Study guide · Exam overview
The format, passing score, and five content domains for Microsoft Certified: Azure AI Apps and Agents Developer Associate, plus the Foundry and agent tasks worth practicing before you book AI-103.
Exam version notice: This overview follows the Microsoft study guide for Exam AI-103: Developing AI Apps and Agents on Azure, skills measured as of April 16, 2026. Exam AI-102 retired on June 30, 2026. AI-103, which earns Azure AI Apps and Agents Developer Associate, replaced it. AI-102 can no longer be scheduled, and a retired certification cannot be renewed after its retirement date. AI-103 expects Python application experience and familiarity with generative AI and Azure services. Pricing, languages, and delivery options can change. Confirm them on the Microsoft certification site before you register. Items tagged Extra are useful context that the study guide does not name as a measured skill. Introductory concepts are on the AI-901 overview.
Microsoft Certified: Azure AI Apps and Agents Developer Associate checks whether you can plan, build, and operate agents and AI solutions on Microsoft Foundry. The work is associate-level development: choose a model, ground it, give an agent tools and limits, then implement vision, language, speech, and extraction paths in Python.
AI-102 was the Azure AI Engineer Associate exam. Its replacement is not a renumbered copy of that outline. AI-103 centers Foundry projects, generative applications, agent orchestration, and the safety and retrieval controls around them. Classic product drills that the April 16, 2026 skill list does not name are marked Extra in the notes.
Figures below come from the AI-103 certification page and the study guide, reviewed in October 2026.
| Specification | AI-103 |
|---|---|
| Level | Associate. No prerequisite certification is required. Python development experience is expected. |
| Length of test | 120 minutes to complete the assessment. The exam may include interactive components. Confirm the time shown when you schedule. |
| Number of questions | Microsoft does not publish a fixed count for AI-103. Do not treat a practice length as the live exam length. |
| Passing score | 700 or greater on a scale of 1 to 1,000. A scaled score is not the percentage of questions you answered correctly. |
| Recommended experience | Building apps in Python and working with general AI, generative AI, and Azure services, including Microsoft Foundry. |
| Exam price | Price depends on the country or region where the exam is proctored. Confirm the current fee when you schedule. |
| Validity | Associate certifications expire annually. Renew with the free online assessment on Microsoft Learn. Confirm the current policy. |
| Delivery | Schedule through Pearson VUE. English is updated first. If the exam is not offered in your language, you can request an additional 30 minutes. |
A blank answer earns no credit. Microsoft exams do not subtract extra points for a wrong answer, so answer every question. Most questions cover generally available features. Preview features can appear when they are commonly used. Confirm in the exam sandbox whether Microsoft Learn opens during the exam. If it does, the timer keeps running while you use it.
These ranges are the share of scored content in the study guide dated April 16, 2026. The 80 practice questions use shares inside those ranges that add to 100 percent: 27.5%, 32.5%, 13.75%, 13.75%, and 12.5%.
Choose the service, then keep it private, observable, and governed.
The largest share. Build the app and the agent, then operate them.
Create media, read media, and stop unsafe visual instructions.
Language tasks and speech as an agent modality.
Get content into an index, then hand clean fields to an agent.
The study guide names implementation choices more often than product slogans. The same map is in the core notes and flashcards.
| Need | Choose | Remember |
|---|---|---|
| Answers must cite your documents | Retrieval over an index | A larger model does not search a private library by itself. |
| Exact SKU and a paraphrase | Hybrid search | Vector-only search can miss the exact token. |
| No keys in source code | Managed identity | Pair it with private networking when the endpoint must stay off the public internet. |
| A tool may draft but must not send | Approval and oversight | Autonomy without a gate fails a side-effecting action. |
| Edit one region of a photo | Masked inpainting | Generating a new scene is a different task. |
| Text hidden in an image overrides instructions | Indirect prompt-injection detection | Do not treat pixels as a higher-priority system prompt. |
| Scans have no text layer | OCR in the ingestion path | Indexing the file name does not read the page. |
| Extra LUIS, QnA Maker, or a Custom Vision training project from AI-102 | Not the AI-103 midpoint map | Useful history. April 16, 2026 skills center Foundry, agents, and the five domains above. |
Microsoft says you should be familiar with the exam sandbox, and the certification page says this exam may include interactive components. This site’s practice questions are single-answer multiple choice. They are not a claim about the live item mix.
