What is prompt engineering?

What is a prompt?

A prompt is the instruction or question you give to an AI model. It is the input on the basis of which the model generates an answer. A simple prompt produces a simple answer; a well-considered prompt produces an accurate, usable result.

Prompt engineering is the field that deals with systematically designing those instructions to deploy AI models as effectively as possible.

Why does the wording matter so much?

Illustration created with Canva AI

Large language models generate text based on probability calculations. They try to produce the most probable continuation of text given the context. That context is you — your prompt determines the direction in which the model will think.

A vague question leads to a vague answer. A specific, structured prompt produces a specific, structured answer.

Important prompt techniques

• Assigning a role — "You are an experienced marketer" puts the model in the right 'mindset'
• Giving context — Describe the situation, the goal, and the target audience
• Providing examples (few-shot) — Show what you mean with 1 or 2 concrete examples
• Step-by-step thinking (chain of thought) — Ask the model to show its reasoning: "Think step by step"
• Specifying output format — "Give your answer as a numbered list in a maximum of 200 words"

Zero-shot, one-shot, and few-shot

These terms describe how many examples you include in the prompt:

• Zero-shot: no example, only the instruction
• One-shot: one example for illustration
• Few-shot: two or more examples to demonstrate the desired pattern

Few-shot prompting works particularly well for structured tasks such as classifying text or converting data to a certain format.

Common mistakes

• Too vague: "Write something about AI" gives a generic answer
• Asking too much at once in one prompt
• Not providing context about who you are or what you want to achieve
• Not correcting the model when the answer deviates

Prompt engineering as a skill

Now that AI tools are widely available, prompt engineering is becoming a basic skill — comparable to knowing how to search on Google. Those who know how to direct AI models get considerably more value from them than those who do not.