- AI-102 and its Azure AI Engineer Associate credential retired June 30, 2026; AI-103 is the live successor.
- AI-103 awards Microsoft Certified: Azure AI Apps and Agents Developer Associate and costs $165 USD at standard US pricing.
- Implement generative AI and agentic solutions is the heaviest domain at 30-35% of the exam.
- The exam runs 120 minutes, is proctored, and requires a score of 700 on a 100-1000 scale to pass.
What Is AI-102, Exactly?
If you're searching "what is a AI-102," you've likely landed in the middle of a naming transition. Historically, AI-102 referred to Designing and Implementing a Microsoft Azure AI Solution, the exam behind the Microsoft Certified: Azure AI Engineer Associate credential. That exam and its associated certification were owned by Microsoft and delivered through Pearson VUE, either at physical test centers or via online proctoring.
Here's the critical detail: that version of the credential is no longer earnable. Microsoft retired Exam AI-102 and the Azure AI Engineer Associate certification, along with the renewal assessment, on June 30, 2026 at 11:59 PM Central. In its place stands Exam AI-103, Developing AI Apps and Agents on Azure, which awards the Microsoft Certified: Azure AI Apps and Agents Developer Associate credential. When people say "AI-102" today, they're almost always referring to this successor path, and that's the version this site and our AI-102 study guide focus on.
The Retirement You Need to Know About
If you already hold the Azure AI Engineer Associate certification earned before the retirement date, it doesn't disappear from your Microsoft transcript. It remains listed in your Active Certifications section until its printed expiration date. What changes is that there's no renewal path for it anymore - once it expires, you can't extend it through the old renewal assessment. The only forward path is earning the new credential through AI-103.
For anyone starting from scratch, this matters practically: don't invest study time preparing for content tied to the old exam number. The skills, tools, and domain structure measured today belong to AI-103, and they reflect a meaningfully different technology stack - one built around agentic AI, orchestration frameworks, and Foundry-based tooling rather than the standalone cognitive services model of the earlier exam.
Key Takeaway
Treat any AI-102 prep material dated before the retirement announcement with caution. Cross-check it against the current domain list before relying on it for exam-day content.
The Five Domains That Define the Exam
The exam is organized into five weighted domains. Understanding the weighting isn't optional trivia - it should directly shape how you allocate study hours. For a full breakdown of subtopics inside each area, see our AI-102 exam domains guide.
Domain 1: Plan and manage an Azure AI solution (25-30%)
Covers the foundational decisions around provisioning, securing, and governing AI resources before any application logic gets written.
- Resource planning across Azure AI services
- Security, monitoring, and cost considerations for AI workloads
Domain 2: Implement generative AI and agentic solutions (30-35%)
The single heaviest domain and the one most candidates underestimate. It centers on building working agentic systems, not just calling a model API.
- RAG (retrieval-augmented generation) implementation
- Multi-agent orchestration and function calling with tool schemas
- Conversation memory, agent evaluation, and error analysis
- Tracing, token analytics, and autonomous workflows with approval controls
Domain 3: Implement computer vision solutions (10-15%)
Applying Azure's vision capabilities to real scenarios such as image analysis and object detection within an AI application.
Domain 4: Implement text analysis solutions (10-15%)
Focuses on extracting sentiment, key phrases, and language understanding signals from unstructured text.
Domain 5: Implement information extraction solutions (10-15%)
Pulling structured data out of documents and forms using extraction-oriented Azure tooling.
Notice how Domain 2 alone carries nearly as much weight as Domains 3, 4, and 5 combined. That's not an accident of exam design - it reflects where Microsoft expects working developers to spend their time today. If you're wondering how tough this actually plays out in practice, our difficulty guide digs into the question style further.
Registration, Format, and Scoring
The exam is 120 minutes long, proctored, and may include interactive components rather than pure multiple-choice items - expect scenario-based questions that mirror real development decisions. Microsoft doesn't publish a fixed item count or an official pass rate, so treat any specific number you see elsewhere skeptically; for a data-grounded look at what's actually known, see our pass rate breakdown.
Standard US pricing sits at $165 USD, though the amount varies by the country or region where you sit the exam. A full cost breakdown, including regional variation, is covered in our certification cost guide.
