Azure AI-102

The source page carries a warning that the certification and the renewal assessment are retired, with a link to learn more. It does not state an effective date for that retirement. Confirm the current status and retirement timing on Microsoft Learn before you book, and consider whether a newer credential better matches your goals.

You have 100 minutes to complete the assessment. It is proctored, and Microsoft states you may have interactive components to complete as part of it. A sandbox is available so you can try the different question types in the same interface used during the exam.

Yes. You can retake 24 hours after the first attempt. For subsequent retakes, the amount of time varies. The full details are in Microsoft's exam retake policy, which is worth reading before you schedule a second attempt so you plan your study window correctly.

Neither figure is stated on the source page. The published material covers duration, proctoring, retake timing, languages, and the assessed domains, but not a passing score, exam fee, question count, or certification validity period. Check your regional pricing and score-report details on Microsoft Learn or in your certification dashboard.

Microsoft positions this as intermediate level for AI engineers. You should have experience developing in Python or C#, be able to use REST APIs and SDKs to build secure image, video, natural language, knowledge mining, and generative AI solutions on Azure, understand the Azure AI portfolio and data storage options, and be able to apply responsible AI principles.

The source page states that no training is available for this exam. What is offered instead is a practice assessment to gauge style, wording, and difficulty, an exam sandbox to experience the interface, and exam prep videos with tips and strategies. You will need to assemble the rest of your study plan yourself.

The source lists six assessed areas - planning and managing an Azure AI solution, generative AI, agentic solutions, computer vision, natural language processing, and knowledge mining and information extraction - but publishes no percentage weights. Do not assume an even split; treat all six as examinable and distribute study time by your own weak spots.

Start with planning and management, since governance, provisioning, security, and monitoring decisions underpin every other area. Then work through generative AI and agentic solutions together, as they share model, prompt, and orchestration concepts. Finish with computer vision, NLP, and knowledge mining, which build on shared resource and storage fundamentals.

Two patterns dominate: treating the exam as recall of service names rather than decisions about which service and configuration fit a scenario, and skipping hands-on practice with SDKs and REST calls. The role description emphasizes building, deploying, integrating, tuning, and monitoring, so expect scenario-based items that reward practical implementation experience over memorized feature lists.

It is offered in English, Japanese, Chinese (Simplified), Korean, German, French, Spanish, Portuguese (Brazil), Chinese (Traditional), and Italian. If you test in a non-native language, allow extra time for reading scenario-heavy items carefully, since the 100-minute limit is the same regardless of language.