Skip to main content

Chapter 3: Use Examples (Few-Shot)

Show, Don't Just Tell​

Sometimes the best way to get exactly what you want is to show the AI an example. This technique is called "few-shot learning" because you're giving the AI a few shots (examples) to learn from.

What is Few-Shot Learning?​

Instead of describing what you want, you show it:

Without example (zero-shot):

"Write product descriptions in an exciting style"

With example (few-shot):

"Write product descriptions like this example:

Example: 'Meet the ThermoMax: your coffee stays volcano-hot for 12 hours, or your money back. Engineered for the busy professionals who refuse to drink lukewarm disappointment.'

Now write one for a laptop stand."

The AI now understands exactly what "exciting style" means to you.

The Power of Examples​

Examples help AI understand:

What You WantHow Examples Help
FormatShow the exact structure you need
ToneDemonstrate the voice you want
LengthProvide a model of appropriate size
StyleShow specific word choices and patterns
Quality levelSet the bar for sophistication

How to Provide Examples​

Pattern 1: Single Example​

Here's an example of what I want:
[Your example]

Now create something similar for [your topic].

Pattern 2: Multiple Examples​

Here are examples of the style I need:

Example 1: [first example]
Example 2: [second example]
Example 3: [third example]

Now write one for: [your topic]

Pattern 3: Input-Output Pairs​

I want you to transform text like this:

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

Input: [another example input]
Output: [another example output]

Now transform this:
Input: [your actual input]

Real Examples in Action​

Example 1: Email Subject Lines​

Prompt:

Write email subject lines for my newsletter. Here are examples of my style:

  • "5 productivity hacks that actually work (tested by me)"
  • "I made $500 by accident. Here's what happened."
  • "The app that replaced 3 others (and saved me hours)"

Write 3 subject lines for my email about morning routines.

Why it works: AI learns your personal style—casual, specific, curiosity-driving.

Example 2: Code Comments​

Prompt:

Write code comments in this style:

Example:

// Validate user input before processing
// Returns: cleaned string or null if invalid
// Note: strips HTML to prevent XSS

Now write a comment for a function that calculates shipping costs.

Why it works: AI matches your documentation style exactly.

Example 3: Product Reviews​

Prompt:

I'm responding to customer reviews. Match this tone:

Example review response: "Thank you so much for the kind words, Sarah! We're thrilled your daughter loved the dinosaur puzzle. 🦕 Can't wait to see what adventure you choose next!"

Write a response for a review that says: "Great quality, arrived fast. Will buy again."

Why it works: AI captures your brand voice including emoji use and enthusiasm level.

Example 4: Data Formatting​

Prompt:

Convert company information into this format:

Input: Apple Inc, Cupertino, California, Technology Output: Apple Inc (Technology) - Cupertino, CA

Input: Toyota Motor Corporation, Toyota City, Japan, Automotive
Output: Toyota Motor Corporation (Automotive) - Toyota City, Japan

Now convert this: Input: Microsoft Corporation, Redmond, Washington, Technology

Why it works: AI learns the exact transformation pattern.

When Examples Help Most​

✅ Use Examples When:​

  • You need a specific format that's hard to describe
  • You want to match an existing style
  • You're doing repetitive transformations
  • Describing what you want is complicated
  • You've tried describing it and AI got it wrong

⚠️ Skip Examples When:​

  • The task is simple and clear
  • You want AI to be creative (examples can limit creativity)
  • You don't have a good example to share
  • One-time tasks that don't need style matching

The Input-Output Pattern​

This is especially powerful for transformations:

Transform [type of input] into [type of output]:

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

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

---
Now transform this:
Input: [your actual input]

Use Cases:​

  • Converting formats (dates, addresses, names)
  • Rewriting text (formal→casual, long→short)
  • Extracting information
  • Standardizing data

🎯 Try It Now: Create an Example-Based Prompt​

Step 1: Choose a writing style you want to replicate

  • A favorite blog
  • Your own previous writing
  • A brand voice you admire

Step 2: Find or write 2-3 examples of that style

Step 3: Create a prompt:

Write [what you need] in this style:

Example 1: [paste example]
Example 2: [paste example]

Now write: [your specific request]

Step 4: Test it and see how closely AI matches your examples

5 Few-Shot Prompt Templates​

1. Style Matching​

Write in this style:

Example: "[Your example text]"

Now write [what you need] about [your topic].
Keep the same tone and length.

2. Format Replication​

Format information like this:

Example input: [sample input]
Example output: [sample output]

Now format this: [your input]

3. Response Templates​

Respond to messages like this example:

Their message: [example incoming message]
My response: [example response]

Now respond to: [actual message you received]

4. Content Generation​

Create [content type] similar to these:

Example 1: [sample]
Example 2: [sample]

Key elements to keep: [what to preserve]
Create one for: [your specific need]

5. Translation/Adaptation​

Adapt content like this:

Original: [original version]
Adapted: [adapted version]

Key changes made: [transformation you want]
Now adapt this: [content to adapt]

Quality of Examples Matters​

Good Examples:​

  • Representative of what you want
  • Clear and unambiguous
  • Consistent with each other
  • High quality (if you want high-quality output)

Bad Examples:​

  • Too varied (confuses AI about the pattern)
  • Low quality (AI will match that quality)
  • Contradictory styles
  • Too long (key patterns get lost)

How Many Examples?​

NumberWhen to Use
1 exampleSimple style matching
2-3 examplesBest balance—shows pattern without overwhelming
4+ examplesComplex patterns or when 2-3 weren't enough

More isn't always better. 2-3 good examples usually outperform 5-6 mediocre ones.

Quick Summary​

  • Few-shot learning = showing examples, not just describing
  • Examples help AI understand format, tone, style, and quality
  • The input-output pattern is great for transformations
  • Use 2-3 high-quality, consistent examples
  • Examples work best when describing what you want is difficult

Module 2 Complete!​

You now know the three core techniques:

  • ✅ Be Clear and Specific
  • ✅ Provide Context
  • ✅ Use Examples (Few-Shot)

What's Next?​

Test your knowledge with the Module 2 Quiz, then move to Module 3: Practical Applications where you'll apply these skills to real tasks.