AI-103: Azure AI Apps and Agents Developer
Exam version: Skills measured as of April 2026 | Passing score: 700/1000 | Duration: ~120 minutes | Cost: ~$20-50 for labs
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
| Domain | Weight | Challenges |
|---|---|---|
| Plan and manage an Azure AI solution | 15–20% | 01–10 |
| Implement generative AI and agentic solutions | 35–40% | 11–23 |
| Implement computer vision solutions | 10–15% | 24–30 |
| Implement text analysis solutions | 15–20% | 31–39 |
| Implement information extraction solutions | 15–20% | 40–49 |
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%)
| Topic | Key Skills |
|---|---|
| Architecture | Secure, scalable Azure AI solution design |
| Deployment | Planning, provisioning, and monitoring AI workloads |
| Compliance | Responsible AI, governance, and regulatory requirements |
| Security | Identity, access controls, and data protection |
| Monitoring | Performance tracking, cost management, alerting |
Domain 2: Implement Generative AI and Agentic Solutions (35–40%)
| Topic | Key Skills |
|---|---|
| Azure OpenAI + Foundry | Deploy and manage LLM models |
| RAG pipelines | Retrieval-Augmented Generation with Azure AI Search |
| Agent architecture | Design multi-step, goal-oriented agents |
| Agent orchestration | Multi-tool, task completion, and reasoning chains |
| Prompt engineering | System/user prompts, chaining, context window management |
| Agent monitoring | Tracking behavior, debugging, and optimization |
| Azure AI Agent Service | Building and deploying production agents |
Domain 3: Implement Computer Vision Solutions (10–15%)
| Topic | Key Skills |
|---|---|
| Azure AI Vision | Image/video analysis, classification, detection |
| Custom Vision | Training custom models for specific use cases |
| OCR | Document text extraction |
| Face API | Facial recognition and analysis |
| Production integration | Deploying vision models in apps |
Domain 4: Implement Text Analysis Solutions (15–20%)
| Topic | Key Skills |
|---|---|
| NLP tasks | Sentiment, key phrases, language detection |
| Entity extraction | Named entities, PII detection, summarization |
| Content moderation | Safety and compliance filtering |
| Speech services | Speech-to-text, text-to-speech, translation |
| Custom models | CLU (Conversational Language Understanding) |
Domain 5: Implement Information Extraction Solutions (15–20%)
| Topic | Key Skills |
|---|---|
| Document Intelligence | Structured data from documents and forms |
| Content Understanding | Extracting info from unstructured sources |
| Azure AI Search | Index design, skillsets, vector search |
| Multimodal extraction | Handling text, images, and mixed-content documents |
| Custom extractors | Training models for domain-specific extraction |
How to use this prep
Each challenge follows a developer-focused format:
| Section | Purpose |
|---|---|
| Exam Skills Covered | Official skills this challenge addresses |
| Scenario | Real-world problem that frames the challenge |
| Architecture | Diagram of the solution components |
| Tasks | Step-by-step with Python / C# / REST tabs |
| Validation | Programmatic checks to verify your solution |
| Break & Fix | Deliberate misconfigurations to troubleshoot |
| Knowledge Check | Exam-style questions |
| Cleanup | Scripts to delete resources and avoid costs |
Prerequisites
- Programming experience — Python (primary) or C#
- Azure subscription — Azure 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)
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
| Resource | Link |
|---|---|
| Official Study Guide | AI-103 Study Guide |
| Free Practice Assessment | Practice Questions |
| Exam Sandbox | Try the exam interface |
| Schedule the Exam | Pearson VUE |
Microsoft Learn paths
| Learning Path | Focus |
|---|---|
| Develop AI solutions with Azure OpenAI | GenAI fundamentals |
| Build AI agents with Azure AI Agent Service | Agentic architecture |
| Implement RAG with Azure AI Search | RAG pipelines |
| Create computer vision solutions | Vision services |
| Develop NLP solutions | Text analysis |
What changed from AI-103?
| Aspect | AI-103 (retired) | AI-103 (current) |
|---|---|---|
| Focus | Traditional AI services | GenAI + Agents first |
| Heaviest domain | Evenly distributed | GenAI/Agents (35-40%) |
| Bot Framework | Included | Removed |
| Agent architecture | Not covered | Core skill |
| RAG patterns | Basic coverage | Deep focus |
| Microsoft Foundry | Not covered | Central platform |
| LUIS | Included | Replaced by CLU |
| Duration | 100-120 minutes | ~120 minutes |
Ready? Start with Challenge 01 to plan and provision your first Azure AI solution.