Pular para o conteúdo principal

AI-103: Azure AI Apps and Agents Developer

Exam Details

Exam version: Skills measured as of April 2026 | Passing score: 700/1000 | Duration: ~120 minutes | Cost: ~$20-50 for labs

Replaces AI-103

This exam replaced AI-103 (retired June 30, 2026). It shifts focus from traditional Azure AI services to generative AI, agentic architectures, and Microsoft Foundry. If you previously studied AI-103, much of the vision/NLP/search knowledge still applies, but the center of gravity is now GenAI + Agents.

Who is this for?

The AI-103 validates the ability to develop, deploy, and manage AI apps and agents using Azure. This is a hands-on, code-first certification focused on modern AI development patterns:

  • AI developers building generative AI applications
  • Software engineers integrating LLMs and agents into products
  • Cloud architects designing agentic AI solutions
  • DevOps engineers deploying AI workloads on Azure
  • Data scientists transitioning to production AI engineering

Skills at a glance

DomainWeightChallenges
Plan and manage an Azure AI solution15–20%01–10
Implement generative AI and agentic solutions35–40%11–23
Implement computer vision solutions10–15%24–30
Implement text analysis solutions15–20%31–39
Implement information extraction solutions15–20%40–49
What's New vs AI-103

The heaviest domain (35-40%) is now GenAI + Agents — covering RAG pipelines, agent orchestration, multi-tool agents, prompt engineering, and Microsoft Foundry. Traditional NLP and Computer Vision remain but with reduced weight.

Domain breakdown

Domain 1: Plan and Manage an Azure AI Solution (15–20%)

TopicKey Skills
ArchitectureSecure, scalable Azure AI solution design
DeploymentPlanning, provisioning, and monitoring AI workloads
ComplianceResponsible AI, governance, and regulatory requirements
SecurityIdentity, access controls, and data protection
MonitoringPerformance tracking, cost management, alerting

Domain 2: Implement Generative AI and Agentic Solutions (35–40%)

TopicKey Skills
Azure OpenAI + FoundryDeploy and manage LLM models
RAG pipelinesRetrieval-Augmented Generation with Azure AI Search
Agent architectureDesign multi-step, goal-oriented agents
Agent orchestrationMulti-tool, task completion, and reasoning chains
Prompt engineeringSystem/user prompts, chaining, context window management
Agent monitoringTracking behavior, debugging, and optimization
Azure AI Agent ServiceBuilding and deploying production agents

Domain 3: Implement Computer Vision Solutions (10–15%)

TopicKey Skills
Azure AI VisionImage/video analysis, classification, detection
Custom VisionTraining custom models for specific use cases
OCRDocument text extraction
Face APIFacial recognition and analysis
Production integrationDeploying vision models in apps

Domain 4: Implement Text Analysis Solutions (15–20%)

TopicKey Skills
NLP tasksSentiment, key phrases, language detection
Entity extractionNamed entities, PII detection, summarization
Content moderationSafety and compliance filtering
Speech servicesSpeech-to-text, text-to-speech, translation
Custom modelsCLU (Conversational Language Understanding)

Domain 5: Implement Information Extraction Solutions (15–20%)

TopicKey Skills
Document IntelligenceStructured data from documents and forms
Content UnderstandingExtracting info from unstructured sources
Azure AI SearchIndex design, skillsets, vector search
Multimodal extractionHandling text, images, and mixed-content documents
Custom extractorsTraining models for domain-specific extraction

How to use this prep

Each challenge follows a developer-focused format:

SectionPurpose
Exam Skills CoveredOfficial skills this challenge addresses
ScenarioReal-world problem that frames the challenge
ArchitectureDiagram of the solution components
TasksStep-by-step with Python / C# / REST tabs
ValidationProgrammatic checks to verify your solution
Break & FixDeliberate misconfigurations to troubleshoot
Knowledge CheckExam-style questions
CleanupScripts to delete resources and avoid costs

Prerequisites

  • Programming experience — Python (primary) or C#
  • Azure subscriptionAzure Free Account with $200 credit recommended
  • Development environment — VS Code with Azure extensions, or GitHub Codespaces
  • Azure CLI — Authenticated and configured
  • AI-901 knowledge — Understanding of Azure AI Foundry concepts (recommended)
Lab Costs

Some challenges create billable resources (Azure OpenAI, Azure AI Search, Document Intelligence). Always run cleanup scripts after completing a challenge. Estimated total cost: $20-50 with careful cleanup.

Study resources

ResourceLink
Official Study GuideAI-103 Study Guide
Free Practice AssessmentPractice Questions
Exam SandboxTry the exam interface
Schedule the ExamPearson VUE

Microsoft Learn paths

Learning PathFocus
Develop AI solutions with Azure OpenAIGenAI fundamentals
Build AI agents with Azure AI Agent ServiceAgentic architecture
Implement RAG with Azure AI SearchRAG pipelines
Create computer vision solutionsVision services
Develop NLP solutionsText analysis

What changed from AI-103?

AspectAI-103 (retired)AI-103 (current)
FocusTraditional AI servicesGenAI + Agents first
Heaviest domainEvenly distributedGenAI/Agents (35-40%)
Bot FrameworkIncludedRemoved
Agent architectureNot coveredCore skill
RAG patternsBasic coverageDeep focus
Microsoft FoundryNot coveredCentral platform
LUISIncludedReplaced by CLU
Duration100-120 minutes~120 minutes

Ready? Start with Challenge 01 to plan and provision your first Azure AI solution.