Introduction

The AI-102 Microsoft Azure AI Engineer Associate Exam is an exam that validates one’s understanding of Microsoft’s various AI product offerings. I took this exam in January 2026, an important thing to consider when reviewing this post as the world of AI is constantly evolving. Specifically the version of the exam I took was as of December 23, 2025. Like most exams this is an important piece to consider when reviewing study material, perhaps meaning more to this exam then others.

Preparation

This was perhaps the hardest piece, to find good study material for the exam. There is the Microsoft practice exam and of course MS Learn. Everyone learns different and I tend to prefer hands on or video walkthroughs when learning something new. For me I leveraged a lot of what I learned in my AI-in-a-Box solution and found a series of videos.

Be warned that there is a lot of information out there and study guides that are severely outdated. I came across videos discussing how prompt flow would be key or how to best leverage Foundry Hubs. Both of which have fallen off the product direction and as such are, for the most part, no longer on the exam.

Experience

This exam for sure focused on the product offerings Microsoft provides with the AI space as opposed to how to best develop or architect solutions. Things like file type and size limitations by services was questioned. Other comments around how to structure and use language translation for text and speech will also appear on the exam.

Perhaps one of the hardest parts is the terminology used is not in sync with what was announced at Ignite. (i.e. Azure Foundry rebranding to Microsoft Foundry) For that matter, as of this writing, all the new features and capabilities discussed at Ignite will not appear on the exam.

As I alluded to earlier Prompt Flow does not appear on the exam; however, references to how to us Azure Open AI Services did appear, though not in depth. Just recognition how to interact with it was. Additionally Foundry Hubs appears as well but specifically around network security as that is not of equal footing as Foundry Projects, yet.

One thing that was different and I hadn’t seen before on a Microsoft Exam was that you had the ability for your coding questions to appear in .NET or in python. Once you choose you can’t go back so be sure going into the exam (and while studying) which version of the exam you’d like to take. I chose .NET just because I’ve had more experience in that tech stack.

Study Guide

Since this exam was about AI, I decided to do something new with a study guide. I outline my experience after the exam and decided then to feed that outline into GitHub CoPilot, scoped on the azure-docs repository. This created a fun and detailed experience that re-enforced my learnings and thoughts via the support of official documentation. I have posted it out on Github JFolberth/Microsoft-AI-102-Study-Guide Here is the Table of Contents for the Study Guide:

• 1. File Types and Size Limitations
  • Document Intelligence (Form Recognizer)
  • Custom Vision
  • Language Services (Text Analytics)
  • Speech Services
  • Azure AI Search
  • Azure Storage (for AI services)
  • Azure Video Indexer (Video processing)
• 2. Control Plane – REST API Calls
  • API key note
  • Common REST Endpoint structure
  • Document Intelligence — Analyze (REST example)
  • Language Services — AnalyzeTextAsync (overview & REST)
  • Custom Vision — Prediction (REST)
  • Speech-to-Text (REST)
  • Azure OpenAI — Chat completions (REST)
• 3. Content Filtering and Message Flagging
• 4. Multi-Language Support for Audio and Video
• 5. Model Selection Criteria (Simplified)
• 6. Azure Security: Key Rotation
• 7. Synonyms in Document Intelligence
• 8. Confidence Scoring (Short)
• 9. Azure AI Language (Question Answering)
• 10. Token Calculation & Max Token Behavior
• 11. Image Model Training (Vision)
• 12. Text / Image Processing Methods (REST)
• 13. Semantic Kernel — Prompt Template Formats
• 14. SSML for Speech — Styles & Example
  • SSML attribute quick reference (Azure Speech)
• 15. AI Search Indexing (Summary)
  • What resource types can be queried with Azure AI Search?
  • Leverage auto tag and classification in vision search
• Video (Azure Video Indexer & video processing)
  • When to run Video Indexing (and what you get)
• 16. Quick Reference Tables
  • API Key Quick Reference
  • File Size / Limits Summary
  • Video File Outputs (Video Indexer)
  • Model Selection (cheat sheet)
  • AnalyzeTextAsync() Snapshot
  • Token Behavior Summary
  • Confidence Score Thresholds (guidance)
  • JSONPath Wildcard Examples
• 17. Entity Linking (Wikipedia links from text)
• 18. Import/Export Azure AI Language Projects
• 19. Foundry: Scope model access for multiple APIs

For those looking to have some with it I’d advise asking CoPilot to create “a question bank of multiple choice questions covering the various areas. The questions should not repeat. After answering each question provide a detailed answer as to why the answer is correct and feel free to leverage MS Learn Docs to provide rationale into why the other answers are incorrect”. The last part of querying MS Learn only works if you have the Microsoft Learn MCP server configured.

Conclusion

This exam does a good job of testing knowledge on the various concepts and product offerings. It will expose one to what is available and when to use offerings such as: Document Intelligence, AI Vision, Speech to Text, and Azure Language.

This exam is weak on architecting these solutions. This could be in part due to the ever-changing landscape. For example there wasn’t anything around design, observability, high availability, or general architecture concepts that tend to appear on the majority of Microsoft exams.

If you like what you read around exam certifications feel free to check out my other blogs on this topic. As always I am curious to hear what others have to say on their experience or concerns taking the exam!