Skip to content
HUB

🎯 Advanced Prompting Techniques

Level up your prompting skills with proven advanced techniques

🌱 What You’ll Learn

Section titled “🌱 What You’ll Learn”

This guide covers professional-grade prompting techniques:

  • Chain-of-thought reasoning
  • Few-shot learning examples
  • Structured output formats
  • Prompt chaining
  • Meta-prompting basics

🧠 Chain-of-Thought (CoT) Prompting

Section titled “🧠 Chain-of-Thought (CoT) Prompting”

What It Is

Section titled “What It Is”

Chain-of-thought prompting guides the AI through step-by-step reasoning before answering. This improves accuracy on complex problems.

Basic Pattern

Section titled “Basic Pattern”
Solve this problem step by step:
[YOUR PROBLEM]


Think through:
1. What information do I have?
2. What approach should I use?
3. Work through each step
4. Verify the answer makes sense

Example: Complex Calculation

Section titled “Example: Complex Calculation”
Calculate the monthly payment for a $300,000 mortgage at 6.5% interest over 30 years.


Think through:
1. Identify the formula needed (monthly payment formula)
2. Convert annual rate to monthly: 6.5% / 12 = 0.542% per month
3. Convert rate to decimal: 0.00542
4. Calculate number of payments: 30 years × 12 = 360 months
5. Apply formula: M = P[r(1+r)^n]/[(1+r)^n-1]
6. Show your work step by step
7. Verify answer is reasonable ($300K over 30 years should be ~$2000/month range)


Now calculate and show each step.

💡 When to Use CoT

Section titled “💡 When to Use CoT”

✅ Use for:

  • Math problems
  • Logic puzzles
  • Multi-step analysis
  • Complex reasoning
  • Fact-checking

❌ Skip for:

  • Simple questions
  • Creative writing
  • Quick lookups
  • Brainstorming

📝 Few-Shot Learning

Section titled “📝 Few-Shot Learning”

What It Is

Section titled “What It Is”

Provide 2-5 examples of what you want, then the AI follows the pattern for new inputs.

Basic Pattern

Section titled “Basic Pattern”
I'll show you examples, then you do the same for my input.


Example 1:
Input: [example input]
Output: [example output]


Example 2:
Input: [example input]
Output: [example output]


Example 3:
Input: [example input]
Output: [example output]


Now do this for:
Input: [your actual input]
Output: ?

Example: Data Formatting

Section titled “Example: Data Formatting”
Format customer data into this structure. Here are examples:


Example 1:
Input: "John Smith, email: john@email.com, purchased: laptop"
Output: {name: "John Smith", contact: "john@email.com", item: "laptop", status: "customer"}


Example 2:
Input: "Sarah Lee, phone: 555-1234, inquired about: tablets"
Output: {name: "Sarah Lee", contact: "555-1234", item: "tablets", status: "lead"}


Example 3:
Input: "Mike Chen, email: mike@test.com, returned: headphones"
Output: {name: "Mike Chen", contact: "mike@test.com", item: "headphones", status: "return"}


Now format:
Input: "Amy Davis, phone: 555-9876, purchased: keyboard"
Output: ?

💡 Few-Shot Best Practices

Section titled “💡 Few-Shot Best Practices”
  1. Use diverse examples - Cover edge cases
  2. Keep consistent format - Make pattern clear
  3. Include 3-5 examples - 2 is minimum, 5+ is often overkill
  4. Show variations - Different inputs, same output format
  5. Label clearly - Make input/output obvious

📊 Structured Output Formats

Section titled “📊 Structured Output Formats”

JSON Output

Section titled “JSON Output”
Provide your answer as valid JSON with this structure:
{
  "summary": "brief overview",
  "key_points": ["point 1", "point 2", "point 3"],
  "recommendations": ["rec 1", "rec 2"],
  "confidence": "high/medium/low",
  "sources": ["source 1", "source 2"]
}


Question: [YOUR QUESTION]

Table Format

Section titled “Table Format”
Organize the information in a markdown table with these columns:
| Feature | Option A | Option B | Option C | Winner |


Include:
- 5-7 key comparison features
- Objective ratings
- Brief explanations
- Clear recommendation in "Winner" column


Compare: [YOUR COMPARISON REQUEST]

List Format

Section titled “List Format”
Provide your answer as a prioritized list:


Format:
1. [ITEM] - Priority: High/Medium/Low
   - Why: [reason]
   - Action: [what to do]
   - Timeline: [when]


2. [ITEM] - Priority: High/Medium/Low
   ...


Request: [YOUR REQUEST]

🔗 Prompt Chaining

Section titled “🔗 Prompt Chaining”

What It Is

Section titled “What It Is”

Break complex tasks into multiple prompts, using output from one as input to the next.

