terms, and keep the answer under 300 words” gives measurable direction.
Module 4
Improve a weak AI response
The first response is a draft. Diagnose the main problem before asking for a revision. A focused follow-up preserves good work and makes the change easier to evaluate.
| Problem | Focused follow-up |
|---|---|
| Too generic | “Keep the structure, but replace general advice with one example for a two-person service business.” |
| Too long | “Preserve the five recommendations and all numbers. Reduce the explanation to 250 words.” |
| Wrong audience | “Rewrite for a first-time user. Define each technical term when it first appears.” |
| Unsupported claims | “Separate statements supported by my notes from assumptions. Remove claims that cannot be verified.” |
| Wrong tone | “Use a warm, direct tone. Avoid exaggerated promises, slang, and exclamation marks.” |
A simple revision loop
- Name what is already useful.
- Identify the largest remaining problem.
- State the exact change.
- Repeat any facts or constraints that must remain protected.
- Review the new version before making another change.
Knowledge check: Why is “make it better” a weak follow-up?
It does not define what “better” means. The model must guess whether you want changes to length, accuracy, tone, structure, detail, or audience.
Module 5
Verify the output and complete a real task
A well-written prompt can improve relevance; it cannot guarantee truth. Review names, dates, quotations, calculations, links, product capabilities, legal requirements, and other consequential claims using current authoritative sources.
Final review checklist
- Does the response complete the requested task?
- Does it follow the audience, length, tone, and format requirements?
- Did it preserve the facts and constraints you supplied?
- Did it add unsupported claims or assumptions?
- Have you verified important information independently?
- Would you be comfortable taking responsibility for the final version?
Capstone exercise
Choose a low-risk task that matters to you: a checklist, short email, study outline, meeting agenda, or comparison. Write a first prompt using the five-part framework. Save the first response. Use two focused follow-ups, then compare the versions and record which instruction produced the largest improvement.
Knowledge check: Does a detailed prompt eliminate the need for verification?
No. Prompt quality improves direction and consistency, but the model can still misunderstand the task, use outdated information, or generate inaccurate details.
Frequently asked questions
What is an AI prompt?
An AI prompt is the instruction and context you give an AI system. It may include a question, task, source material, constraints, and the format you want returned.
Do I need special prompt-engineering commands?
No. Most everyday tasks improve when you clearly define the outcome, relevant context, constraints, and desired format. Specialized technical work may require additional methods, but clarity comes first.
How long should a prompt be?
Use enough detail to remove important ambiguity. A simple transformation may need one sentence; a consequential plan may need several paragraphs and source material. Longer is not automatically better.
Should I tell the AI to act as an expert?
A role can help establish perspective or vocabulary, but it does not give the system verified expertise or make its claims correct. Concrete task requirements are more important.
Can I reuse the same prompt?
Yes. Save a template when the task repeats, but update the context and review the output each time. Do not assume a template will work equally well across different AI systems.
What should I do when the AI asks for missing information?
Provide the missing information if it is relevant and safe to share. Otherwise, ask the AI to proceed using clearly labeled assumptions or narrow the task.
Continue with a practical guide
Recognize the most frequent prompt-writing problems and see how small changes improve the result.
Read: 10 Common AI Prompt Mistakes Beginners Make—and How to Fix Them
About this learning path
Developed with AI assistance and reviewed, edited, and tested by the AI Learning Gym Editorial Team. Every learning path is checked for clarity, accuracy, practical usefulness, and responsible AI guidance. It includes original examples, exercises, and knowledge checks designed to help learners complete a specific real-world task.
Published: September 25, 2026
Last reviewed: September 25, 2026