Prompt of the Day: Meta-Prompting -- Let AI Write and Execute Your Optimal Prompt
You type into ChatGPT or Claude: 'Write me a project description.' You get something generic back. You try again, add details, rephrase. After the third attempt, the result is okay -- but not good. You know: a better prompt would have delivered a better result. But how do you write the perfect prompt?
The problem: The quality of your AI results depends directly on how good your prompt is. But good prompt engineering is a skill that takes time to develop. You want results, not a degree in prompt formulation.
The solution: Meta-prompting. Instead of crafting the perfect prompt yourself, you ask AI to create the optimal prompt for your task -- and then execute it. AI knows better than you what structure, context, and instructions it needs to deliver the best result.
Why does this work so well?
- AI knows its own strengths and weaknesses
- It automatically adds important details you would have forgotten (audience, tone, structure, length)
- It structures the task into logical steps
- Studies show 10-25% better results compared to direct requests
What you can use this for:
- Writing: Blog articles, emails, presentations, reports, social media posts
- Analysis: Market analyses, competitor comparisons, SWOT analyses, data interpretation
- Creative: Brainstorming, concept development, naming, storylines
- Planning: Project plans, strategy papers, workshop concepts, agendas
- Communication: Difficult emails, feedback conversations, negotiation strategies
- Learning: Explanations of complex topics, summaries, study plans
How it works:
1. Describe your task in 1-2 sentences -- just as you would tell a colleague
2. Add optional context (target audience, industry, previous attempts)
3. AI creates the optimal prompt, explains why it is better, and delivers the result directly
Pro tips:
- Use iteratively: If the first result is not perfect, say: 'Revise your prompt based on this feedback: [your feedback]. Execute the improved prompt again.'
- Build a prompt library: Save the generated prompts that worked well. You build a personal prompt collection without needing to become a prompt expert yourself.
- For recurring tasks: 'Create a reusable prompt template for [task type] that I can use every week. With placeholders for the variable parts.'
- Comparison: 'Create 2 different prompts for this task -- a short one and a detailed one. Execute both and explain which fits better and why.'
- Learning effect: Read the generated prompt carefully. You will learn which elements make good prompts -- role assignment, context, structure, format requirements, constraints.
You are an expert in prompt engineering. Your task has two parts: **Part 1: Create the optimal prompt** My task: [Describe your task in 1-2 sentences, e.g., 'I need to write a convincing project description for a new sustainability project that will excite our board' or 'I need an analysis of the pros and cons of remote work for my team meeting'] Additional context (optional): - Target audience: [e.g., 'Management', 'Customers', 'Colleagues', 'General public'] - Industry/Field: [e.g., 'IT', 'Marketing', 'Education', 'Healthcare'] - Desired format: [e.g., 'Email', 'Presentation', 'Report', 'Bullet points'] - Tone: [e.g., 'Professional', 'Casual', 'Persuasive', 'Factual'] - Previous attempts: [e.g., 'I already tried asking directly but the result was too generic'] Create the best possible prompt to perfectly solve this task. The prompt should: - Include a clear role assignment - Provide the necessary context - Break the task into logical steps - Define quality criteria (length, style, depth) - Set concrete constraints (what to avoid) Show the prompt in a marked block and explain in 3-4 sentences why this prompt works better than a simple direct request. **Part 2: Execute the prompt** Now execute the created prompt yourself and deliver the complete result. **Part 3: Reflection** Finally, answer: - What makes this prompt better than a direct request? - Which 2 elements of the prompt have the greatest impact on quality? - How could the prompt be improved further?