AI-103 Skills Measured: Skill Areas, Weights and Study Notes

Microsoft's AI-103 (Developing AI Apps and Agents on Azure) study guide shows "Skills measured as of April 16, 2026" with 5 skill areas: Plan and manage an Azure AI solution (25 to 30%), Implement generative AI and agentic solutions (30 to 35%), Implement computer vision solutions (10 to 15%), Implement text analysis solutions (10 to 15%), Implement information extraction solutions (10 to 15%). AI-103 leads to Microsoft Certified: Azure AI Apps and Agents Developer Associate.

Which version of the outline we used

When we checked on 17 November 2026, the study guide (last updated on 04/15/2026) showed the skills measured as of April 16, 2026, and we found no announced change after that date. Microsoft says it always updates the English version first and that localised versions follow approximately eight weeks later. Our AI-102 to AI-103: what changed covers how AI-103 differs from the exam it follows.

Every skill area and sub-heading below is copied word for word from the study guide, with a short plain-English note on what it covers. We do not reproduce Microsoft's task lists. To test yourself as you go, try our free AI-103 practice test, and see our round-up Microsoft certifications ranked by difficulty.

Who the AI-103 exam is written for

The audience profile describes an Azure AI engineer who builds, manages and deploys agents and AI solutions that use Microsoft Foundry. Microsoft says you should have experience developing apps with Python and be familiar with the capabilities of general AI, generative AI and Azure services. Microsoft also says most questions cover features that are generally available (GA), and that the exam may contain questions on Preview features if those features are commonly used. That is why it helps to read current Microsoft documentation as well as the outline.

AI-103 exam facts from Microsoft

Detail AI-103 Source
Questions No fixed number published Microsoft exam experience page
Time 120 minutes Microsoft exam page
Passing score 700 or greater, on a scale of 1 to 1,000 Microsoft scoring page
Price (US) 165 USD, varies by country or region Microsoft exam page
Renewal Expires annually, free online renewal Microsoft study guide
Retake 24 hours after the first attempt, then waits vary Microsoft scoring page

Microsoft does not publish a fixed question count for AI-103. Its exam experience page says most Microsoft Certification exams typically contain between 40 and 60 questions, and that the number can vary. Prices can change, so check Microsoft before you book. For how the exams compare, see our IT certification exam facts.

AI-103 skill areas and weights

Microsoft shows each weight as a range. Microsoft does not publish how many questions come from each skill area.

Skill area Weight Sub-headings
Plan and manage an Azure AI solution 25 to 30% 4
Implement generative AI and agentic solutions 30 to 35% 3
Implement computer vision solutions 10 to 15% 3
Implement text analysis solutions 10 to 15% 2
Implement information extraction solutions 10 to 15% 2

Microsoft notes that the bullets under each skill "are intended to illustrate how we are assessing that skill" and that related topics may be covered. Treat the summaries below as a map, not a complete list.

Plan and manage an Azure AI solution (25 to 30%)

4 sub-headings.

Choose the appropriate Foundry services for generative AI and agents

Matching the model and the Foundry service to the job, including retrieval, indexing and what an agent needs to remember.

Set up AI solutions in Foundry

Planning the Azure setup for AI apps and agents, choosing deployment options and connecting Foundry projects to CI/CD.

Manage, monitor, and secure AI systems

Keeping AI workloads healthy and safe: quotas and cost, monitoring quality and drift, and securing access.

Implement responsible AI across generative AI and agentic systems

Safety filters and guardrails, evaluating and auditing AI behaviour, and limiting what agents may do.

Implement generative AI and agentic solutions (30 to 35%)

3 sub-headings.

Build generative applications by using Foundry

Building apps on deployed models, including retrieval-augmented generation (RAG), workflows, evaluation and the Foundry SDKs.

Build agents by using Foundry

Defining and building agents with tools, memory and retrieval, then combining and monitoring them, with safeguards.

Optimize and operationalize generative AI systems

Tuning prompts and model settings and watching a running system through tracing and other observability.

Implement computer vision solutions (10 to 15%)

3 sub-headings.

Design and implement image- and video-generation solutions

Generating and editing images and video from prompts and reference media.

Design and implement multimodal understanding workflows

Having models read images and video: captions, visual question answering, alt text and Content Understanding pipelines.

Implement responsible AI for multimodal content

Applying safety and policy controls to visual content, including attacks hidden in images.

Implement text analysis solutions (10 to 15%)

2 sub-headings.

Apply language model text analysis

Using language models to pull structure out of text, judge tone and safety, and translate.

Implement speech solutions

Speech to text and text to speech as a way for agents and apps to interact.

Implement information extraction solutions (10 to 15%)

2 sub-headings.

Build retrieval and grounding pipelines

Ingesting and indexing content, then searching it so answers are grounded, using semantic, hybrid and vector search.

Extract content from documents

Turning documents into clean, structured content for agents and RAG with OCR, layout analysis and Content Understanding.

Our view: how to use the weights

This section is opinion, not Microsoft's advice. The two biggest ranges are Implement generative AI and agentic solutions (30 to 35%) and Plan and manage an Azure AI solution (25 to 30%), so we would spend the most hands-on time in Foundry building a RAG app and an agent. The three smaller areas (vision, text and information extraction) are 10 to 15% each, but their sub-headings overlap with the larger ones, for example Content Understanding and retrieval pipelines, so skipping them would be a gamble. Microsoft says the exam is written for Python developers, so we would be comfortable reading Python snippets before booking.

AI-103 FAQ

Which version of the AI-103 skills outline applies?

The study guide we checked on 17 November 2026 shows "Skills measured as of April 16, 2026" and was last updated on 04/15/2026.

How many questions are on the AI-103 exam?

Microsoft does not publish a fixed number. It says most Microsoft Certification exams typically contain between 40 and 60 questions, and that the number can vary by exam.

How long is the AI-103 exam?

Microsoft's page says you will have 120 minutes to complete the assessment.

What is the AI-103 passing score?

700 or greater, on a scale of 1 to 1,000. Microsoft notes that this is a scaled score, so it may not equal 70% of the points.

Does the AI-103 certification expire?

Microsoft says associate, expert and specialty certifications expire annually, and you can renew by passing a free online assessment on Microsoft Learn.

How much does the AI-103 exam cost?

Microsoft's page shows $165 USD for the United States. The price is based on the country or region in which the exam is proctored.


Sources: checked on 17 November 2026 on learn.microsoft.com: the AI-103 study guide ("Skills measured as of April 16, 2026", last updated 04/15/2026), the AI-103 exam and certification page, the exam scoring page and the exam duration and experience page. Skill area names, weights and sub-headings are quoted from the study guide and the summaries are our own wording. Microsoft does not publish a fixed question count or the number of questions per skill area. The duration and price come from that page.


Preparing for the exam? See our Microsoft AI-103 practice questions and the free AI-103 practice test.

What people say about this pack

Great resource for AI-103.
Liu Sharma
No complaints, passed AI-103.
ROHAN NAIR
Back to blog