One of the most powerful prompt techniques costs you just one sentence: give the AI a role. Tests show that role prompts improve response quality by up to 40% — because they focus the model on a specific area of expertise.
📖 Definition: With role prompting, you assign the AI a specific expertise and perspective. The model then activates the most relevant knowledge from its training and adjusts tonality, detail depth, and vocabulary accordingly.
Without role:
"How can I improve my newsletter?" → Generic tips found everywhere
With role:
"You are an email marketing expert with 15 years of B2B experience who has optimized newsletters for Fortune 500 companies. Analyze my newsletter and provide 5 concrete, prioritized improvement suggestions." → Specific, actionable recommendations at expert level
A good role definition has three components:
| Component | What It Does | Example |
|---|---|---|
| Title/Position | Professional classification | "Senior Financial Controller" |
| Experience | Depth of knowledge | "with 10 years of experience in the DACH region" |
| Specialization | Focus on relevant knowledge | "specialized in SaaS companies with 20-200 employees" |
💡 Tip: The more specific the role, the better the result. "Marketing expert" is good, "performance marketing specialist for B2B SaaS in the enterprise segment" is better.
Here are five immediately usable templates:
🔍 ANALYST
You are a Senior Business Analyst experienced in [INDUSTRY].
You analyze data methodically, identify patterns, and
formulate data-driven recommendations. Communicate clearly
and only use technical terms with explanations.
✏️ EDITOR
You are an experienced editor and copywriter focused on
business communication. You improve texts for clarity,
conciseness, and impact. You maintain the brand voice
and correct grammar and style.
💼 CONSULTANT
You are a management consultant specializing in [AREA].
You think strategically, weigh pros and cons, and provide
prioritized action recommendations. You ask questions
when important information is missing.
🎯 COACH
You are a leadership coach with 10+ years of experience.
You give constructive, concrete feedback. You ask
reflection questions and help with perspective shifts.
Tone: supportive but honest.
⚖️ CRITIC
You are a Devil's Advocate. Your task is to find
weaknesses, blind spots, and risks in proposals.
You are constructively critical and support your
objections with concrete arguments.
📖 Definition: A system prompt is a special instruction that defines the AI's base behavior for an entire conversation. Unlike a regular prompt, it applies to all subsequent messages.
System prompts are primarily used in API integrations and custom chatbots:
System prompt for an internal finance assistant:
You are a finance assistant for [COMPANY NAME].
Rules:
- Always respond in English
- Use technical terms only with brief explanations
- Explicitly flag uncertainty
- NEVER answer legal or tax questions
→ Refer to the legal department instead
- Base answers only on provided data
- Maximum response length: 300 words
⚠️ Caution: System prompts are powerful but not infallible. For applications with user inputs, you should additionally implement input validation to prevent prompt injection attacks.
For complex tasks, you can use multiple roles sequentially:
🏢 Real-world example: A sales team uses three standard roles: the "Sales Analyst" evaluates pipeline data, the "Deal Coach" gives advice on critical deals, and the "Email Pro" drafts follow-up messages. Result: 25% higher response rate on cold outreach.
🎯 Exercise: Create three role templates for your most common tasks. Test each role with a real task and compare results with a prompt without a role.
Next lesson: Chain-of-Thought and Few-Shot — advanced techniques for more thoughtful AI responses.