Hello, you are using an old browser that's unsafe and no longer supported. Please consider updating your browser to a newer version, or downloading a modern browser.
The Microsoft Certified: Azure AI Apps and Agents Developer Associate is Microsoft's certification for Python developers who build generative AI apps and agents on Microsoft Foundry. This October 2026 guide covers the AI-103 exam, the five domains and weights in the April 16, 2026 outline, the 120-minute format and 700 passing score, the $165 fee, the free annual renewal, how it replaced AI-102, and the roles it leads to.
The Microsoft Certified: Azure AI Apps and Agents Developer Associate, earned by passing exam AI-103, is Microsoft's role-based certification for developers who build, deploy and manage generative AI apps and agents on Microsoft Foundry.
The exam's full title is Developing AI Apps and Agents on Azure. You pick the model and Foundry services for a task, build retrieval-augmented generation (RAG) into an app, give agents tools, memory and approval controls, and orchestrate several agents together. The outline also reaches into image and video generation, speech, translation, document extraction and the search indexes that ground an agent's answers. Microsoft lists it as the replacement for AI-102, the Azure AI Engineer Associate exam, which retired on June 30, 2026.
AI-103 carries no formal prerequisites, but Microsoft expects experience developing apps in Python and familiarity with generative AI and Azure services. It is issued and maintained by Microsoft. The exam opened as a beta in April 2026, and the current skills-measured outline is dated April 16, 2026.
Foundry services and models, deployments, cost, security and responsible AI. 25 to 30%.
RAG, agents with tools and memory, multi-agent orchestration and tracing. The largest domain at 30 to 35%.
Image and video generation, multimodal understanding and Content Understanding. 10 to 15%.
Entities, sentiment, translation, and speech as an agent modality. 10 to 15%.
Ingestion, semantic, hybrid and vector search, and document extraction. 10 to 15%.
Four things that set the Azure AI Apps and Agents Developer apart from the exam it replaced in June 2026.
AI-102 split its outline into six skill areas, with generative AI and agentic solutions listed separately. AI-103 folds them into one domain worth 30 to 35 percent, the largest on the exam, and planning and managing the solution in Foundry adds another 25 to 30. Vision, text and extraction share what's left at 10 to 15 percent each.
The AI-102 page named Python and C#. AI-103 asks for experience developing apps in Python, and the exam runs 120 minutes instead of AI-102's 100. The objectives read as build tasks, such as connecting an app to a Foundry project or defining an agent's tool schemas, so working code helps more than flashcards.
AI-103 is the prerequisite for the new AI-500 Multi-Agent AI Solutions Expert, and it sits beside AI-200 and AI-300 at the associate tier. Our look at how Microsoft certification evolved covers the other 2026 replacements.
AI-103 is valid for one year and renews free through an unproctored, open-book assessment on Microsoft Learn. The renewal window opens six months before your expiration date, and there is no limit on renewal attempts. Microsoft doesn't claim a DoD 8140 mapping for AI-103. The current matrix is published at the DoD Cyber Exchange.
Employers judge it on the work. Teams putting RAG, agents and multimodal features into production on Azure need developers who can build them in Foundry, and AI-103 is the Microsoft credential written for that job. AI-102 holders have a clock running too, because a retired certification can't be renewed.
The certification, the AI-103 exam, and how to keep it current, as of October 2026.
AI-103 is the associate developer credential in Microsoft's AI lineup. From here you can go expert on multi-agent design, move into AI operations, or add the back-end developer exam. The three below are the common next steps.
The expert step. AI-500 tests architecting, building, evaluating and securing multi-agent systems, and Microsoft expects familiarity with Microsoft Agent Framework, Model Context Protocol (MCP) and LangGraph. You must hold the Azure AI Apps and Agents Developer Associate to earn it, so AI-103 is the direct on-ramp.
The operations move. AI-300 covers MLOps and GenAIOps on Azure Machine Learning and Microsoft Foundry, with GitHub Actions pipelines and Bicep along the way. The evaluation, tracing and monitoring objectives in AI-103 are a natural lead-in.
The back-end partner. AI-200 is Microsoft's listed replacement for AZ-204, which retired July 31, 2026, and it covers containerized solutions, Azure data management services, vector databases and messaging. Pair it with AI-103 if you own both the agent and the services it calls.
