Models are pattern-completers, so the shortest way to specify a pattern is to show it. Three examples of 'this review → that reply' teach tone, length and format more reliably than a paragraph of adjectives about them. Zero-shot asks with words; few-shot shows with cases.
The tradeoffs are practical: examples eat context tokens on every call, and a biased example set quietly biases every answer. In production systems the examples are chosen and versioned like code, not pasted in once and forgotten.
Related terms
Prompt engineering
The practice of writing model inputs so the output is reliably what you need — structure, examples, constraints and tone rather than hope.
System prompt
The standing instructions sent ahead of every conversation, setting what the model is supposed to do and how it should behave.
Context window
The maximum amount of text a model can take into account at once, counting the instructions, the conversation and the answer together.
The bench this belongs to
AI automationThe repetitive half of your week, handed to software that does not get bored. Inbox triage, follow-ups, reporting, data entry between tools that were never meant to talk.
