- Why "AI-102" Now Means AI-103
- Exam Format, Registration, and Fees
- Domain 1: Plan and Manage an Azure AI Solution
- Domain 2: Implement Generative AI and Agentic Solutions
- Domain 3: Implement Computer Vision Solutions
- Domain 4: Implement Text Analysis Solutions
- Domain 5: Implement Information Extraction Solutions
- Who Actually Hires for This Skill Set
- Mapping a Study Timeline to Domain Weights
- FAQ
- The old AI-102 exam and its Azure AI Engineer Associate credential retired June 30, 2026; AI-103 is the live path.
- AI-103 covers five domains, with "Implement generative AI and agentic solutions" weighted heaviest at 30-35%.
- Passing score is 700 on a 100-1000 scale; the exam runs 120 proctored minutes via Pearson VUE.
- Expect deep coverage of Microsoft Foundry, Foundry Agent Service, and Microsoft Agent Framework, not just single-service APIs.
Why "AI-102" Now Means AI-103
If you've been researching this credential for a while, you may still see it referenced by its old exam code. Here's the status that matters for anyone planning a study schedule in 2026: Exam AI-102, Designing and Implementing a Microsoft Azure AI Solution, and its associated credential, Microsoft Certified: Azure AI Engineer Associate, both retired on June 30, 2026 at 11:59 PM Central. The renewal assessment retired alongside it, which means the credential can no longer be earned or renewed through that path. If you certified before the retirement date, it stays listed in the Active Certifications section of your Microsoft transcript until its printed expiration - but there is no renewal route afterward.
The live successor is Exam AI-103, Developing AI Apps and Agents on Azure, which awards Microsoft Certified: Azure AI Apps and Agents Developer Associate. The training material followed the same path: the four-day AI-103T00-A course replaced AI-102T00 at the instructor-led training level. Anyone still searching under the old name should treat this article, and the rest of this site, as coverage of the current AI-103 domains while the old code phases out of search results and job postings.
Exam Format, Registration, and Fees
The exam is owned by Microsoft and delivered through Pearson VUE, either at a physical test center or via online proctoring. It runs 120 minutes and may include interactive components rather than pure multiple-choice items - expect drag-and-drop ordering, code completion, and scenario-based question sets alongside standard single/multiple-answer items. Microsoft does not publish a fixed item count or a pass rate for this exam, so treat any specific number you see elsewhere with skepticism; our AI-102 Pass Rate article covers what can and can't be verified.
Standard US pricing is $165 USD, and the fee varies by the country or region where you sit the exam. For a full breakdown of what that price does and doesn't include, see AI-102 Certification Cost.
The passing score is 700 on a 100-1000 scale. That threshold applies across every domain combined - there's no requirement to pass each domain individually, though a weak showing in the 30-35% weighted domain is harder to compensate for than a weak showing in a 10-15% domain. Our AI-102 Passing Score guide walks through the scoring mechanics in more detail.
If you fail on your first attempt, you can retake after 24 hours, with progressively longer waiting periods for each subsequent attempt. Once certified, the credential expires annually and renews free through an unproctored assessment on Microsoft Learn - no need to resit the full proctored exam every year.
Key Takeaway
Skills measured are current as of April 16, 2026. Most questions cover generally available features, but commonly used preview features can appear - don't assume preview-labeled services are automatically out of scope.
Domain 1: Plan and Manage an Azure AI Solution (25-30%)
This domain sets the foundation before any code gets written. It tests whether you understand resource provisioning, security boundaries, cost and monitoring controls, and responsible AI governance across an Azure AI solution - not just how to call an API.
Plan and Manage an Azure AI Solution
Candidates need working fluency with provisioning and securing Azure AI resources, along with the governance layer around them.
- Selecting and provisioning the right Azure AI resource types for a given workload
- Securing access with keys, Microsoft Entra ID, managed identities, and network isolation
- Monitoring, logging, and cost management across AI services
- Applying responsible AI principles and content safety controls at the planning stage
Because this domain sits at 25-30%, it's close in weight to several of the implementation domains combined - treat it as a first-class study area, not a warm-up chapter.
Domain 2: Implement Generative AI and Agentic Solutions (30-35%)
This is the single heaviest domain on the exam, and it's also the area where the shift from the old AI-102 to AI-103 shows up most clearly. Expect deep, applied coverage of agent-based architecture rather than isolated model calls.
Implement Generative AI and Agentic Solutions
This domain expects you to design, build, and troubleshoot agentic systems end to end, not just prompt a single model.
- RAG (retrieval-augmented generation) implementation patterns using Azure AI Search and Azure OpenAI in Foundry Models
- Multi-agent orchestration using Foundry Agent Service and Microsoft Agent Framework
- Function calling and defining tool schemas that agents can invoke reliably
- Conversation memory design for multi-turn and multi-session interactions
- Agent evaluation and error analysis, plus tracing and token analytics for cost and performance tuning
- Autonomous workflows that include human approval controls
Because this section is so heavily weighted and so different from legacy AI-102 material, it's worth cross-referencing your prep against a structured plan - see the AI-102 Study Guide for a first-attempt-focused approach built around this exact weighting.