AI-103 sits at the associate tier of Microsoft's AI path. AI-901 covers the concepts underneath it, and from here the path branches toward multi-agent design, AI operations and cloud development.
Optional groundwork
The AI developer credential
Pick your path
Two questions to answer before you commit: can you sit it, and should you pursue AI-103 specifically. Here's a straight answer to both.
No prerequisite credential required.
AI-103 has no formal prerequisites, and AI-901 is optional. You register through Pearson VUE and test at a center or online with a proctor. Microsoft advises booking with a personal Microsoft account rather than a work or school account, because exam records tied to an organizational account are lost for good if you leave that organization.
The outline assumes you write code.
Microsoft expects experience developing apps in Python, and the objectives are written as build tasks: deploy a model, wire retrieval into an app, give an agent tools and memory, add an approval step. Build a small RAG app and a tool-calling agent in your own Foundry project before you book. Reading about agents tends to stall on scenario questions about them.
AI-103 maps to the developer and engineer roles that build AI into applications, from RAG chat tools to agents that call APIs and pause for a human to approve.
The role Microsoft writes AI-103 around. Builds, manages and deploys AI apps and agents on Microsoft Foundry, working with solution architects, data scientists, DevOps engineers and cloud security engineers along the way.
Builds applications on large and small language models, adds retrieval-augmented generation so answers come from the company's own content, and evaluates output for fabrications, relevance and safety before it ships.
Defines agent roles, goals and tool schemas, connects agents to APIs, search and memory, and orchestrates several agents with approval steps. That work sits in the largest AI-103 domain, worth 30 to 35 percent.
The Python developer whose product team has started shipping AI features. AI-103 covers connecting an existing application to a Foundry project through the SDKs and connectors, plus the quota and rate-limit controls that keep the bill in check.
Builds voice and chat experiences, with speech to text, text to speech, custom speech models and translation feeding an agent. The text analysis domain covers each of those, along with sentiment and tone detection.
Owns the ingestion and indexing behind grounded answers, from semantic, hybrid and vector search to OCR, layout extraction and Content Understanding analyzers. A natural fit for developers who came from search or document processing work.
AI-103 replaced AI-102 and sits beside AI-200 at the associate tier, with AI-901 underneath both. Here's how the four line up.
| AI Apps and Agents Developer (AI-103) | Azure AI Engineer (AI-102) | Azure AI Cloud Developer (AI-200) | Azure AI Fundamentals (AI-901) | |
|---|---|---|---|---|
| Issuer | Microsoft | Microsoft | Microsoft | Microsoft |
| Status | Current (replaced AI-102) | Retired June 30, 2026 | Current (replaced AZ-204) | Current (replaced AI-900) |
| Level | Associate | Associate | Associate | Fundamentals |
| Focus | AI apps and agents in Microsoft Foundry | Azure AI services, AI Search and Azure OpenAI | Back-end services for AI apps | AI concepts and Foundry basics |
| Coding Expected | Python | Python or C# | Python | Basic Python syntax |
| Skill Areas | 5 | 6 | 4 | 2 |
| Exam Time | 120 min | 100 min | 120 min | 45 min |
| Passing Score | 700 of 1000 | 700 of 1000 | 700 of 1000 | 700 of 1000 |
| Cost (US) | $165 | No longer offered | $165 | $99 |
| Validity | 1 year, free renewal | Lapses at expiry, no renewal | 1 year, free renewal | Does not expire |
| Best For | Python developers building AI apps and agents | Existing holders only | Cloud and back-end developers | Newcomers to AI development |
Costs are US prices as of October 2026 and vary by region and taxes. Microsoft claims no DoD 8140 mapping for any of the four. Fundamentals certifications don't expire, associate credentials renew free each year, and a retired AI-102 stays on your transcript but can't be renewed.
Certification Matchups: the published facts, strengths and weak spots of 33 certifications, side by side.
As an authorized Microsoft training partner, our Azure AI Apps and Agents Developer boot camp covers all five AI-103 domains in four days of live, instructor-led training, in person or online. You work from official Microsoft courseware in hosted Foundry projects with labs every day, practice on questions mapped to the five domains, and keep instructor access for 180 days after class. Your AI-103 exam voucher and an exam retake are included.
