2.4 Module 2 · Prompt Engineering

Few-Shot & Zero-Shot Strategies

The difference between vague AI output and precise, consistent results often comes down to one thing: examples. Learn when to provide them and when to go without.

Example Injector Tool Strategy Decision Tree

Understanding Shot Strategies

In prompt engineering, "shots" refer to examples you provide the AI. Zero-shot means no examples. One-shot means one example. Few-shot means two to five examples. The more examples you provide, the more tightly you constrain the AI's output format, tone, and reasoning pattern.

Zero-Shot

No examples. Relies entirely on the AI's training. Best for straightforward tasks where format doesn't matter much.

One-Shot

One example input/output pair. Sets the pattern. Good when the task is clear but you want a specific format.

Few-Shot

Two to five examples. Demonstrates edge cases and nuance. Essential for classification, evaluation, or complex formatting tasks.

Defence context: Security classification, document triage, and briefing note formats all benefit enormously from few-shot prompting. One well-chosen example can be worth a paragraph of instructions.

Example Injector Tool

Build a prompt step by step. Add examples and see how zero-shot, one-shot, and few-shot versions compare. Toggle examples on and off to see the difference.

Strategy Decision Tree

Not sure which strategy to use? Walk through this interactive decision tree. Click each node to explore your options.

?

Do you have example outputs you want to match?