1. What This Objective Really Means
Using Azure AI Foundry, you should know how to:
Build generative AI applications (e.g. chatbots)
Create agents that can take actions
Use models, prompts, and data together
Deploy solutions to users
🤖 2. Generative AI Apps vs Agents
🧾 Generative AI Apps
Focus on creating content
Examples:
Chatbots answering questions
Report or email generation
Document summarisation
👉 Behaviour:
Input → Prompt
Output → Generated response
🧠AI Agents
Go beyond generation → take actions
Capabilities:
Multi-step reasoning
Call external tools/APIs
Automate workflows
Examples:
Read email → extract request → create ticket
Query data → generate report → send email
👉 Behaviour:
Think → Decide → Act
🧩 3. Key Components You Must Recognise
🤖 Models
Typically Large Language Models (LLMs)
Provide reasoning and generation capability
✍️ Prompts
Define behaviour and tasks
Include:
System prompt (rules, tone)
User prompt (request)
đź”— Grounding (Your Data)
Connect AI to real business data
Prevents hallucinations
👉 Example:
Chatbot using company policies
🧰 Tools (for Agents)
Agents can interact with:
APIs
Databases
External systems
👉 This enables action-taking
🔄 Orchestration
Chain steps together:
Understand request
Retrieve data
Generate response
Perform action
🚀 Deployment
Publish as:
Chatbot
Web app
API
🔄 4. Typical Flow (Exam-Friendly)
Generative AI App:
User asks question
Prompt sent to model
Model generates response
Response returned
Agent:
User request
Model interprets intent
Calls tool / retrieves data
Performs action
Returns result
🎯 5. Example Scenarios (Very Important)
Scenario 1
“Build a chatbot that answers HR questions using company documents”
Generative AI app
Uses grounding
No actions required
Scenario 2
“An AI assistant that books meetings and sends emails”
Agent
Requires tools + orchestration
Scenario 3
“Generate a summary of a long report”
Generative AI app
⚠️ 6. Common Exam Traps
❌ Confusing apps with agents
✔️ Agents take actions, apps just generate content
❌ Ignoring grounding
✔️ Critical for enterprise use
❌ Thinking agents are fully autonomous
✔️ They operate within defined tools and constraints
🧠7. Simple Memory Trick
“Generate vs Act”
Generate → Generative AI app
Act → Agent
âś… Summary
To implement generative AI apps and agents using Foundry:
Use Azure AI Foundry
Build:
Apps → generate content
Agents → take actions
Combine:
Models
Prompts
Data (grounding)
Tools (for agents)
Deploy to users