How the 2026 Microsoft retirements played out, where AI-103 sits among AI certifications, and the agent security risks behind its guardrail objectives.
Which Microsoft exams people are actually sitting in 2026, why AI-103 and AI-901 are climbing fast, and when a developer should pick AI-200 or AI-103 first.
AI-103 next to the AWS, Google and NVIDIA build-side credentials, and why it is more than a renumbered AI-102 now that agents in Microsoft Foundry are the focus.
From product exams to job roles to Foundry, and how the 2026 replacements, AI-103 for AI-102 and AI-200 for AZ-204 among them, fit the longer pattern.
The full Microsoft map by role and level, refreshed after the 2026 retirements, with AI-103 at the associate tier of the AI and data track.
The ten security risks of agents that plan, remember and call tools. Useful background for the AI-103 objectives on guardrails, tool-access controls and prompt injection.
Why a scaled passing score of 700 on a Microsoft exam does not equal 70 percent, and how to read practice results before you book the real thing.
ID, breaks, online proctoring and retake rules compared across vendors, including Microsoft's waiver process when a Pearson VUE session fails.
The AI-103 exam is organized into five domains. Generative AI and agentic solutions is the heaviest at 30 to 35 percent, and planning and managing the solution adds another 25 to 30. Click any domain for what it covers.
Choosing the right model for a task (large and small language models, multimodal models and Foundry Tools), the Foundry services for grounding, vector search and agent workflows, and the memory, tool and knowledge services an agent needs. It also covers infrastructure and deployment options, CI/CD for Foundry projects, quotas, rate limits and cost, monitoring for drift, safety events and grounding quality, managed identity, private networking and keyless credentials, and responsible AI controls such as guardrails, evaluators, trace logging, approval workflows and tool-access limits.
The heaviest domain. Deploying and consuming language, code and multimodal models, building retrieval-augmented generation (RAG) into an app, multistep reasoning pipelines, and evaluating output for fabrications, relevance and safety. On the agent side you define roles, goals and tool schemas, add retrieval, function calling and conversation memory, orchestrate multi-agent solutions with approval flow controls, and then tune prompts and parameters, add self-critique loops, and set up tracing, token analytics and latency breakdowns.
Generating images and videos from text prompts and reference media, inpainting and mask-based edits, captions and alt-text for accessibility, and question answering grounded in visual evidence. Azure Content Understanding in Foundry Tools handles visual extraction and video analysis in single-task and pro-mode pipelines. The responsible AI piece covers filtering unsafe images, catching indirect prompt injection hidden in image text, and enforcing visual policy rules such as watermarks and brand usage.
Extracting entities, topics, summaries and structured JSON with generative prompting and Foundry Tools, detecting sentiment, tone and sensitive content, translating text with Azure Translator in Foundry Tools or an LLM-powered flow, and customizing output for domain tasks like compliance summarization. Speech sits here too, with speech to text and text to speech for agents, custom speech models, reasoning over audio input and speech translation.
Ingesting and indexing documents, images, audio and video, configuring semantic, hybrid and vector search for grounding, enrichment with built-in or custom skills, and RAG ingestion that uses OCR. You also connect retrieval pipelines directly to workflows and agent tools, and use Content Understanding analyzers to turn documents into clean structured or markdown output an agent can reason over.
Domains and weights reflect the Microsoft AI-103 skills-measured outline dated April 16, 2026. Microsoft notes the exam may include Preview features that are commonly used, and it refreshes role-based exams periodically, so confirm the current outline on Microsoft Learn before scheduling.
The questions candidates ask most often when researching the Microsoft Azure AI Apps and Agents Developer certification.
The Microsoft Certified: Azure AI Apps and Agents Developer Associate, earned by passing exam AI-103, is Microsoft's role-based certification for developers who build, deploy and manage generative AI apps and agents on Microsoft Foundry. It covers planning and securing an AI solution, generative AI and agents, computer vision, text and speech analysis, and information extraction. It replaced the Azure AI Engineer Associate (AI-102), which retired on June 30, 2026.
