Update, July 2026. AI-102 retired on June 30, 2026, so it can no longer be booked. Everything below is kept as a record of the decision candidates faced at the time. If you are starting now, go straight to AI-103.
You have five days.
If you’ve been studying for AI-102, the window to take it closes June 30, 2026, at 11:59 PM CST. The exam is gone after that. So is the certification it earns.
Five days is enough to decide well, and this is how.
The Certification Itself Is Also Retiring
This isn’t just an exam number update.
AI-102 is the exam. “Microsoft Certified: Azure AI Engineer Associate” is the certification it earns. Both retire June 30. The replacement isn’t just a renumbered version of the same thing. It’s a different exam earning a different credential with a new name: Azure AI Apps and Agents Developer Associate, earned by passing AI-103.
If you pass AI-102 before June 30, you get “Azure AI Engineer Associate” on your transcript. The credential expires annually. After June 30, there’s no way to earn it again. AI-103 is the only forward path.
What AI-103 Actually Tests
Put the two weightings side by side.
AI-102 (retiring June 30):
| Domain | Weight |
|---|---|
| Plan and manage an Azure AI solution | 20–25% |
| Implement generative AI solutions | 15–20% |
| Implement an agentic solution | 5–10% |
| Implement computer vision solutions | 10–15% |
| Implement natural language processing solutions | 15–20% |
| Implement knowledge mining and information extraction | 15–20% |
AI-103 (the replacement):
| Domain | Weight |
|---|---|
| Plan and manage an Azure AI solution | 25–30% |
| Implement generative AI and agentic solutions | 30–35% |
| Implement computer vision solutions | 10–15% |
| Implement text analysis solutions | 10–15% |
| Implement information extraction solutions | 10–15% |
Two things stand out. Generative AI and agents are now combined into one 30–35% domain. In AI-102, agentic content was 5–10%, a corner of the exam. In AI-103, it’s the centerpiece. NLP as a standalone discipline (15–20% of AI-102) is folded into “text analysis” at 10–15%, with implementation now happening through Azure AI Foundry tools rather than directly configuring individual language services.
Passing score: 700 out of 1000 on both exams.
Take AI-102 Before June 30 If…
You’re already deep in prep. If you’ve been studying for weeks and you’re scoring consistently within range on practice questions, book it today. Abandoning that preparation to restart with a structurally different exam isn’t the right trade.
Your organization recognizes the Azure AI Engineer title specifically. Some Microsoft partner certifications and specializations list “Azure AI Engineer Associate” by name. If your employer or partner program references that credential, earning it now gives you a year of validity to work with.
You’re not currently building agents or deploying models in Azure AI Foundry. AI-103’s 30–35% generative and agentic domain expects hands-on experience with Foundry: defining agent roles, building RAG pipelines, configuring multi-agent orchestration. If that’s not your current day job, the exam difficulty gap is real.
Wait for AI-103 If…
You haven’t started studying. Don’t rush a credential with no forward path. AI-103 maps to where Azure AI is heading.
You’re already building agents or working in Azure AI Foundry. AI-103 tests what you’re likely already doing. The 30–35% generative and agentic domain maps directly to deploying models, building agents with tool schemas, implementing RAG, and operating multi-agent workflows. Your live work is your study material.
You want a credential that stays current. “Azure AI Apps and Agents Developer Associate” is the cert Microsoft will maintain and update. “Azure AI Engineer Associate” has a fixed expiry and no renewal path once it’s retired.
If You’re Taking AI-102 This Week
With five days left, you can’t cover everything. Focus here.
The generative AI domain (15–20%) is where most candidates underperform. Questions aren’t “what is RAG”. They’re scenario-based: which Foundry deployment option fits a specific constraint, how to evaluate a model for a production use case, which approach grounds output in your data. Know the Azure AI Foundry project and hub structure.
The agentic domain (5–10%) is small but specific. Know what the Foundry Agent Service does, how multi-agent orchestration works conceptually, and when you’d use it instead of a prompt flow.
Knowledge mining (15–20%) catches people who skimmed it. Azure AI Search architecture, specifically skillsets, indexers, and knowledge store projections, comes up in scenario questions more than most candidates expect.
Pass score: 700. Microsoft’s official AI-102 study guide has the full objective list.
AI-900 holders: that exam also retires June 30. We covered the AI-900 vs AI-901 decision here.
Pass-IT has AI-102 adaptive practice questions with per-domain tracking. Alex can walk through any concept on the exam. Free to start, no card for your first session. With five days left, finding out exactly where you’re losing points is the fastest path to the score you need.
Moving to AI-103 instead? Pass-IT now has AI-103 practice questions for Developing AI Apps and Agents on Azure: Foundry, agents, RAG, and the rest of the new blueprint.
