Skip to content
← All terms
Working with language modelsLast reviewed

Prompt Engineering

Prompt engineering means designing inputs to an AI so that the results are reliably usable: with a clear task, context, examples and the desired format.

How it works

A model makes the best of your input, but it cannot read minds. Good prompts are therefore specific: they state the task, give context, describe the desired format and show examples where needed. Typical building blocks are a role (“You are …”), the task, the background, examples and rules for the output.

A practical example

Instead of “Write something about our new product”, this works better: “Write a newsletter text of at most 120 words for business customers, friendly and factual, with a clear call to action. Product: …”.

What you should know

  • Prompts are work products: if you collect, version and test them, you get consistent quality.
  • A good prompt does not replace checking. You should always review important outputs.
  • For recurring tasks, a fixed system prompt is worth it instead of ever-new inputs.

Matching tools

Questions about your project?

Get in touch →