- AI-102 (Azure AI Engineer Associate) retired June 30, 2026 at 11:59 PM Central; the live path is AI-103.
- AI-103 covers five domains, with "Implement generative AI and agentic solutions" weighted heaviest at 30-35%.
- The exam runs 120 minutes, costs $165 USD at standard US pricing, and requires a 700/1000 to pass.
- Holders who certified before retirement keep AI-102 as an Active Certification until it prints its expiration, with no renewal path.
What Is AI-102, Really?
"AI-102" refers to Designing and Implementing a Microsoft Azure AI Solution, the Microsoft exam that awarded the Microsoft Certified: Azure AI Engineer Associate credential. It was a proctored, scenario-heavy assessment delivered through Pearson VUE, built for developers who build applications on top of Azure's AI services rather than data scientists who train models from scratch.
Here is the part that trips up almost everyone searching for this exam in 2026: AI-102 and its associated Azure AI Engineer Associate credential retired on June 30, 2026 at 11:59 PM Central, along with the renewal assessment that used to keep it active. That means AI-102 can no longer be earned or renewed by anyone starting today. If you already passed it before the retirement date, it stays listed in the Active Certifications section of your Microsoft transcript until its printed expiration date - but there is no renewal path once that date arrives.
The successor is Exam AI-103: Developing AI Apps and Agents on Azure, which awards Microsoft Certified: Azure AI Apps and Agents Developer Associate. If you're researching "what is AI-102" today, the honest answer is: it's the retired predecessor, and AI-103 is the credential you actually need to pursue. This site treats "AI-102" as shorthand for that whole lineage, and our content is built around the live AI-103 exam so you're studying material that still counts.
The AI-102 to AI-103 Transition, Explained
Microsoft retires exams when the underlying platform shifts enough that the old blueprint no longer reflects how the job is actually done. AI-102 was built around discrete Azure AI services called individually from application code. AI-103 reflects a different reality: developers now orchestrate agents, tool calls, and retrieval pipelines through unified platforms like Microsoft Foundry rather than wiring together isolated APIs.
At the training level, this shows up directly: the four-day AI-103T00-A course replaced the older AI-102T00 course. That's a strong signal Microsoft considers this a full generational replacement, not a minor refresh. For a deeper breakdown of what actually changed content-wise, see our AI-102 Certification overview, which maps the old and new blueprints side by side.
| Attribute | AI-102 (Retired) | AI-103 (Current) |
|---|---|---|
| Credential | Azure AI Engineer Associate | Azure AI Apps and Agents Developer Associate |
| Status | Retired June 30, 2026, 11:59 PM Central | Live, earnable now |
| Training Course | AI-102T00 | AI-103T00-A (four days) |
| Focus | Individual AI service integration | Agentic and generative AI app development |
| Renewal | No longer possible | Free annual renewal via Microsoft Learn assessment |
Key Takeaway
If you have not started studying yet, don't start with AI-102 material - go straight to AI-103. It's the only version of this credential you can actually earn.
Exam Format, Fees, and Registration Mechanics
The live exam is delivered by Microsoft through Pearson VUE, either at a physical test center or via online proctoring, so you can choose whichever format fits your schedule. A few mechanics matter more than people expect:
- Duration: 120 minutes, and the session may include interactive question components rather than plain multiple choice.
- Price: $165 USD at standard US pricing; the exact fee varies depending on the country or region where you sit the exam.
- Passing score: 700 on a 100-1000 scale. Microsoft does not publish a fixed question count or a pass rate, so treat any specific number you see elsewhere with skepticism.
- Retakes: a failed first attempt can be retried after a 24-hour wait, with progressively longer waits required for each subsequent attempt.
- Renewal: like other Microsoft associate certifications, it expires annually, but renewal is free and happens through an unproctored online assessment on Microsoft Learn - no need to retake the full proctored exam.
For a full walkthrough of every fee scenario and regional variation, our AI-102 Certification Cost 2026: Complete Pricing Breakdown article goes deeper than this section. And if you want the exact scoring mechanics spelled out further, read AI-102 Passing Score 2026: Exactly What You Need to Pass.
There are no formal prerequisites enforced at registration. That said, Microsoft's audience profile assumes app development experience in Python plus familiarity with general AI, generative AI concepts, and Azure services broadly. If you're unsure whether your background is sufficient, our AI-102 Requirements 2026: Eligibility, Prerequisites & How to Qualify guide breaks down what "sufficient experience" actually looks like in practice.
The Five Exam Domains
The current blueprint organizes skills into five domains, and the weighting tells you exactly where to spend your study hours. Skills measured are current as of April 16, 2026, and most questions target generally available features - though commonly used preview features can still show up, so don't assume anything labeled "preview" is off-limits.
Domain 1: Plan and manage an Azure AI solution (25-30%)
Covers resource provisioning, security, monitoring, and cost management decisions you make before writing a single line of application code.
- Selecting and configuring the right Azure AI resources for a given scenario
Domain 2: Implement generative AI and agentic solutions (30-35%)
This is the single heaviest domain and the clearest signal of how much the job has changed. It covers RAG implementation, multi-agent orchestration, function calling and tool schemas, conversation memory, agent evaluation and error analysis, tracing and token analytics, and autonomous workflows with approval controls.
