1. role assignment
2. few shot examples
3. cot prompting
4. constraint setting

1. Role Assignment

Defines the persona, expertise level, and perspective the AI should adopt. This shifts the model’s underlying context to generate more tailored responses.

• Format: “You are an expert [Role] specializing in [Topic].”
• Example: “You are an elite financial analyst who specializes in corporate risk assessment.”

2. Few-Shot Examples

Provides the model with 1 to 3 concrete input-output pairings. This shows the model exactly what the desired output looks like without needing to retrain it.

• Format: Give specific ‘Q:’ and ‘A:’ pairs so the model learns the expected structure and tone.

3. Chain-of-Thought (CoT) Prompting

Instructs the AI to break a problem down and show its work sequentially before delivering a final answer. This drastically reduces hallucinations and improves logic, especially in math or coding.

• Format: Require the model to “think step by step”. In few-shot examples, explicitly write out the reasoning steps.

4. Constraint Setting

Defines boundaries, formats, and rules that the AI must never violate. This helps filter out unnecessary text or prevents the AI from acting unsafely or inaccurately.

Format: “Do not include [X]. Always format the output as [Y].”