Prompt Engineering for beginners

The quality of the results you receive from AI tools like ChatGPT depends significantly on how you formulate your queries. This ability to write effective instructions for AI systems is called "[Prompt Engineering](/Performance/ki trainings)". In this guide you will learn the basics of this important competence and get practical tips to achieve better results immediately.

What is Prompt Engineering?

A "prompt" is the input or instruction you give to an AI. Prompt Engineering is the art and science to formulate these inputs so that the AI delivers exactly what you need. It is like learning a new language – the language of the AI.

The 5 golden rules for better prompts

1. Be specific and precise

Vage inquiries lead to vague answers. The more accurate you describe what you want, the better the result.

Bad

"Write me something about marketing"

"Write a 300-word blog post on social media marketing strategies for small craft businesses, with 3 concrete tips"

2. Specify context

Explain the AI the background of your request. Who is the target group? What is the output used for?

✓ Example with context: "I am a human resource manager in a medium-sized IT company. We are currently introducing ChatGPT as a tool for all employees. Write an email to the staff who explains the benefits and addresses possible concerns. The sound should be friendly and encouraging."

3. Define the format

Tell the AI how the answer should be structured: as a list, table, flow text, with headings etc.

✓ Example with format specification: "Create a comparison table of the 5 most popular project management tools. Columns: name, price, main functions, advantages, disadvantages, suitable for."

4. Make a role

Ask the AI to take a specific perspective or expertise. This often significantly improves the quality and relevance of the answers.

✓ Example with role assignment: "You are an experienced SEO expert with 15 years of experience. Analyze the following website structure and give me 5 concrete suggestions for better Google rankings."

5.

The first prompt rarely delivers the perfect result. Refine your inquiry based on the answer. Give feedback and ask for adjustments.

✓ Example for iteration: "This is good, but please shorten the text to half and use a more formal sound. Also add a call-to-action at the end."

The CRISP Framework for Perfect Prompts

Use this simple framework as checklist:

• Context: background information on the situation

• Role (roll): What expertise should the AI take?

• Instructions: What exactly should be done?

• Specifics (Details): format, length, style, target group

• Purpose: Where