How to Pass AI-103: A 3-Week Study Plan for Microsoft’s Azure AI Apps and Agents Exam
Give AI-103 about three weeks if you already work with Azure and Python, or six to eight if you don’t. Put two of those weeks into the two biggest domains, build in Microsoft Foundry instead of only reading about it, and let timed practice tell you when you’re ready.
The exam went into beta in April 2026 and is now generally available, so most prep advice for it is thin or recycled from AI-102. That’s your advantage: the skills outline and domain weights are public, and almost nobody else studying has a structured plan. Here’s one, built directly from Microsoft’s blueprint, plus an honest answer to how hard AI-103 is and a one-screen cheat sheet of the published outline.
AI-103 at a glance
| Certification | Microsoft Certified: Azure AI Apps and Agents Developer Associate |
| Level | Intermediate (associate) |
| Exam length | 120 minutes, proctored. It may include interactive components. |
| Cost | $165 in the US. The price depends on the country where you sit it (£106 in the UK, ₹4,865 in India). Full cost breakdown |
| Passing score | 700 or higher (scaled) |
| Where | Pearson VUE, online or at a test center |
| Open book? | Partly. Microsoft Learn is available inside the exam, but the clock keeps running. |
| Retakes | 24 hours after a first attempt, then 14 days between attempts; at most 5 attempts in 12 months, each one paid |
| Renewal | Free online renewal assessment on Microsoft Learn every year |
| Prerequisites | None formal. Microsoft expects experience building apps in Python. |
| Replaces | AI-102, retired June 30, 2026 (what changed) |
Checked against Microsoft Learn’s AI-103 certification page, study guide, exam pricing and retake policy on October 8, 2026.
Get an email when the Azure AI Apps and Agents Developer Associate (AI-103) exam changes
We check the official exam pages every day. If the price, version, format or retirement date changes, you’ll get a short email. At most one a week, plus the HOW TO // AI newsletter on Thursdays.
Unsubscribe anytime. Privacy policy.
Try it now, without leaving this page: 10 original AI-103 practice questions written from Microsoft’s public exam guide, each with a plain-English explanation. No sign-up.
FREE PRACTICE QUIZ
Try 10 free AI-103 practice questions
Answer at your own pace. Every question shows the correct answer and a plain-English explanation, and that's where the learning happens. Unlock 10 more questions free at the end.
What you’re up against
| Domain | Weight |
|---|---|
| Implement generative AI and agentic solutions | 30–35% |
| Plan and manage an Azure AI solution | 25–30% |
| Implement computer vision solutions | 10–15% |
| Implement text analysis solutions | 10–15% |
| Implement information extraction solutions | 10–15% |
The language of the exam is Python, and the platform at its center is Microsoft Foundry. Read the weights like a study budget: roughly two-thirds of your points live in the top two domains, so that’s where two of your three weeks go.
Is AI-103 hard?
Honestly: it’s a real associate-level exam, not a vocabulary test, but it’s very passable with hands-on prep. Microsoft rates it Intermediate and doesn’t publish pass rates, so treat any pass rate you see quoted with suspicion. What makes it hard:
- You need Python. Microsoft’s audience profile expects experience developing apps with Python.
- It’s broad. The outline spans five domains and 14 skill groups, from agent orchestration and retrieval pipelines to image and video generation and speech.
- It tests doing, not describing. The exam may include interactive components, and questions can cover Preview features when they’re commonly used.
What makes it easier: Microsoft Learn is open during the exam (the clock keeps running), Microsoft’s official practice assessment is free, and the pass mark is a scaled 700. If your background is classic Azure AI services rather than agents, expect most of your new study to land there, since generative AI and agentic solutions alone are 30–35% of the exam. Weighing whether it’s worth the effort? See is AI-103 worth it?
How long to study
- You build on Azure and write Python: about three focused weeks, the plan below.
- You write Python but are new to Azure: four to six weeks. Add a week of Azure basics (subscriptions, resource groups, keys and identity) before Week 1.
- Neither yet: six to eight weeks. Learn basic Python first, and consider the fundamentals exam, AI-901, as a warm-up.
In hours, the three-week plan below assumes roughly 8 to 10 hours a week, or about 25 to 30 hours in total. Most of that should be building in Foundry and taking timed practice, not watching videos.
Week 1: Foundations and the “plan and manage” domain
- Set up a real environment: an Azure subscription and a Microsoft Foundry project. Every hour spent hands-on beats two hours of videos.
- Work the 25–30% domain: choosing and provisioning the right Azure AI services, securing them, responsible AI practices, monitoring, and cost management.
- End the week with a free AI-103 sample set to baseline yourself. Score honestly and note the domains that hurt.
Week 2: The big one, generative AI and agents
- This is a third of your exam. Build, don’t just read: a small generative AI app in Foundry, covering model selection, grounding with your own data, orchestration and deployment.
- Then make it agentic: tool use, multi-step tasks, the agent patterns Foundry supports. If you can explain when an agent beats a plain completion call, you’re thinking like the exam.
- Fold in the two language domains (text analysis, information extraction) late in the week. They’re smaller, and they reuse concepts you’ll have just practiced.
Week 3: Vision, mocks and readiness
- Close the computer vision domain (10–15%): image and video generation and editing, multimodal understanding of images and video (including Azure Content Understanding), and responsible AI controls for visual content.
- Shift to timed mock exams and let a readiness score tell you when you’re consistently clearing the bar. Our AI-103 practice platform tracks this per domain, and Microsoft’s own free practice assessment is on AI Skills Navigator (sign-in required).
- Final days: re-drill only your weakest domain. Resist the urge to re-study what you already pass.
