Exam guide
How to Pass the Microsoft AI-103 in 2026: Format, Cost, Domains and Study Plan
Updated 11 Nov 2026 · 5 min read
On this page
What the Microsoft AI-103 is and who it is for
The AI-103, Developing AI Apps and Agents on Azure, is the exam behind the Microsoft Certified: Azure AI Apps and Agents Developer Associate certification. Microsoft Learn lists it at intermediate level for the AI Engineer and Developer roles, with Azure and Microsoft Foundry as the products. The certification page describes it as validating your expertise in designing, developing and deploying advanced Azure AI solutions using Python and Microsoft Foundry.
Microsoft describes the candidate as an Azure AI engineer who builds, manages and deploys agents and AI solutions that take advantage of Microsoft Foundry. You should have experience developing apps by using Python and be familiar with the capabilities of general AI, generative AI and Azure services. The role covers planning and managing Azure AI solutions and implementing generative AI and agentic solutions, computer vision, text analysis and information extraction. You work with business stakeholders, solution architects, data scientists, DevOps engineers and cloud security engineers.
Microsoft AI-103 at a glance
| Item | Detail |
|---|---|
| Exam code | AI-103 |
| Time allowed | 120 minutes |
| Passing score | 700 or greater |
| Exam fee | US$165 in the US, priced by the country or region where you sit it |
| Languages | English, Chinese (Simplified), Chinese (Traditional), French, German, Japanese, Korean, Italian, Portuguese (Brazil), Spanish |
| Where you sit it | Pearson VUE, proctored |
| Certification valid for | 1 year, renewed with a free online assessment |
Microsoft does not publish a question count for the AI-103. It says most of its certification exams typically contain between 40 and 60 questions, but that the number can vary by exam. The exam may include interactive components.
What is on the exam
These weights come from the skills measured in Microsoft's study guide as of 16 April 2026. Microsoft updates exams periodically, so check the study guide for the version that applies on your exam date.
Implement generative AI and agentic solutions (30 to 35%). The heaviest domain. You build generative applications with Foundry, including retrieval-augmented generation (RAG), multistep reasoning pipelines and evaluation of fabrications, relevance, quality and safety. You also build agents that use retrieval, function calling and conversation memory, implement orchestrated multi-agent solutions, add safeguards and approval flows, and optimise systems through prompt tuning, tracing, token analytics and latency breakdowns.
Plan and manage an Azure AI solution (25 to 30%). Choosing models and Foundry services for each task, designing infrastructure and deployments, and connecting Foundry projects to CI/CD pipelines. It also covers quotas, rate limits and cost, monitoring for drift and safety events, security with managed identity, private networking and role policies, and responsible AI through safety filters, evaluators, audit trails and tool-access controls.
Implement computer vision solutions (10 to 15%). Generating and editing images and video from prompts, building multimodal understanding workflows such as captions, alt text and visually grounded question answering, using Content Understanding for video analysis, and filtering unsafe visual content including indirect prompt injection hidden in images.
Implement text analysis solutions (10 to 15%). Extracting entities, topics, summaries and structured JSON with generative prompting, detecting sentiment and sensitive content, translation, and speech solutions including speech to text, text to speech and custom speech models.
Implement information extraction solutions (10 to 15%). Building ingestion and indexing pipelines with semantic, hybrid and vector search, enrichment skills, OCR, and extracting content from documents with Content Understanding analyzers that produce clean output for agents and RAG.
Why people fail it
The first problem is breadth. Two domains carry up to 65% of the marks between them, but the other three each carry 10 to 15% and cover separate topics: vision, language and speech, and extraction. Candidates who have only built chat or RAG apps can lose marks across the lighter domains in aggregate.
The second is treating agents as prompts. The blueprint asks about tool schemas, conversation tracking, memory, orchestration across multiple agents and approval flow controls. If you have not defined an agent with tools and traced what it did, those objectives are guesswork.
The third is the operational side. Quotas, rate limits, managed identity, private networking, drift monitoring and CI/CD are in the heaviest planning domain, and they are easy to skip when you are focused on model behaviour. Microsoft also says the exam may contain questions on Preview features if they are commonly used, so read current Foundry documentation rather than older Azure AI material.
A study plan that fits the exam
Seven weeks suits an experienced Python developer. The time is split in line with the domain weights, and the Foundry material comes first because the planning and building domains are worth up to 65%.
- Week 1: Foundry services, models and deployment. Choose models for different tasks, deploy one, and connect an app to a Foundry project. Take the free AI-103 practice questions to find your gaps.
- Week 2: managing, securing and governing. Work through quotas and rate limits, managed identity, private networking, monitoring, safety filters and evaluators.
- Week 3: generative apps and RAG. Build a RAG application, evaluate it for relevance and fabrication, and tune prompts and model parameters.
- Week 4: agents. Build an agent with tools and memory, then a multi-agent workflow with an approval step, and read its traces.
- Week 5: vision, text and speech. Generate and edit an image, caption images with a multimodal model, extract structured JSON from text and try speech to text and translation.
- Week 6: information extraction. Build an index with hybrid and vector search, add enrichment skills and OCR, and try a Content Understanding analyzer on documents.
- Week 7: timed runs. Work through the questions-only copy in the AI-103 practice question pack in 120 minute blocks and read the explanation for every miss.
On exam day
You schedule through Pearson VUE and the exam is proctored. Microsoft strongly recommends registering with a personal Microsoft account, because exam records tied to a work or school account are lost if you leave that organisation. You get 120 minutes. Microsoft's duration page lists 140 minutes of seat time for associate and expert exams that may contain labs, which covers instructions, the candidate agreement and comments afterwards.
Five minutes for breaks are built into the exam time, but the clock keeps running while you are on a break. Microsoft provides an exam sandbox where you can try the question types and interface before you sit it.
Frequently asked questions
What happens if I fail the AI-103?
You can retake it after 24 hours. After that, Microsoft imposes a 14 day wait between subsequent attempts, and you cannot take a given exam more than five times in the 12 months from your first attempt. Check the retake policy on Microsoft Learn before you book.
Is the practice question pack enough on its own?
No. It is practice, not a course. The 187 real exam questions with explanations show how Microsoft frames decisions about Foundry, agents and retrieval, and where your gaps are, but the AI-103 rewards hands-on time building with Python. Build the solutions yourself, then use the questions to check yourself and rehearse under time.
How long does the certification last?
Microsoft's study guide says associate, expert and specialty certifications expire annually, and that you can renew by passing a free online assessment on Microsoft Learn.
What about the older Azure AI Engineer Associate certification?
Microsoft Learn marks the Azure AI Engineer Associate certification and its renewal assessment as retired. Study for the AI-103 from its own study guide, because the skills measured are built around Microsoft Foundry, agents and Python.
Try the free questions first, then get all 187 AI-103 practice questions with full explanations for US$39, pass or your money back: get the AI-103 pack.
Sources
The official pages the facts in this guide were checked against.
- learn.microsoft.com/en-us/credentials/certifications/resources/study-guides/ai-103
- learn.microsoft.com/en-us/credentials/certifications/azure-ai-apps-and-agents-developer-associate/
- learn.microsoft.com/en-us/credentials/support/exam-duration-exam-experience
- learn.microsoft.com/en-us/credentials/support/retake-policy
- learn.microsoft.com/en-us/credentials/certifications/azure-ai-engineer/
- learn.microsoft.com/en-us/credentials/exam-pricing.json
