Course OverviewFree AI CourseBlogReserve Bootcamp Seat
Prompt Engineering · Day 1 of 5 ~50 minutes

The Foundation — Why 90% of Prompts Fail

Most people write prompts like Google searches. That's the mistake. Today you'll learn the 4-element framework that makes every prompt better — and see 10 real before/after examples.

1
Day 1
2
Day 2
3
Day 3
4
Day 4
5
Day 5
What You'll Do Today

Learn the 4-element framework (Role, Context, Task, Format), see 10 before/after prompt comparisons that show exactly why the good ones work, and rewrite 10 prompts of your own using the framework.

1
The Problem

Why Most Prompts Fail

When someone opens ChatGPT or Claude and types "summarize this report," they're treating a powerful AI like a search engine. Search engines work on keywords. AI language models work on context, role, and structure.

The result is generic output that doesn't match what you actually needed — which leads most people to conclude "AI isn't that useful for my work." That conclusion is wrong. The prompt is wrong.

There are four elements that every good prompt has. Missing any one of them degrades the output significantly.

2
The Framework

The 4-Element Framework: RCTF

R
Role
Tell the AI who it is. "You are a senior financial analyst with 15 years at Goldman Sachs." This activates relevant knowledge and communication style.
C
Context
Provide the situation. Who is the audience? What's the goal? What constraints exist? "This is for a board presentation to non-technical executives."
T
Task
State clearly what you want done. Be specific. "Summarize the key financial risks from this Q3 earnings report" beats "summarize this."
F
Format
Specify how you want the output. "3 bullet points, each under 20 words, starting with a financial metric" gets you usable output, not an essay.

You don't always need all four. For simple tasks, Role + Task is enough. For complex, high-stakes outputs, use all four. The point is to think about each one deliberately.

3
Before / After

10 Before/After Examples

✗ Before
summarize this report
No role, no context, no format. You'll get 3 paragraphs of generic summary that don't match your needs.
✓ After
You are a management consultant summarizing findings for a CEO who has 3 minutes to read this. Summarize the following Q3 earnings report. Focus on: revenue growth, cost trends, and top 2 risks. Format: 3 bullet points, each under 25 words.
Role (consultant), Context (CEO, 3 min), Task (specific focus areas), Format (3 bullets, length limit). The output is immediately usable.
✗ Before
write me an email
Write an email to who? About what? In what tone? The AI has to guess everything.
✓ After
You are a professional communications writer. I need to follow up with a client (Sarah Chen, VP of Operations at Acme Corp) who has gone quiet after our proposal 2 weeks ago. We submitted a $47K proposal for AI workflow automation. Write a short follow-up email that is warm but direct. Format: subject line + 3-sentence body. No attachments mention.
Every element is filled in. The AI can write exactly the right email because it has all the context it needs.
✗ Before
explain machine learning
Who is this for? A developer? A 10-year-old? A business executive? The AI will pick one at random.
✓ After
You are a patient technology educator. Explain machine learning to a non-technical HR director who needs to understand it well enough to champion an AI initiative internally. Avoid jargon. Use one analogy from everyday business life. Keep it under 150 words.
Now the explanation will be at exactly the right level, with the right framing, at the right length.
✗ Before
review my code
What kind of review? Security? Performance? Style? What language? What level of severity?
✓ After
You are a senior Python engineer doing a production readiness review. Review the following Python function for: (1) security issues, (2) error handling gaps, (3) performance problems. For each issue found: state the line number, severity (critical/major/minor), and a 1-sentence fix recommendation. Skip style/naming comments.
Specific role, specific scope, specific output format. You get a structured code review, not generic commentary.
✗ Before
help me with my presentation
Help how? Write slides? Review content? Suggest structure? The AI has no idea what to do.
✓ After
You are a presentation coach who has helped Fortune 500 executives present to boards. I have a 10-slide deck about our Q4 product roadmap for an investor meeting. Review the following slide titles and tell me: (1) which slides are missing, (2) which should be combined, (3) whether the narrative arc builds to a clear ask. Format: numbered list, one insight per slide.
Specific expert role, clear context, defined task, structured format. Immediately actionable feedback.
Today's Exercise

Rewrite 10 Real Prompts

Take prompts you've actually used at work in the past month — or use the ones below if you need a starting point. Apply the RCTF framework to each one. Then run both the original and rewritten version in Claude or ChatGPT and compare the outputs.

  • "write a job description for a data analyst"
  • "summarize the meeting notes"
  • "help me respond to this customer complaint"
  • "write a performance review"
  • "explain the difference between AI and machine learning"
  • "create a project plan"
  • "analyze this data"
  • "write a LinkedIn post about our product"
  • "draft a contract clause about data privacy"
  • "help me prepare for this negotiation"

What You Learned Today

  • Why most prompts fail: treating AI like a search engine instead of a collaborator
  • The 4-element framework: Role, Context, Task, Format
  • How adding each element changes the quality of the output
  • 10 before/after examples you can adapt for your own work today
Course Progress
Day 1 of 5 — 20%
Day 1 Complete

Day 2: Chain of Thought and Few-Shot

Tomorrow you'll learn the techniques that make AI reason better and give consistent outputs.

Start Day 2
← Course Overview Join the Bootcamp