AI-103 cheat sheet: the published outline on one screen
This is Microsoft’s skills outline (as of April 16, 2026), condensed. It tells you what the exam measures; it is not exam content. Print it, and tick off each skill group once you’ve built something that uses it.
| Domain (weight) | Skill groups Microsoft lists |
|---|---|
| Plan and manage an Azure AI solution (25–30%) | Choose the right Foundry services and models; set up AI solutions in Foundry (deployments, CI/CD); manage, monitor and secure AI systems (quotas, cost, managed identity, private networking); implement responsible AI (safety filters, evaluations, tracing, agent oversight) |
| Implement generative AI and agentic solutions (30–35%) | Build generative apps (RAG, multistep flows, evaluation, Foundry SDKs); build agents (tools, memory, multi-agent orchestration, approval flows); optimize and operationalize (prompt tuning, tracing, token analytics, latency) |
| Implement computer vision solutions (10–15%) | Image and video generation and editing; multimodal understanding, including Azure Content Understanding; responsible AI for visual content, including prompt injection hidden in images |
| Implement text analysis solutions (10–15%) | Language model text analysis (entities, summaries, JSON output, sentiment, translation); speech solutions (speech to text, text to speech, speech as an agent modality) |
| Implement information extraction solutions (10–15%) | Retrieval and grounding pipelines (ingestion, semantic, hybrid and vector search, OCR); extracting content from documents with Content Understanding |
Exam-day numbers to remember:
- 120 minutes, scaled pass mark of 700, question count not published.
- Microsoft Learn is available, but the timer keeps running.
- Speech sits inside the text analysis domain, so don’t skip it.
- If the exam isn’t offered in your language, Microsoft lets you request an extra 30 minutes.
Which AI-103 courses and practice tests to use
- Free from Microsoft: the AI-103 study guide, the exam sandbox (a demo of the interface) and the official practice assessment on AI Skills Navigator, which needs a sign-in.
- Free from us: AI-103 practice questions with explanations and the AI-103 study guide, both mapped to the published outline.
- Paid video courses (Pluralsight, Coursera, Udemy and others): check that the course is built for AI-103 and the April 16, 2026 outline, not an AI-102 course with a new title.
Exam day: what to expect
- Microsoft Learn is open, but it costs you time. Associate exams let you open learn.microsoft.com in a split screen. No extra time is added and the timer keeps running, so use it to check one detail, not to answer every question. Q&A, practice assessments and your profile are blocked.
- Breaks lock what you’ve seen. You can take a break without asking, but the clock keeps running and you can’t go back to any question you viewed before the break. Finish and review first.
- Register with a personal Microsoft account. Microsoft recommends it: exam records tied to a work or school account are lost if you leave that organization.
- Try the exam sandbox first. Microsoft’s free demo shows the question types and the interface, so nothing on the day is a surprise.
One honest note on question sources
Because AI-103 is new, “dumps” sites are already selling recycled AI-102 questions with the labels swapped. The two exams measure different things, so that’s worse than studying nothing. Every question in our AI-103 bank is original and written from the published skills outline. Use the study guide as your domain map, and if you were studying for AI-102, see what changed when it retired.
Related guides
- AI-103 Study Guide & Cheat Sheet (Azure AI Apps & Agents Developer)
- Free AI-103 Practice Questions (Azure AI Apps & Agents Developer)
- Is the AI-103 Certification Worth It? An Honest Look at Microsoft’s Newest AI Cert
- AI-102 vs AI-103: AI-102 Is Retired — Here’s What Changed and What to Do
- AI-103 Dumps: Why You Don’t Need Them
Frequently asked questions
How long does it take to prepare for AI-103?
With existing Azure and Python experience, about three focused weeks at roughly 8 to 10 hours a week. Python without Azure, plan four to six weeks. Starting with neither, plan six to eight weeks and build the fundamentals first.
Is AI-103 hard?
It is an intermediate, associate-level exam with real hands-on expectations: Python, Microsoft Foundry and agent patterns across five domains. Microsoft does not publish pass rates; the generative AI and agents domain (30–35%) decides most outcomes, and Microsoft Learn is available during the exam.
How much does the AI-103 exam cost?
$165 in the US. Microsoft sets the price by the country where you take the exam, for example £106 in the UK, €126 in Germany and ₹4,865 in India.
How long is the AI-103 exam?
120 minutes. Microsoft doesn't publish the question count for AI-103; it says most of its certification exams have 40 to 60 questions. You need a scaled score of 700 or higher to pass.
Can you use Microsoft Learn during the AI-103 exam?
Yes. Associate and expert exams let you open Microsoft Learn in a split screen. No extra time is added and the timer keeps running, and Q&A, practice assessments and your profile are blocked.
How soon can I retake AI-103?
You can retake it 24 hours after your first attempt, then you wait 14 days between later attempts, with at most five attempts in 12 months. Each attempt is paid.
Do I need AI-901 before AI-103?
There's no formal prerequisite. But AI-103 assumes working comfort with Azure AI concepts, so fundamentals-level knowledge is effectively expected even though it isn't required.
Are there official Microsoft practice tests for AI-103?
Yes. Microsoft's free official practice assessment for AI-103 is on AI Skills Navigator, and you need to sign in to launch it. Our free AI-103 question bank is also available, written from the published skills outline.
Is there an AI-103 cheat sheet?
Yes: this page condenses Microsoft's published AI-103 skills outline (as of April 16, 2026) into one table of domains, weights and skill groups. Generative AI and agents (30–35%) plus planning and managing Azure AI solutions (25–30%) are about two-thirds of the exam.