The current exam is AI-103: Developing AI Apps and Agents on Azure, on the skills-measured outline dated April 16, 2026. It opened as a beta in April 2026, and Microsoft's announcement said it was expected to go live in June 2026. As of October 2026, the Microsoft Learn certification page and study guide no longer label it beta. Microsoft updates the English version first and localized versions about eight weeks later.
Yes. Microsoft lists the Azure AI Apps and Agents Developer Associate (AI-103) as the replacement for the Azure AI Engineer Associate (AI-102). The AI-102 exam, the certification and its renewal assessment all retired on June 30, 2026. If you earned AI-102, it stays on your Microsoft Learn transcript, but it can no longer be renewed, so it lapses at its expiration date. Microsoft's retirement announcement describes no upgrade or transition exam, so a current Microsoft AI developer credential now means passing AI-103.
No. AI-103 has no formal prerequisites, so anyone can register and sit the exam. Microsoft expects experience developing apps in Python and familiarity with general AI, generative AI and Azure services. AI-901 Azure AI Fundamentals covers the concepts underneath it, but it isn't required.
The five domains are Plan and manage an Azure AI solution (25 to 30 percent), Implement generative AI and agentic solutions (30 to 35 percent), Implement computer vision solutions (10 to 15 percent), Implement text analysis solutions (10 to 15 percent), and Implement information extraction solutions (10 to 15 percent). Generative AI and agents carry the most weight.
Microsoft doesn't publish a question count for AI-103. Its general guidance is that most Microsoft certification exams have 40 to 60 questions, with the number varying by exam. You get 120 minutes, the length Microsoft gives role-based exams that may contain labs, and Microsoft says to plan on 140 minutes of seat time. The exam is proctored and may include interactive components.
You need 700 on a scale of 1 to 1,000. Microsoft uses scaled scoring, so 700 doesn't mean 70 percent of the questions. Microsoft sets the passing line against question difficulty, so an easier set of questions takes more points to pass than a harder one.
The AI-103 exam costs $165 USD in the United States as of October 2026. Microsoft sets the price by the country where you test, and local taxes can apply. You schedule it through Pearson VUE, at a test center or online with a proctor. Training Camp's Azure AI Apps and Agents Developer boot camp includes the exam voucher in the program fee, along with an exam retake.
No DoD 8140 mapping is claimed for it. Microsoft doesn't list AI-103 against a DoD 8140 work role, so don't count on it to satisfy a cyber-coded position requirement on its own. What it does show an employer is that you can build AI apps and agents in Microsoft Foundry. See the full DoD 8140 work role paths.
AI-103 is valid for one year. You renew it free by passing an unproctored, open-book online assessment on Microsoft Learn before your certification expires. The assessment opens six months before your expiration date, and there's no limit on attempts, although after the second try you wait at least 24 hours between attempts.
AI-103 maps to roles such as AI Engineer, Generative AI Developer, AI Agent Developer, Conversational AI Developer and Search and Retrieval Engineer, along with software developers adding AI features to existing applications. Microsoft tags the certification for the AI Engineer and Developer job roles.
Take the one that matches the code you write. AI-103 is about the AI layer itself, meaning models, RAG, agents, vision, speech and document extraction in Microsoft Foundry. AI-200, the Azure AI Cloud Developer Associate that replaced AZ-204, covers the back end those apps run on, such as containers, Azure data services, vector databases and messaging. Both are 120-minute associate exams at $165, so the choice comes down to which half of the app you build.
Most of the difficulty is range. AI-103 runs from agent orchestration and safety guardrails to image generation, speech, translation and OCR pipelines, and it assumes you write Python. Microsoft also notes that the exam may cover Preview features if they're commonly used, so hands-on time in Foundry counts for more than memorizing service names. You can open Microsoft Learn during the exam, but the clock keeps running, so use it to confirm a parameter or setting, not to learn a service from scratch.
For Python developers building AI features on Azure, yes. It's the Microsoft credential written around building apps and agents in Foundry, it costs $165 with a free annual renewal, and it's the required prerequisite for AI-500 Multi-Agent AI Solutions Expert. It's less useful if you don't write code, where AI-901 is the lighter fit, or if your development work is mostly back-end services, where AI-200 is closer.
Whether you're moving off AI-102, weighing AI-103 against AI-200, or arranging AI developer training for a team, tell us where you are and we'll help you map out the right path.