Step 3 of 16
Not magic prompts, but the thinking framework behind good prompts — role, context, constraint, format.
ไม่ใช่ 10 prompt เด็ด แต่เป็นกรอบคิดที่ทำให้คุณเขียน prompt ได้ดีขึ้น — role, context, constraint, format
Every effective prompt has four elements. You don't need all four every time, but the more you include, the more predictable the output.
You are a senior product manager at a tech startup.
You are a Thai language teacher correcting student essays.
You are a financial analyst reviewing quarterly earnings.
Our company sells B2B SaaS to banks in Southeast Asia.
I'm writing an email to a client who complained about delayed delivery.
This is a follow-up to last week's meeting where we agreed on 3 action items.
Write a 3-paragraph email apologizing and offering a 10% discount.
Summarize this article into 5 bullet points.
Compare option A and option B in a table format.
Respond in Thai. Use professional tone.
Output as a markdown table with columns: Feature, Option A, Option B.
Keep it under 200 words. Use bullet points, not paragraphs.
Bad prompt:
Write me an email about the project delay.
Good prompt:
You are a project manager at a software consultancy.
Context: Our team is 2 weeks behind on the mobile app project
for Client X. The delay was caused by a third-party API change
that required unexpected rework.
Task: Write an email to the client informing them about the delay,
explaining the cause, and proposing a revised timeline.
Format: Professional but warm tone. Under 300 words.
End with a clear ask for a 15-minute call this week.
Write in English.
Don't rewrite the whole prompt. Fix one thing at a time:
| Problem | Fix |
|---|---|
| Too vague | Add more context |
| Wrong tone | Specify tone explicitly |
| Too long/short | Set word count or bullet limit |
| Off-topic | Narrow the task scope |
| Wrong format | Specify table, list, paragraph, code |
Example iteration:
First attempt: "Summarize this article"
→ Too long
Second attempt: "Summarize this article in 5 bullet points"
→ Missing key points
Third attempt: "Summarize this article in 5 bullet points,
focusing on the financial impact and next steps"
→ Good
These patterns come from AI research but are practical for anyone. Use them when a simple prompt isn't enough.
No examples, just instructions. Works for simple, common tasks where AI already knows the pattern.
Classify this email as "urgent" or "normal":
[paste email]
If the result is wrong or inconsistent, switch to few-shot.
Give 2-3 examples so AI learns the exact pattern you want. This is the single most effective way to improve output quality.
Classify these emails:
Email: "Server down, all users affected"
Label: Urgent
Email: "Weekly newsletter subscription"
Label: Normal
Email: "Payment failed for order #1234"
Label: Urgent
Email: [paste your email]
Label:
Rules for good examples:
Ask AI to reason before answering. Dramatically improves accuracy for complex questions, math, planning, and analysis.
Think step by step.
1. First, identify the root cause of the problem
2. Then, list 3 possible solutions
3. Then, compare pros and cons of each
4. Finally, recommend the best one with reasoning
Problem: [your problem]
You can also just add "Think step by step before answering" to any prompt.
Sometimes telling AI what to avoid is more effective than telling it what to do.
Write a product description.
- Do NOT use superlatives (best, amazing, incredible)
- Do NOT mention pricing
- Do NOT exceed 100 words
- Focus on features and use cases only
Don't ask AI to do everything in one prompt. Chain prompts for multi-step work:
Step 1 prompt: "Analyze this report and list the 3 key findings"
Step 2 prompt (paste findings): "For each finding, suggest one action item"
Step 3 prompt (paste action items): "Write a summary email to the team"
Each step gives AI focused context, producing better results than one giant prompt.
Good prompting is good communication. If a human colleague wouldn't understand your request, AI won't either. Be specific, provide context, and iterate.