Pattern

Section titled “Pattern”
Step 1: Research Phase
"List the top 5 approaches to [PROBLEM]. For each:
- Core concept
- Pros and cons
- Best use case"


[Take output, analyze, then:]


Step 2: Selection Phase
"Based on this analysis: [PASTE STEP 1 OUTPUT]


Recommend the best approach for my specific situation:
- [Constraint 1]
- [Constraint 2]
- [Goal]"


[Take recommendation, then:]


Step 3: Implementation Phase
"Create a detailed implementation plan for [SELECTED APPROACH].
Include:
- Step-by-step actions
- Timeline
- Resources needed
- Success metrics"

Example: Content Strategy

Section titled “Example: Content Strategy”
Prompt 1: "Analyze these 3 blog posts and extract:
- Target audience
- Writing style
- Key themes
- Engagement tactics
[PASTE POSTS]"


Prompt 2: "Using this analysis [PASTE RESULTS], create:
- Content calendar for next month
- 10 blog post titles matching this style
- Optimal posting schedule"


Prompt 3: "Take the top 3 titles and create:
- Full outlines
- SEO keywords
- Meta descriptions"

🎭 Role-Based Prompting

Section titled “🎭 Role-Based Prompting”

Assign Expertise

Section titled “Assign Expertise”
You are a [SPECIFIC EXPERT] with [X YEARS] experience in [DOMAIN].


Your expertise includes:
- [Skill 1]
- [Skill 2]
- [Skill 3]


Approach: [analytical/creative/practical]
Tone: [professional/casual/technical]


Now help me with: [YOUR REQUEST]

Example Roles

Section titled “Example Roles”

Data Analyst:

You are a senior data analyst with 10 years in e-commerce analytics.
You specialize in conversion optimization and A/B testing.
You communicate insights clearly for non-technical stakeholders.


Analyze this data: [DATA]
Provide: insights, recommendations, and next steps.

Content Strategist:

You are a content strategist for B2B SaaS companies.
You focus on thought leadership and SEO-driven content.
Your recommendations are data-backed and actionable.


Create a content strategy for: [COMPANY/PRODUCT]

⚡ Power Techniques

Section titled “⚡ Power Techniques”

Negative Prompting

Section titled “Negative Prompting”

Tell the AI what NOT to do:

Write a product description for [PRODUCT].


Do:
- Focus on benefits, not just features
- Use specific numbers and data
- Include customer pain points
- Keep it scannable


Don't:
- Use hype or exaggeration
- Include technical jargon
- Make it longer than 150 words
- Use passive voice

Constraint-Based Prompting

Section titled “Constraint-Based Prompting”

Add specific constraints for better control:

Write a blog intro about [TOPIC].


Constraints:
- Exactly 3 sentences
- First sentence must be a question
- Include the word "proven" once
- Reading level: 8th grade
- Hook must relate to reader's pain point

Iterative Refinement

Section titled “Iterative Refinement”
First Draft:
"Write [REQUEST]"


[Review output]


Refinement 1:
"Revise the above to:
- Make it 30% shorter
- Add more specific examples
- Strengthen the conclusion"


[Review again]


Refinement 2:
"Final polish:
- Improve headline
- Add transition sentences
- Ensure consistent tone"

🔬 Advanced Patterns

Section titled “🔬 Advanced Patterns”

Meta-Prompting Preview

Section titled “Meta-Prompting Preview”
Before you answer my question, tell me:
1. What approach will you use?
2. What assumptions are you making?
3. What information might be missing?
4. What's your confidence level?


Then answer: [YOUR QUESTION]

Self-Correction

Section titled “Self-Correction”
Answer this question: [QUESTION]


After answering:
1. Identify any assumptions you made
2. Point out potential weaknesses in your answer
3. Suggest what additional information would improve it
4. Rate your confidence: high/medium/low

📚 Practice Exercises

Section titled “📚 Practice Exercises”

Exercise 1: Chain-of-Thought

Section titled “Exercise 1: Chain-of-Thought”

Take this simple question and turn it into a CoT prompt: “Which smartphone should I buy?”

Exercise 2: Few-Shot

Section titled “Exercise 2: Few-Shot”

Create a few-shot prompt to format addresses consistently.

Exercise 3: Structured Output

Section titled “Exercise 3: Structured Output”

Design a JSON schema for book recommendations that includes:

  • Title, author, genre
  • Why it’s recommended
  • Similar books
  • Where to buy

🎯 When to Use What

Section titled “🎯 When to Use What”

Quick Answer Needed:

  • Simple, direct prompt

Complex Problem:

  • Chain-of-thought

Consistent Format:

  • Few-shot learning

Integration with Code:

  • Structured output (JSON)

Multi-Step Task:

  • Prompt chaining

Specialized Domain:

  • Role-based prompting

💡 Pro Tips

Section titled “💡 Pro Tips”
  1. Save successful prompts - Build a personal library
  2. Test variations - Small changes can have big effects
  3. Combine techniques - Use CoT + structured output together
  4. Iterate quickly - Don’t overthink, just try it
  5. Document what works - Note context and results

📚 Related Guides

Section titled “📚 Related Guides”

Next Level:

Apply This:

Part of the Intermediate AI Mastery Cookbook - HUB by CURATIONS

← Back to Intermediate