Start Lesson
Here is a prompt most people would write:
Write me an email about the missed meeting.
And here is what they get back: a 300-word generic mess that opens with "I hope this email finds you well," apologizes three times, and says nothing useful. Now look at this version:
You are a senior account manager at a digital marketing agency.
Your client, a mid-size e-commerce brand, missed your last
scheduled meeting without explanation.
Write a 3-paragraph follow-up email that:
1. Acknowledges the missed meeting without over-apologizing
2. Reaffirms the value of the partnership with one specific result
3. Proposes two new meeting times
Tone: professional but warm. Under 150 words. Do not use the
phrase "I hope this email finds you well."
Same task. The second prompt takes 30 extra seconds to write and saves 10 minutes of editing. The difference is structure, and that structure has a name.
After this lesson, you will be able to: break any prompt into three components (Context, Goal, Constraints) and write prompts that produce usable first drafts instead of generic filler.
Every effective prompt has three parts: Context, Goal, and Constraints. Miss any one of them, and the output suffers. This is the foundation everything else in this course builds on.
An AI model starts every conversation with zero knowledge of your situation. It does not know your job, your industry, your audience, or what happened five minutes ago. Every detail you leave out is a gap the model fills with the most generic, average version of what you asked for.
Context answers four questions:
Notice these are not vague descriptions. Each one gives the model a concrete fact it can use to shape the output.
"Help me with this" is not a goal. A goal tells the AI what to produce, in what format, and what structure to follow.
Here is the test: if someone handed you the output, how would you know it was done well? Your goal statement should answer that.
| Vague Goal | Precise Goal | |---|---| | "Write something about the missed meeting" | "Write a 3-paragraph email that acknowledges, reaffirms, and proposes" | | "Give me some marketing ideas" | "List 5 Instagram Reel concepts for a coffee brand targeting remote workers, each with a hook and CTA" | | "Help me with my resume" | "Rewrite the Experience section using quantified achievements in the format: Action verb + result + metric" |
The precise version tells the model the format, the structure, and the success criteria.
Constraints keep the AI from wandering. They are not limitations -- they are specifications that cut your editing time.
Four types of constraints that change output quality immediately:
Exclusions are the most underused constraint. If you keep editing the same phrases out of AI output, add them as exclusions and stop repeating that work.
[CONTEXT]
I am a [your role] at [company type]. [1-2 sentences about
the situation].
[GOAL]
Write a [format] that [specific structure or sections].
[CONSTRAINTS]
Tone: [tone]. Length: [word count or item count].
Do not include: [specific things to exclude].
Paste this into Claude, GPT-4, or Gemini. Fill in the brackets. The output will be dramatically better than an unstructured request.
I am a [role] evaluating [decision]. I need a brief that
covers:
1. The two best options with pros and cons for each
2. The key risk for each option
3. A recommendation with one sentence of reasoning
Audience: [who will read this]. Length: under 250 words.
Use a comparison table for the options.
Expected output: A short document with a two-row comparison table, a risk flag for each option, and a single clear recommendation. This format works for vendor selection, tool evaluation, or strategic choices.
I wrote the following [document type] for [audience]:
---
[Paste your draft here]
---
Review it for: [2-3 specific criteria, e.g., clarity, tone
match, missing information].
For each issue, quote the specific sentence and suggest a
concrete revision. Do not rewrite the whole document.
Expected output: A numbered list of specific issues, each with the original sentence quoted and a suggested replacement. This is faster than asking someone to "take a look" and hoping they give useful feedback.
Pick a real task you need to do this week -- an email, a document, a plan. Write it as a CGC prompt using Template 1 above. Then check your work:
If your output still feels generic, the fix is almost always in the Context. Add one more specific detail about the situation and run it again.
You now have the CGC framework -- the skeleton of every good prompt. In the next lesson, you will learn role prompting: how adding "Act as a [specific expert]..." to the Context section changes the entire angle and depth of the output. CGC gives you structure. Roles give you expertise.