1. Where This Fits
Using Azure AI Foundry, you typically:
Choose a model (e.g. LLM)
Create system + user prompts
Test and refine responses
Deploy your solution
👉 Prompt design is one of the most important skills tested.
🧩 2. System Prompt vs User Prompt
⚙️ System Prompt (Controls Behaviour)
đź“Ś What it does
Defines how the AI should behave
🔑 Typical contents
Role (“You are a financial analyst…”)
Tone (“Respond formally…”)
Constraints (“Only use provided data…”)
Rules (“Do not make assumptions…”)
âś… Example
“You are a professional data architect. Provide precise, structured answers using bullet points.”
👉 Think:
“Set the personality and rules”
👤 User Prompt (The Task / Question)
đź“Ś What it does
Tells the AI what to do
🔑 Typical contents
Question or instruction
Context or input data
âś… Example
“Summarise the following report in 3 bullet points…”
👉 Think:
“What do I want right now?”
⚖️ 3. Key Differences (Exam Critical)
Feature System Prompt User Prompt
Purpose Define behaviour Request task
Scope Persistent Per request
Example “You are a tutor…” “Explain AI in simple terms”
🎯 4. Characteristics of Effective Prompts
âś… Be Clear and Specific
❌ “Explain this”
✅ “Explain this in 3 bullet points for a beginner”
âś… Provide Context
Include relevant background data
Reduces hallucinations
âś… Define Output Format
Bullet points
Table
Short summary
âś… Use Constraints
“Use only the provided data”
“Limit to 100 words”
âś… Use Examples (Few-shot Prompting)
Provide sample inputs/outputs
🧪 5. Prompt Engineering Techniques (AI-901 Level)
đź“Ś Zero-shot
No examples
“Summarise this text”
đź“Ś Few-shot
Provide examples
“Here are examples… now do the same”
đź“Ś Role-based prompting
“You are an expert…”
🚀 6. Example (Exam-Style Scenario)
Scenario:
“Create a chatbot that answers questions using internal documentation and responds formally”
Correct setup:
System prompt:
“You are a professional assistant. Answer using only the provided company data. Respond formally.”
User prompt:
“What is our refund policy?”
⚠️ 7. Common Exam Traps
Confusing system vs user prompt
Forgetting to specify output format
Prompts too vague → poor answers
Thinking prompts = model training ❌
🧠8. Simple Memory Trick
“System = Setup, User = سؤال (Question)”
System → sets behaviour
User → asks task
âś… Summary
To create effective prompts in AI-901:
Use system prompts to define behaviour, tone, and rules
Use user prompts to specify the task
Make prompts:
Clear
Specific
Context-aware
Structured