About this module
The focus here is practical rather than theoretical. Learners start with the difference between zero-shot and few-shot prompting, then practice deciding when to ask directly and when to provide examples of the desired result. The risk is giving examples that accidentally teach the AI the wrong pattern, so the module keeps review and judgment close to the work. By the end, learners can use examples when consistency matters and know what to check before moving an AI-assisted result forward.
Key takeaways
Every prompt makes a choice. Ask cold, or show examples first. That one decision changes the quality of what you get back. Ask a model for a summary with zero setup, and you get its default style.
Show it two examples of the summary you actually want first, and the output changes completely. Zero-shot prompting is the fastest option. You state the task plainly, and the model draws entirely on its own training to decide how to answer, with no examples to guide it.
This is zero-shot. One instruction, nothing else. The model has to guess your tone, length, and structure entirely on its own, and it usually guesses generic. Few-shot prompting adds examples before the ask.
Show the model two or three input and output pairs in the style you want, and it mirrors that pattern on the new input. Same product, same model, but now it has seen the pattern twice. The output matches that punchy, benefit-first style instead of a generic paragraph. Zero-shot is fast but unpredictable, fine for simple, well-known tasks.
Few-shot takes a little more setup, but it locks in tone and format when consistency actually matters. Reach for zero-shot when the task is common, when you just need a fast first draft, or when there's no strict format to match yet.
Reach for few-shot when tone matters, when you're producing many similar items, or when zero-shot keeps coming back too generic to use. Two to three examples is usually the sweet spot. Fewer, and the pattern isn't clear. More, and you're burning tokens the model doesn't need.
One common mistake: giving examples that don't match each other in format or tone. Inconsistent examples confuse the model more than giving no examples at all. Zero-shot for speed and simple tasks, few-shot for consistency and specific tone, two or three well-matched examples, and always keep your format the same throughout.
Next time you prompt, try it cold first, then add two matching examples and ask again. Notice how much more precise the second answer becomes.
