The basics
A prompt is the text you enter into an AI system. It is your instruction, your assignment, your question. Everything you type before you press Enter.
The term comes from the English "to prompt" - to urge, to incite to action. And that is exactly what you do: you prompt the AI model to produce something.
Why the prompt is decisive
AI models are tools. Just like a search engine: the quality of your search term determines the quality of the results. Type "marketing" into Google and you get millions of vague results. Type "B2B content marketing strategy SaaS 2026" and you get exactly what you need.
Prompts work the same way. "Write a text" produces something generic. "Write an introduction of 80 words for our new recruitment page. Target audience: senior developers. Tone: informal, direct, no corporate speak" produces something usable.
The anatomy of a good prompt
An effective prompt contains four elements:
1. Role or context Tell the AI from which perspective it should work. "You are an experienced copywriter specializing in tech companies." This steers the style and knowledge level.
2. Task What needs to happen? Be concrete. Not "help me with my presentation" but "create an outline of 10 slides for a pitch to investors about our AI tool for recruiters."
3. Constraints Set boundaries. Length, tone, format, what you specifically do not want. "Maximum 200 words, no jargon, no bullet points."
4. Examples (optional but powerful) Show what you expect. Give an example of good output. The AI recognizes the pattern and applies it.
Types of prompts
Direct prompts A simple question or assignment. "What is the capital of Japan?" or "Translate this text to English." Works well for simple tasks.
Instruction prompts A detailed assignment with context and constraints. This is the most commonly used form for professional work.
Few-shot prompts You give 2-3 examples and ask for more in the same style. Example: "Here are three product descriptions I like: [examples]. Write two more in the same style for these products: [products]."
Chain-of-thought prompts You ask the AI to reason step by step. "Analyze this problem step by step before giving a solution." This produces better results for complex questions.
System prompts A background prompt that permanently steers the AI's behavior. For AI agents, this is the personality configuration - the tone of voice, rules, and context that the agent always carries with it.
Common mistakes
- Too vague - "Write something fun" gives the AI no direction
- Too much at once - Break large tasks into smaller steps
- No context - The AI knows nothing about you unless you tell it
- Not iterating - Your first prompt is a starting point, not the end result
- Copy behavior - A prompt that works for someone else does not necessarily work for you. Adapt to your own context
Prompts for an AI agent
With your own AI agent, you do not write the prompt from scratch every time. The agent has a system prompt that is always active. It defines who the agent is, how it communicates, and what it may and may not do.
Your daily interaction then becomes much shorter: "Summarize yesterday's meeting" instead of explaining who you are and what you expect every time.
Getting started
1. Pick a task you already have to do this week 2. Write a prompt with role, task, and constraints 3. Look at the result and refine 4. Save your best prompts for reuse
