Module 3: Advanced Prompting Techniques

Once you have mastered the 4 Pillars, you are ready for two of the most powerful techniques in prompt engineering: few-shot prompting and chain-of-thought reasoning. These techniques dramatically improve output quality for complex tasks.

Lesson 3.1 — Few-Shot Prompting

Few-shot prompting means giving the AI one or more examples of the output you want before asking it to produce its own. You are essentially showing it the pattern, then asking it to follow that pattern.

Why It Works

LLMs are pattern-matching machines. When you provide examples, you are not just describing what you want — you are demonstrating it. This is far more effective than description alone for tasks involving specific formats, styles, or structures.

Structure of a Few-Shot Prompt

1. State the task
2. Provide 1–3 examples (input → output pairs)
3. Provide your actual input and ask for the output

Example: Classifying Customer Feedback

Task: Classify each customer comment as Positive, Negative, or Neutral.

Example 1: “The delivery was fast and the product works perfectly.” → Positive

Example 2: “I waited 3 weeks and the item arrived damaged.” → Negative

Example 3: “The package arrived on the expected date.” → Neutral

Now classify: “The product is okay but the instructions were confusing.”

By providing three examples, you have shown the AI exactly how to apply the classification — including the nuance that “okay but” is Neutral, not Positive.

When to Use Few-Shot Prompting

• Classifying or categorising data
• Reformatting text into a specific structure
• Writing in a specific style or voice
• Extracting specific fields from unstructured text

Lesson 3.2 — Chain-of-Thought (CoT) Prompting

Chain-of-thought prompting asks the AI to show its reasoning step by step before giving a final answer. This dramatically improves accuracy for tasks that involve logic, analysis, or multi-step decisions.

Why It Works

When an AI is forced to reason through a problem step by step, it is less likely to jump to an incorrect conclusion. The intermediate steps act as a self-check mechanism.

How to Trigger Chain-of-Thought

Simply add one of these phrases to your prompt:

• “Think step by step.”
• “Walk me through your reasoning before giving a final answer.”
• “Explain your thinking, then provide the conclusion.”

Example: Analysing a Business Decision

Without CoT: “Should we expand to the Trinidad market this year?”

With CoT: “We are a Barbados-based SaaS company with 200 customers. We are considering expanding to Trinidad this year. Think step by step through the key factors — market size, competition, operational costs, and revenue potential — before giving a recommendation.”

The CoT version forces the AI to consider each factor systematically before concluding, producing a far more useful and defensible analysis.

Combining Few-Shot and CoT

For complex analytical tasks, you can combine both techniques: provide an example of the reasoning process you want (few-shot), then ask the AI to apply the same reasoning to your specific problem (CoT). This is one of the most powerful combinations in prompt engineering.