- Building agents that call tools, retain memory, and operate within human-approval boundaries
Domain 3: Implement computer vision solutions (10-15%)
Image analysis, object detection, and vision-based application patterns built on Azure's vision services.
- Choosing between prebuilt and custom vision models for a given use case
Domain 4: Implement text analysis solutions (10-15%)
Sentiment, key phrase extraction, entity recognition, and language-centric processing pipelines.
- Selecting the right text analysis feature for structured versus unstructured input
Domain 5: Implement information extraction solutions (10-15%)
Pulling structured data out of documents and unstructured content using Foundry Tools like Document Intelligence and Content Understanding.
- Designing extraction pipelines that feed downstream generative workflows
For a domain-by-domain study plan rather than just the weighting, see AI-102 Exam Domains 2026: Complete Guide to All 5 Content Areas. And if you're trying to gauge how tough this actually is relative to other associate-level exams, How Hard Is the AI-102 Exam? Complete Difficulty Guide 2026 covers that in detail.
Core Technologies You're Expected to Know
Domain names only tell part of the story. What actually shows up in exam scenarios is hands-on familiarity with a specific technology stack:
- Microsoft Foundry - the unified platform for building, deploying, and governing AI applications
- Foundry Agent Service and the Microsoft Agent Framework - for building and orchestrating multi-agent systems
- Azure OpenAI in Foundry Models - for generative model access and prompt-driven app logic
- Azure AI Search - the retrieval layer behind most RAG implementations tested in Domain 2
- Azure Content Understanding in Foundry Tools and Azure Document Intelligence in Foundry Tools - for information extraction scenarios
- Azure Speech and Azure Translator - for speech and language-adjacent application features
Notice how many of these tools sit under the "Foundry Tools" umbrella. Microsoft has consolidated a lot of what used to be separate, individually-branded services into this unified naming, and exam questions expect you to recognize that consolidation rather than think in terms of the old service names.
Who Actually Earns This Certification
The audience profile Microsoft describes isn't a data scientist - it's an application developer who is comfortable in Python and already has some exposure to general AI, generative AI, and Azure services, and now needs to build production features on top of agentic and generative AI platforms. That typically includes:
- Backend and full-stack developers adding AI features to existing Azure-hosted applications
- Cloud solution developers responsible for integrating Foundry-based agents into customer-facing products
- Platform and integration engineers building tool-calling and orchestration layers between internal systems and generative models
If you're evaluating whether this credential actually translates into better roles or compensation, our AI-102 Salary Guide 2026: Complete Earnings Analysis and Is the AI-102 Certification Worth It? Complete ROI Analysis 2026 articles walk through that decision in more depth, and AI-102 Jobs looks at the kinds of postings that reference this credential specifically.
A Realistic Way to Sequence Your Prep
Given that Domain 2 alone accounts for 30-35% of the exam, it deserves more calendar time than any other section - but it shouldn't be studied in isolation, since it depends on concepts from Domain 1 (resource setup) and Domain 5 (extraction feeding into agent context).
Foundation and Planning
- Provision Foundry resources and review security/monitoring patterns from Domain 1
- Get comfortable with the Foundry portal before touching agent logic
Generative AI and Agents (Domain 2)
- Build a RAG pipeline against Azure AI Search
- Implement function calling and tool schemas in Foundry Agent Service
- Practice reading tracing and token analytics output
Vision, Text, and Extraction
- Work through computer vision and text analysis scenarios
- Pair Document Intelligence extraction output with a downstream agent call
Review and Simulation
- Run full-length timed practice exams under the 120-minute limit
- Revisit weak domains identified from practice results
If you want a condensed, day-of-review companion once you've finished the structured plan above, keep our AI-102 Cheat Sheet 2026: One-Page Review of Must-Know Facts handy, and for the full methodology behind this sequencing, see the AI-102 Study Guide 2026: How to Pass on Your First Attempt. You can also run timed practice tests throughout this five-week window to track whether your recall is actually improving or just your familiarity with the material.
Before you lock in a testing date, double-check the current testing windows and any registration deadlines on AI-102 Exam Dates 2026: Testing Windows, Deadlines & Scheduling, since scheduling constraints can shift how you pace the weeks above.
Frequently Asked Questions
No. AI-102 and the Azure AI Engineer Associate credential retired on June 30, 2026 at 11:59 PM Central, along with the renewal assessment. New candidates must pursue AI-103 instead.
No. It remains in the Active Certifications section of your Microsoft transcript until its printed expiration date. There is simply no way to renew it once that date passes.
The content is different rather than strictly harder - it shifts weight toward generative AI and agentic orchestration (30-35% of the exam), which requires hands-on comfort with Foundry Agent Service and tool calling.
There's no formal prerequisite, but the exam's audience profile assumes app development experience in Python plus familiarity with general AI and Azure services, so lacking that background will make several scenarios harder to reason through.
It's a 120-minute proctored exam, possibly including interactive question types, and you need a score of 700 out of a possible 1000 to pass.