Lesson 2 of 6·7 min read

Role Prompting and System Prompts 🎭

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.


🎯 What You'll Learn

  • Why role prompting is so effective
  • How to define impactful roles
  • What system prompts are and when you need them
  • Ready-made role templates for immediate use

The Power of Roles 💪

📖 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


Defining Effective Roles 🎯

A good role definition has three components:

ComponentWhat It DoesExample
Title/PositionProfessional classification"Senior Financial Controller"
ExperienceDepth of knowledge"with 10 years of experience in the DACH region"
SpecializationFocus 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.


Ready-Made Role Templates 📋

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.

System Prompts: The Foundation 🏗️

📖 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.


Combining Roles 🔗

For complex tasks, you can use multiple roles sequentially:

  1. Analyst → Analyze data, identify patterns
  2. Consultant → Derive strategic recommendations
  3. Critic → Check recommendations for weaknesses
  4. Writer → Format results for presentation

🏢 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.


📋 Summary

  • Role prompting focuses the AI on relevant expert knowledge
  • A good role has title + experience + specialization
  • System prompts define base behavior for conversations and API integrations
  • Create 5-10 standard roles for your team's most common use cases
  • Combine roles sequentially for complex tasks

🎯 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.