What is the difference between AI-103 and AI-200?

AI-103 (Azure AI Apps and Agents Developer Associate) certifies building AI applications and agents in Microsoft Foundry, with generative AI and agentic solutions its heaviest domain at 30-35%. AI-200 (Azure AI Cloud Developer Associate) certifies the back-end side: containerised solutions, Azure data management services, connecting to Azure services, and securing and monitoring solutions. Both are associate level, 120 minutes, and pass at a scaled 700.

Does AI-103 count towards AB-100?

Yes. To earn Microsoft Certified: Agentic AI Business Solutions Architect Expert you must pass AB-100 and hold at least one associate certification from Microsoft's list, and AI-103 is on it. AI-300, AB-620 and several Dynamics 365 and Power Platform associates are also accepted.

Should I take AB-100 or AI-103?

They are different levels rather than alternatives. AB-100 is an expert-level architect exam that assumes hands-on experience and requires a listed associate certification to award the credential. AI-103 is an associate developer exam and is one of the certifications that meets that requirement, so for most people it comes first.

Only for Microsoft Learn. AI-200 is a role-based associate exam, so like AI-103 it lets you open learn.microsoft.com in a split screen during the exam. Other websites are blocked and no extra time is added.

Know if you're actually ready. Take the Azure AI-103 quiz → get your AI readiness report.Take the free test →

Study guide · Azure AI-103

AI-103 vs AI-200 vs AB-100: Which Microsoft AI Certification?

Three Microsoft AI certifications that sound interchangeable and are not. Two are associate exams for different developers; the third is an expert credential that AI-103 can count towards.

Examifyr·2026·7 min read

The short answer

AI-103 is for the developer who builds AI applications and agents in Microsoft Foundry. AI-200 is for the developer who builds the back end those AI solutions run on — containers, data services, messaging, monitoring. AB-100 is not a third option alongside them: it is the exam for an expert-level architect credential, and an associate certification such as AI-103 is one of the ways to meet its prerequisite. So the real choice is between AI-103 and AI-200, and AB-100 is where either kind of experience can lead later.

                  AI-103                 AI-200                    AB-100
Credential ...... Azure AI Apps and      Azure AI Cloud            Agentic AI Business
                  Agents Developer       Developer Associate       Solutions Architect
                  Associate                                        Expert
Level ........... Associate              Associate                 Expert
Builds .......... agents, AI apps        back-end services for     architecture across
                  in Foundry             AI solutions              Dynamics 365, Power
                                                                   Platform, Copilot
                                                                   Studio, Azure AI
Duration ........ 120 minutes            120 minutes               not stated on the
                                                                   exam page
Passing score ... 700 (scaled)           700 (scaled)              700 (scaled)
Prerequisite .... none                   none                      one listed associate
                                                                   cert (AI-103 counts)

NoteEvery figure here is from the Microsoft Learn certification and exam pages, checked 14 October 2026. Microsoft scores all three on a scale where 700 passes; that is not a percentage of questions answered correctly.

AI-103 vs AI-200: which developer are you?

Read the two audience profiles side by side and the split is clean. AI-103 expects Python and familiarity with generative AI and Azure services, and its heaviest domain is implementing generative AI and agentic solutions at 30-35%, alongside planning and managing an Azure AI solution and the vision, text analysis and information extraction workloads. AI-200 expects proficiency in Azure SDKs, Azure data management services, monitoring and troubleshooting, messaging and eventing, vector databases, Python and containerised applications, and it assesses four areas: developing containerised solutions on Azure, developing AI solutions by using Azure data management services, connecting to and consuming Azure services, and securing, monitoring and troubleshooting Azure solutions. If your week is prompts, agents, tools and evaluation, that is AI-103. If it is services, queues, data stores and deployments that an AI feature depends on, that is AI-200.

NoteMicrosoft lists AI-200's four assessed areas without weights on the certification page, so we do not quote any. Its practice assessment was not yet available when we checked — Microsoft says these usually follow within eight weeks of an exam leaving beta.

Where AB-100 fits

AB-100 leads to Microsoft Certified: Agentic AI Business Solutions Architect Expert. Passing it is not enough on its own: you must also hold at least one associate certification from a list Microsoft publishes, and AI-103 is on that list — as are AI-300, AB-620 and a range of Dynamics 365 and Power Platform associates. The exam is pitched at an accomplished solution architect who designs agentic and multi-agent solutions across Dynamics 365, Power Platform, Copilot Studio and Azure AI, and works with open protocols including Agent2Agent and Model Context Protocol. Its assessed areas are planning (25–30%), designing (25–30%) and deploying (40–45%) AI-powered business solutions.

NoteMicrosoft updated the English version of AB-100 on 14 October 2026. Check the study guide linked from the exam page for the current skills list before you start preparing.

Both associate exams are open book — for Microsoft Learn only

AI-103 and AI-200 are role-based exams, so you can open Microsoft Learn in a split screen during the exam and browse anything on learn.microsoft.com except Q&A, practice assessments and your profile. Other websites are blocked and the clock does not stop. That changes how to prepare: memorising an SDK signature matters less than knowing which service solves the problem in front of you, because the lookup is possible and the time to do many lookups is not.

How to choose

Pick the associate exam that matches the work you do now, not the title that sounds most senior. If you are aiming for the architect credential eventually, AI-103 is a natural route because it both proves the hands-on agent skills AB-100 assumes and satisfies its prerequisite. If you are unsure whether you are ready for AI-103 at all, the free readiness test below is weighted to its published blueprint and will tell you domain by domain.

Exam tipDo not book AB-100 first because it sounds like the bigger credential. Without one of the listed associate certifications you cannot be awarded the expert certification even after passing the exam.