Domain 3: Implement Computer Vision Solutions (10-15%)
Implement Computer Vision Solutions
A smaller but still testable domain covering image and video analysis capabilities within the Azure AI ecosystem.
- Image analysis, tagging, and object/face detection workflows
- Custom vision model training and deployment for domain-specific classification
- Video indexing and analysis scenarios
- Integrating vision outputs into broader agentic or application pipelines
Because this domain sits at 10-15%, it doesn't warrant the same study hours as generative AI, but skipping it entirely is risky since the exam draws items proportionally across all five domains.
Domain 4: Implement Text Analysis Solutions (10-15%)
Implement Text Analysis Solutions
This domain covers natural language understanding tasks that sit apart from the generative/agentic workload in Domain 2.
- Sentiment analysis, key phrase extraction, and entity recognition
- Language detection and translation workflows using Azure Translator
- Speech-to-text and text-to-speech scenarios using Azure Speech
- Custom text classification for domain-specific language tasks
Text analysis questions tend to be scenario-driven: you'll be given a business requirement and asked to pick the right service or configuration rather than recite a definition.
Domain 5: Implement Information Extraction Solutions (10-15%)
Implement Information Extraction Solutions
This domain focuses on pulling structured data out of unstructured documents and content - a common enterprise use case layered on top of the AI stack.
- Document processing pipelines using Azure Document Intelligence in Foundry Tools
- Structured and unstructured content parsing with Azure Content Understanding in Foundry Tools
- Prebuilt versus custom extraction model selection
- Combining extraction outputs with downstream RAG or agent pipelines from Domain 2
Note the overlap: information extraction feeds directly into the RAG patterns tested in Domain 2, so weak fundamentals here can cost points in two domains at once.
| Domain | Weight | Core Focus |
|---|---|---|
| 1. Plan and manage an Azure AI solution | 25-30% | Provisioning, security, governance, monitoring |
| 2. Implement generative AI and agentic solutions | 30-35% | RAG, multi-agent orchestration, tool calling, evaluation |
| 3. Implement computer vision solutions | 10-15% | Image/video analysis, custom vision models |
| 4. Implement text analysis solutions | 10-15% | NLP, sentiment, speech, translation |
| 5. Implement information extraction solutions | 10-15% | Document Intelligence, Content Understanding |
Who Actually Hires for This Skill Set
The audience profile behind this exam assumes app development experience in Python, plus familiarity with general AI, generative AI, and Azure services - there are no formal prerequisites, but that background is what the questions assume you already have. In practice, that maps to roles building production applications on top of Microsoft Foundry, Foundry Agent Service, Azure OpenAI in Foundry Models, and Azure AI Search: AI application developers, cloud solution engineers embedding agents into existing products, and platform teams standing up RAG and multi-agent systems for internal or customer-facing tools.
If you're evaluating whether this credential fits your career direction rather than just your current project, the AI-102 Jobs overview and the AI-102 Salary Guide lay out the qualitative picture without inventing numbers that aren't part of the official record.
Mapping a Study Timeline to Domain Weights
A generic four-week study calendar isn't useful on its own - what matters is which domain gets the most calendar time, and that should track the exam's weighting, not your personal comfort level.
Foundations: Domain 1
- Provision and secure Azure AI resources in a sandbox subscription
- Review Entra ID, managed identity, and content safety configuration
Deep dive: Domain 2
- Build a RAG pipeline against Azure AI Search and Azure OpenAI in Foundry Models
- Stand up a multi-agent workflow with Foundry Agent Service and Microsoft Agent Framework
- Practice tracing, token analytics, and agent evaluation on your own test cases
Breadth pass: Domains 3-5
- Run computer vision, text analysis, and Document Intelligence labs back to back
- Focus on where extraction outputs feed into Domain 2 pipelines
Spacing repeated review sessions across these weeks (rather than cramming each domain once) works well here specifically because Domain 2 content compounds - orchestration and evaluation concepts you touch in week 2 get reused when you circle back in week 4. For a longer walkthrough of this approach, see the AI-102 Study Guide, and gauge your overall readiness against How Hard Is the AI-102 Exam? before you lock in a test date.
FAQ
No. Exam AI-102 and the Azure AI Engineer Associate credential retired June 30, 2026. The active path today is Exam AI-103, Developing AI Apps and Agents on Azure, which awards Microsoft Certified: Azure AI Apps and Agents Developer Associate.
Domain 2, "Implement generative AI and agentic solutions," carries the heaviest weight at 30-35% and covers RAG, multi-agent orchestration, and agent evaluation - it deserves the largest share of your prep time.
There are no formal prerequisites, but the exam assumes app development experience in Python along with familiarity with general AI, generative AI, and Azure services. See AI-102 Requirements for the full picture.
You need a 700 on a 100-1000 scale. Microsoft does not publish a fixed item count or pass rate, so focus on domain coverage rather than chasing a specific question count.
You can retake the exam after 24 hours. Subsequent retakes carry progressively longer required waiting periods, so it pays to close domain gaps before rebooking.