Use Examples in your Prompt

Sometimes the fastest way to explain what you want is to stop describing it and just show it. Including examples in your prompt — a sample input paired with the output you’d want back — communicates format, tone, level of detail, and edge-case handling far more precisely than paragraphs of instructions. It’s the prompting equivalent of handing a contractor a photo instead of describing a kitchen.

For developers, examples are the cheapest reliability upgrade available. Instructions get interpreted; examples get imitated. When you need output that follows a house style, a quirky data format, or a subtle judgment call (“this counts as spam, that doesn’t”), a description alone leaves room for drift, while two or three well-chosen demonstrations lock the behavior in. Examples are also self-documenting: anyone reading your prompt later sees exactly what “good” looks like.

In practice, you’ll embed input/output pairs directly in the prompt, clearly delimited: Input: "pymnt failed 2x, plz help" → Output: {"category": "billing", "sentiment": "frustrated"}. Pick examples that cover the tricky cases, not just the easy ones — include one where the right answer is “reject” or “unknown” — and keep them consistent, because the model will copy your formatting down to the punctuation. This habit is the intuition behind the formal few-shot prompting technique you’ll meet later in this module; here, the point is simpler: when words are ambiguous, demonstrate.