AI-103 Study Guide

A recommended approach to preparing for the Azure AI Apps and Agents Developer Associate exam.

How to Use This Guide

Each domain section contains:

• Overview — scope and key exam emphasis
• Key Concepts — critical terms and distinctions you must know
• Azure Services / Foundry Features — what is in scope
• Code Patterns — Python SDK snippets you must understand and apply
• Study Resources — official labs and documentation

Domain Breakdown

Domain	Weight	Priority
1: Plan & Manage	25–30%	⭐⭐⭐ High — appears in every other domain
2: Generative AI & Agents	30–35%	⭐⭐⭐ Highest — core of the exam
3: Computer Vision	10–15%	⭐⭐ Medium
4: Text Analysis	10–15%	⭐⭐ Medium
5: Information Extraction	10–15%	⭐⭐ Medium

Recommended Study Order

1. Domain 1 — Get Foundry set up, understand service selection, responsible AI, security
2. Domain 2 — Build a RAG app, deploy an agent, implement multi-agent orchestration
3. Domain 5 — Content Understanding and retrieval pipelines (connects to Domain 2's RAG work)
4. Domain 3 — Image/video generation and multimodal understanding
5. Domain 4 — Text and speech analysis via Foundry Tools

Suggested Study Timeline (6–8 weeks)

Week	Focus
1	Domain 1: Foundry setup, service selection, security, responsible AI
2	Domain 2 (part 1): Deploy LLMs, implement RAG, Prompt Flow
3	Domain 2 (part 2): Build agents, multi-agent orchestration, safeguards
4	Domain 3: Image/video generation, multimodal models, responsible AI for vision
5	Domain 4: Text analysis, speech workflows, custom language models
6	Domain 5: Content Understanding, retrieval pipelines, document extraction
7–8	Review + practice scenarios + exam sandbox

Essential Hands-on Labs

Lab repo	Focus
mslearn-ai-studio	Foundry, generative AI, agents
mslearn-ai-services	Azure AI service fundamentals
mslearn-knowledge-mining	Azure AI Search and RAG

Core Python Patterns to Know

# 1. Connect to a Foundry project
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential

client = AIProjectClient(
    endpoint="https://<your-project>.services.ai.azure.com",
    credential=DefaultAzureCredential()
)

# 2. Chat with a deployed LLM
from azure.ai.inference import ChatCompletionsClient
from azure.ai.inference.models import SystemMessage, UserMessage
from azure.core.credentials import AzureKeyCredential

chat = ChatCompletionsClient(
    endpoint="https://<your-project>.services.ai.azure.com/models",
    credential=AzureKeyCredential("<key>")
)
response = chat.complete(
    model="gpt-4o",
    messages=[SystemMessage("You are a helpful assistant."), UserMessage("Hello!")]
)

# 3. Create and run an agent
agent = client.agents.create_agent(
    model="gpt-4o",
    name="my-agent",
    instructions="You are a helpful assistant that can search the web.",
    tools=[{"type": "bing_grounding"}]
)
thread = client.agents.create_thread()
client.agents.create_message(thread.id, role="user", content="What is new in Azure AI?")
run = client.agents.create_and_process_run(thread.id, agent.id)

Exam Day Tips

tip

• Service selection is everywhere — know when to use LLM prompting vs. a dedicated Foundry Tool (e.g., Azure Translator vs. LLM translation).
• RAG is heavily tested — understand the full pipeline: ingest → chunk → embed → index → retrieve → generate.
• Agents: know the difference between single-agent vs. multi-agent, and understand tools, memory, and conversation tracking.
• Responsible AI is in every domain — content filters, guardrails, prompt shields, trace logging, and approval workflows.
• Content Understanding covers documents, images, audio, and video — not just PDFs.