Also known as: prompt design
In plain English
Prompt engineering is learning how to ask AI well. Giving it a role, clear steps, examples and the format you want can turn a vague answer into a useful one.
In practice
Good techniques include explaining the audience and purpose, showing examples of good output, asking for structured formats like tables or JSON, and breaking complex tasks into steps. As models improve, clarity matters more than clever tricks.
Under the hood
Techniques include few-shot examples, step-by-step reasoning, structured output schemas, role specification and prompt chaining. Rigorous practice uses evaluation sets to measure whether a prompt change improves results across many cases, not just one.
Example
"With better prompt engineering, the summaries finally matched our template."