Level 2: Prompting masterclass: getting great answers every time

The same AI gives a weak answer to a vague question and an excellent answer to a clear one. Writing clear instructions is called prompting, and it's the most valuable AI skill for everyone, from students to CEOs.

Why vague prompts fail

❌ "Write a business plan."

The AI doesn't know your business, location, budget, audience or what the plan is for. It fills the gaps with generic guesses.

✅ "I'm starting a mobile car wash in Ruiru with KSh 80,000. Write a one-page business plan for a youth fund application: services and prices, target customers, startup costs, monthly costs, expected income, and risks. Use simple English and a table for the numbers."

The RTCFE formula

Good prompts usually include five parts:

You don't need all five every time, but task + context + format fixes most weak answers.

Before and after

Techniques that work

1. Ask for questions first

"Before you write it, ask me 5 questions you need answered to do this well."

2. Give it something to work from

Paste the article, your notes, the job advert or your draft. AI is much better at improving your material than inventing from nothing.

3. Iterate, don't restart

Treat the first answer as a draft: "Shorter.", "More formal.", "Add prices in KSh.", "Remove the jargon.", "Make point 3 clearer."

4. Break big jobs into steps

Instead of "write my whole proposal", go: outline → section 1 → section 2 → review → final edit.

5. Ask it to check itself

"Review your answer. List any mistakes, missing points or assumptions."

6. Control the format

"Answer in a table with columns: Task, Who, Deadline." · "Give me exactly 3 options." · "Reply in Kiswahili." · "Use bullet points, no more than 100 words."

7. Set the audience and tone

"Explain for a farmer with primary education." · "Friendly and professional, like a bank's customer care."

8. Use examples (few-shot prompting)

Show 1 to 3 examples of what you want, and the AI copies the style and structure.

A reusable prompt template

Act as [role].
I need you to [task].
Background: [who it's for, your situation, important facts].
Requirements: [length, tone, must-include points, things to avoid].
Format: [table / bullets / steps / email / language].
First, ask me any questions you need.

Save templates for things you do often (social posts, emails, lesson plans, reports) in a notes app.

Prompting practice in Python

AI apps often build prompts from data. This program fills a template for several products, which is exactly what businesses do to generate many product descriptions at once.

template = (
    "Act as a copywriter for a Kenyan online shop.\n"
    "Write a {words}-word product description for: {name}.\n"
    "Price: KSh {price:,}. Key features: {features}.\n"
    "Tone: friendly. End with a call to order on WhatsApp."
)

products = [
    {"name": "Solar lantern", "price": 1850, "features": "8 hours light, phone charging, 1-year warranty"},
    {"name": "Ceramic jiko", "price": 1200, "features": "saves charcoal, durable, keeps heat long"},
]

for p in products:
    prompt = template.format(words=60, **p)
    print(prompt)
    print("-" * 40)

Common mistakes

• Too short and vague. Add context.
• Trying to do everything in one prompt. Split it.
• Accepting the first answer. Iterate.
• Not checking facts. Especially numbers, laws and names.
• Sharing private data you shouldn't.