A typical stem gives a constraint and asks for the best control. The refund must not be submitted. The SKU must match exactly. The photo contains a hostile instruction. Distractors are nearby features that solve a different problem. Retrieval is not image generation. A safety filter is not a larger context window. Hybrid search is not vector search alone when the exact token matters.
There is no published extra penalty for a wrong guess. A blank answer earns nothing. On a timed practice exam, answer every item, then use the review to see which domain actually needs work.
A practical sequence if AI-102 notes still stop at naming retired services.
Spend the most time on planning and on generative apps and agents. They are more than half the outline.
Deploy a model, connect a Python client, and add identity, a private endpoint, and a repeatable release.
Index content, compare vector and hybrid retrieval with optional semantic reranking, call it as a tool, and require approval before a side effect.
Separate generation from understanding, and speech-to-text from text-to-speech. Extract fields from a scan.
Use the five decks until you can answer before the card flips. Revisit misses the next day.
Work through original questions, review every explanation, and return to the notes on domains below 75 percent. Before booking, check every objective and practice unfamiliar scenarios; a repeated-bank percentage is not evidence of exam readiness. The certification page lists 120 minutes for the assessment; confirm your booking and allow for longer interactive tasks rather than treating each practice question as an equal slice of exam time.
The overview, core notes, flashcards, and practice exams are all published.
Exam format, the 700 passing score, domain weights, and a study plan.
Reference notes for planning, agents, vision, language, speech, and extraction.
40 recall cards in five decks, plus memory notes for the highest-yield distinctions.
80 original questions with custom exams, explanations, and a score report by domain.
The April 16, 2026 objectives remain the basis for this study path. A named feature can change before the exam guide is revised. In particular, the guide still names Content Understanding pro mode, while current documentation identifies its old preview API as retired and describes migration to agentic document analysis. Learn the objective, then check current API support before building a lab.
Use the official AI-103 study guide for assessed skills and the Content Understanding migration guidance for implementation constraints. The notes include retrieval-stage distinctions, response validation, and five guided labs with expected failure cases.
A score of 700 or greater is required to pass. Microsoft reports scores on a scale of 1 to 1,000. A scaled score is not the percentage of questions you answered correctly.
The certification page gives you 120 minutes to complete this assessment. Microsoft does not publish a fixed question count for AI-103. Confirm the time shown when you schedule.
The study guide dated April 16, 2026 measures five areas: 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%), and implement information extraction solutions (10–15%).
No. Exam AI-102, Azure AI Engineer Associate, retired on June 30, 2026. AI-103 earns Microsoft Certified: Azure AI Apps and Agents Developer Associate. A certification already earned can remain on your Microsoft transcript, and you cannot renew a retired certification after its retirement date.
Microsoft associate certifications expire annually. You renew by passing a free online assessment on Microsoft Learn. Confirm the current renewal policy before you rely on it.
No. The practice questions on this site are original. They are not Microsoft exam items, and a practice score does not predict a result.
Service names and portal labels change. Use the current Microsoft study guide as the authority when it differs from these notes.
Useful companions while you study.
Convert structured data between JSON and YAML with formatting and validation.
Encode text and files to Base64 or decode Base64 data with UTF-8 support.
Compare two text blocks and highlight added, removed, and changed content.
Convert CSV spreadsheet data to JSON or turn JSON arrays into downloadable CSV files.
Exam versions, objectives, prices, languages, and retirement dates can change. Verify the details on the Microsoft Learn certification page before you register.
Certicab provides general informational, educational, and convenience resources. Calculations, conversions, estimates, and learning materials may contain errors or become outdated. Financial, tax, medical, legal, and travel information is not professional advice. Verify important results and current requirements with qualified professionals or authoritative sources. Protect sensitive files and personal information, review each tool’s privacy limitations, and use only content you are authorized to process. You are responsible for how you use and share results. Study resources are independent and do not guarantee exam success or imply certification-provider endorsement. Microsoft, Azure, and related marks are trademarks of Microsoft Corporation. Certicab is not affiliated with, endorsed by, or sponsored by Microsoft.
Certi