You need a score of 700 on a 100-1000 scale to pass - our passing score article explains exactly what that number means in context. If you don't pass on the first attempt, you can retake after 24 hours, with progressively longer mandatory waits for each subsequent attempt. Once earned, the associate-level certification follows Microsoft's standard annual expiration, renewable for free by passing an unproctored online assessment on Microsoft Learn - no need to resit the full proctored exam each year.
Who Actually Earns This Credential
There are no formal prerequisites listed for the exam, but the audience profile assumes real, hands-on experience: app development in Python, plus working familiarity with general AI concepts, generative AI patterns, and Azure services more broadly. This isn't an entry-level credential for someone who has only read about AI - it's built for developers already writing code against cloud services who now need to add agentic and generative AI implementation skills.
Concretely, candidates are expected to have working knowledge of tools including Microsoft Foundry, Foundry Agent Service, Microsoft Agent Framework, Azure OpenAI in Foundry Models, Azure AI Search, Azure Content Understanding in Foundry Tools, Azure Document Intelligence in Foundry Tools, Azure Speech, and Azure Translator. That's a wide surface area, and it's why the certification tends to attract application developers, AI-focused backend engineers, and solution builders who are integrating language models and agents into production systems rather than researchers or data scientists. For a closer look at eligibility expectations, see our requirements guide, and for how the credential tends to translate into roles, check AI-102 jobs.
Key Takeaway
If you haven't built anything with function calling, retrieval pipelines, or multi-agent tool orchestration yet, prioritize hands-on practice with those patterns before diving into practice questions on ../.
Mapping a Study Plan to the Domain Weights
Rather than studying domains in the order Microsoft lists them, allocate time proportionally to their weight - and put the heaviest, most conceptually dense domain in the middle of your plan, once you have foundational vocabulary but still have time to iterate.
Foundations (Domain 1)
- Provisioning and securing Azure AI resources
- Cost and monitoring fundamentals
Generative AI and Agents (Domain 2)
- Build a small RAG pipeline end-to-end
- Practice multi-agent orchestration and tool-schema function calling
- Work through agent evaluation, tracing, and approval-control patterns
Vision, Text, and Extraction (Domains 3-5)
- Computer vision scenarios and text analysis workflows
- Document- and form-based information extraction
Review and Practice Exams
- Run timed practice sessions on ../ to simulate the 120-minute format
- Revisit weak domains identified from practice results
This sequencing respects the reality that Domain 2 alone accounts for 30-35% of the exam and touches skills - orchestration, memory, evaluation - that take longer to internalize than single-service tasks like calling a vision API. For a condensed, at-a-glance version of must-know facts to revisit right before test day, bookmark our AI-102 cheat sheet.
AI-102 vs. AI-103 at a Glance
| Attribute | Legacy AI-102 Path | Current AI-103 Path |
|---|---|---|
| Exam name | Designing and Implementing a Microsoft Azure AI Solution | Developing AI Apps and Agents on Azure |
| Credential earned | Microsoft Certified: Azure AI Engineer Associate | Microsoft Certified: Azure AI Apps and Agents Developer Associate |
| Status | Retired June 30, 2026 (no longer earnable or renewable) | Active and earnable |
| Associated training | AI-102T00 | AI-103T00-A (four-day course) |
| Core technology focus | Cognitive services integration | Foundry, agent frameworks, and generative AI orchestration |
If you're deciding whether the effort is worthwhile at all given the shift, our ROI analysis and salary guide both weigh the current credential against the roles it opens up.
Frequently Asked Questions
No. Exam AI-102 and the Azure AI Engineer Associate credential, along with its renewal assessment, retired on June 30, 2026. The only active path forward is Exam AI-103, which awards the Azure AI Apps and Agents Developer Associate credential.
It stays listed in the Active Certifications section of your Microsoft transcript until its printed expiration date. There is no renewal path afterward, so it will eventually expire without an option to extend it.
The exam runs 120 minutes, is proctored, and may include interactive question types. You need a score of 700 on a 100-1000 scale to pass; Microsoft doesn't publish a fixed item count.
There are no formal prerequisites, but the exam assumes practical Python development experience along with familiarity with general AI, generative AI concepts, and Azure services.
You can retake the exam after a 24-hour waiting period. Subsequent failed attempts carry progressively longer mandatory waits before you can try again.
Understanding "what is AI-102" today means understanding a transition: an old exam number pointing at a retired credential, and a live exam - AI-103 - carrying the torch with a heavier emphasis on generative AI and agentic development. Once you separate those two facts, the rest of your prep, including timed runs on ../, becomes far more focused.