Once you've mastered the basics of prompting, you can significantly improve the performance and precision of your AI interactions with advanced techniques.
System prompts define the fundamental personality and behavior of the AI assistant:
System: You are an experienced data analyst with expertise in Python and
statistical analysis. You explain complex concepts clearly and
approach problems systematically. When uncertain, you ask for clarification.
Structure complex prompts hierarchically:
Main Task: Analyze the annual report
Context:
- Industry: E-commerce
- Period: 2023
- Focus: Revenue growth and customer acquisition
Subtasks:
1. Identify key KPIs
2. Compare with previous year
3. Recognize trends and patterns
4. Formulate recommendations
Format: Structured report with visualization suggestions
Have the AI improve its own thinking process:
Before answering, consider:
1. What is the real question behind the request?
2. What assumptions am I making?
3. What information am I missing?
4. How can I structure my response?
Now answer the original question.
Step 1: Create an initial draft of the solution
Step 2: Identify weaknesses in this draft
Step 3: Improve the draft based on the weaknesses
Step 4: Present the final, optimized solution
Consider this problem from three perspectives:
1. As CEO: Focus on strategic impact
2. As CFO: Focus on financial aspects
3. As CTO: Focus on technical feasibility
Summarize the insights and find consensus.
Imagine you're moderating a discussion between:
- An economist
- A privacy expert
- A UX designer
Topic: Implementing AI in customer service
Present the different viewpoints and a synthesis.
Analyze the text and output results in the following JSON format:
{
"main_topic": "...",
"sentiment": "positive/neutral/negative",
"keywords": ["...", "..."],
"recommendations": [
{"priority": "high/medium/low", "action": "..."}
]
}
Compare the three solution approaches in a table:
| Criterion | Solution A | Solution B | Solution C |
|------------|------------|------------|------------|
| Cost | ... | ... | ... |
| Time | ... | ... | ... |
| Complexity | ... | ... | ... |
| Risk | ... | ... | ... |
Write a product description with the following constraints:
- EXACTLY 150 words (±5 words)
- At least 3 benefit arguments
- No superlatives ("best", "greatest", etc.)
- Call-to-action at the end
Explain quantum computing, but:
- Use only everyday analogies
- No technical terms without explanation
- Maximum 3 sentences per paragraph
- Build a continuous metaphor
TEXT: [Insert]
ANALYSIS FRAMEWORK:
1. Summary (2-3 sentences)
2. Identify main arguments
3. Evaluate strengths and weaknesses
4. Uncover implicit assumptions
5. Conclusions and implications
OUTPUT: Structured analysis according to above framework
PROBLEM: [Description]
CONTEXT:
- Affected stakeholders: ...
- Timeframe: ...
- Budget: ...
- Constraints: ...
SOLUTION APPROACH:
1. Problem analysis (identify causes)
2. Solution options (min. 3 alternatives)
3. Evaluation (pros/cons of each option)
4. Recommendation with justification
5. Implementation plan
If you encounter the following problems:
- Unclear requirements: Ask clarifying questions
- Contradictory information: Point it out
- Missing data: Mark assumptions
- Ethical concerns: Explain them
Then proceed with the best possible answer.
After your response:
1. Verify factual accuracy
2. Ensure all requirements are met
3. Identify possible misunderstandings
4. Offer clarifications if needed
Example 1:
Question: Revenue 2022: €1M, 2023: €1.2M. Growth?
Thinking: (1.2M - 1M) / 1M = 0.2 = 20%
Answer: 20% growth
Example 2:
Question: Costs decreased from 500k to 450k. Savings?
Thinking: (500k - 450k) / 500k = 0.1 = 10%
Answer: 10% cost savings
Now your task:
Question: Customer count increased from 1000 to 1350. Growth?
As an experienced project manager, create a project plan:
FORMAT:
# Project Name
## Goals (SMART criteria)
## Milestones
- [ ] Milestone 1 (Date)
- [ ] Milestone 2 (Date)
## Resources
## Risks and Countermeasures
## Success Criteria
- Clarity over complexity - Use complex techniques only when necessary
- Test and iterate - Continuously improve prompts
- Documentation - Save and share successful prompts
- Combination - Combine different techniques situationally
- Feedback loops - Analyze results and learn
Once you've mastered these advanced techniques, you can move on to expert techniques or explore specific use cases.