Few-Shot Prompting: Guiding AI Responses with Examples

From Telling to Showing: Zero, One, and Few-Shot Prompting

So far, the prompts you've constructed have relied entirely on instructions. You've told the AI what persona to adopt, what task to perform, and what format to use. In prompt engineering, this is known as zero-shot prompting—you give zero examples. While effective for many tasks, it often struggles when you need a very specific style or structure that is difficult to describe with words alone.

The solution is to provide examples. This is where we introduce two new concepts:

• One-shot prompting: You provide a single example of the input and desired output.
• Few-shot prompting: You provide multiple examples (typically 2-5) to give the AI a clearer pattern to follow.

The following video provides a concise overview of these concepts and explains why examples are so critical for guiding an AI model.

As the video explains, moving from zero-shot to few-shot is the standard way to improve performance when instructions alone aren't enough. For your work in HR, this is invaluable. Imagine trying to describe your company's precise tone for performance review feedback—it's far easier and more effective to show the AI three examples of well-written feedback than to try and describe that tone abstractly.

The Power of Examples in Action

Few-shot prompting works by giving the model a clear pattern to replicate. Instead of guessing your stylistic preferences, it learns them directly from the examples you provide. This significantly increases the chances that the output will match your expectations in terms of tone, structure, length, and format.

This diagram illustrates the concept: the prompt includes not just your new input, but also a few examples of previous inputs and their correct outputs (labels). The model uses these examples as a guide to process your new request.

The difference can be striking. The image below contrasts a response from a simple, zero-shot prompt with one that used few-shot learning.

For a deeper look into designing effective examples, the following reading from Lyfe AI is excellent. It explains how to move from simple instructions to providing one or more examples to anchor the AI's output.

How to Structure a Few-Shot Prompt for HR

Now, let's integrate this technique into the prompt structures you've already learned. You can think of it as adding an "E" for "Examples" to the frameworks we've discussed. A common structure is Role-Task-Context-Examples-Output.

The key is to clearly label your examples. A simple INPUT: / OUTPUT: or EXAMPLE 1: / EXAMPLE 2: format works very well.

The "AI Prompting Mastery Guide for HR Teams" provides a fantastic, practical template for this.

Let's look at that structure from the guide:

I need behavioral interview questions for a Marketing Manager role. Here are examples of the style I want:

Example 1: (Leadership) → 'Tell me about a time you led a cross-functional team to launch a successful marketing campaign. What was your approach, and what were the results?'
Example 2: (Data-Driven Decision Making) → 'Describe a situation where you used data to make a significant change to a marketing strategy. What was the data, and what was the impact?'

Now generate three new questions for: Creativity, Budget Management, and Stakeholder Communication.

This prompt is effective because it doesn't just ask for "behavioral questions." It shows the AI that a good question probes for the approach, the results, and the impact, using the "Tell me about a time..." or "Describe a situation..." format.

Your Turn: Practice with a Common HR Task

Let's apply this to a sensitive and common HR task: writing candidate rejection emails. You want the emails to be empathetic, professional, and brief, preserving the company's employer brand. A generic AI response is often too cold or verbose. This is a perfect use case for few-shot prompting.

Scenario:
You need to write a rejection email to a candidate who interviewed for a role but was not selected.

Your Task:
Use the provided examples to construct a complete few-shot prompt. Your goal is to generate a new, personalized rejection email for a candidate named Priya Sharma, who interviewed for the HR Coordinator position.

Here are your examples of well-written rejection emails:

• Example 1 Input:

  • Candidate Name: David Chen
  • Role: Marketing Analyst

• Example 1 Output:

  • Subject: Update on your Marketing Analyst application
  • Hi David,
  • Thank you again for taking the time to interview with our team. We sincerely appreciate your interest in our company.
  • While we were very impressed with your skills and experience, we have decided to move forward with another candidate for this role.
  • We wish you the very best in your job search and will keep your information on file for future opportunities.
  • Best regards,
  • The Talent Team

• Example 2 Input:

  • Candidate Name: Sarah Jones
  • Role: Junior Accountant

• Example 2 Output:

  • Subject: Your application for the Junior Accountant position
  • Hi Sarah,
  • Thank you for interviewing with us and for your interest in the Junior Accountant role. It was a pleasure learning more about your background.
  • The team had a difficult decision to make. Although your profile is strong, we have chosen to proceed with a candidate whose experience was a closer match for this specific position's requirements.
  • We encourage you to keep an eye on our careers page and wish you all the best.
  • Sincerely,
  • The Talent Team

Now, in your text editor, construct a prompt using the few-shot structure. Your prompt should include:

1. A Persona/Role.
2. Your two Examples (clearly labeled).
3. The Task and Context for the new email you want to generate for Priya Sharma.
4. A specification for the Output.

Take about 10-15 minutes to build your prompt.

Example Solution:

Here is a well-structured few-shot prompt for this task. Your prompt may be slightly different, but it should follow the same principles.

Persona: Act as a empathetic and professional HR recruitment specialist. Your goal is to communicate a hiring decision clearly while maintaining a positive relationship with the candidate.

Examples:
Here are two examples of the style and tone I want you to use for rejection emails.

INPUT:
Candidate Name: David Chen
Role: Marketing Analyst
OUTPUT:
Subject: Update on your Marketing Analyst application
Hi David,
Thank you again for taking the time to interview with our team. We sincerely appreciate your interest in our company. While we were very impressed with your skills and experience, we have decided to move forward with another candidate for this role. We wish you the very best in your job search and will keep your information on file for future opportunities.
Best regards,
The Talent Team

INPUT:
Candidate Name: Sarah Jones
Role: Junior Accountant
OUTPUT:
Subject: Your application for the Junior Accountant position
Hi Sarah,
Thank you for interviewing with us and for your interest in the Junior Accountant role. It was a pleasure learning more about your background. The team had a difficult decision to make. Although your profile is strong, we have chosen to proceed with a candidate whose experience was a closer match for this specific position's requirements. We encourage you to keep an eye on our careers page and wish you all the best.
Sincerely,
The Talent Team

---

Task: Now, following the style and tone of the examples, write a new rejection email.

Context:
INPUT:
Candidate Name: Priya Sharma
Role: HR Coordinator

Output: Return only the new email, including the subject line.

This prompt effectively teaches the AI the desired length, tone (empathetic but direct), and structure (gratitude, decision, well wishes) before asking it to perform the new task.

Conclusion

You have now added a powerful technique to your prompt engineering toolkit. By providing examples, you move from simply giving instructions to actively training the AI on the specific style you need for a given task.

Key Takeaways:

• Show, Don't Just Tell: Few-shot prompting, which involves providing 2-5 examples, is more effective than instructions alone for tasks requiring a specific style, tone, or format.
• Structure is Key: Organize your examples clearly with labels like INPUT: and OUTPUT: to create a pattern the AI can easily follow.
• Integrate with What You Know: Add an "Examples" section to the prompt structures you've already mastered (e.g., Role-Task-Context-Examples-Output).
• HR Application: This technique is invaluable for maintaining consistency in HR communications, from candidate emails and performance feedback to internal announcements.

In this lesson, you learned to guide the AI's style. But what about tasks that are not just stylistically specific, but also logically complex? In our next lesson, we will address this by learning how to break down a large, multi-step HR project into a sequence of smaller, manageable prompts. This will allow you to leverage AI for more complex analytical and strategic work.