AI-900<\/a> retired on June 30, 2026, but Azure AI Fundamentals did not disappear. Microsoft now uses AI-901 for the current Azure AI Fundamentals exam. The change reflects how quickly the entry-level AI role has moved: candidates are still expected to understand core AI concepts and responsible AI, but the current exam places much more emphasis on Microsoft Foundry, generative and agentic AI, and lightweight implementation skills.<\/p>\n For someone asking what to do after AI-900, there are really two questions. First, what is the current fundamentals credential? That is AI-901. Second, what should come after fundamentals? That depends on whether you want to build AI apps and agents, operate machine-learning systems, design agentic business solutions, or move into AI security and governance.<\/p>\n This change matters for study planning because an old AI-900 objective list can still look plausible while missing the emphasis Microsoft now places on Foundry, generative AI, and agentic workloads. Candidates returning after a break should use AI-900 material to refresh durable concepts such as machine learning, vision, language, and responsible AI, but they should rebuild the final preparation plan around the current AI-901 skills rather than treating the retired exam as a frozen template.<\/p>\n AI-900 is now historical. The old exam measured broad awareness of machine learning, computer vision, natural language processing, responsible AI, and generative AI on Azure. AI-901 keeps the fundamentals role but updates what \u201cfundamentals\u201d means in a Foundry-centered environment.<\/p>\n The current exam still starts with conceptual understanding, yet it also expects basic technical fluency. Candidates should recognize Python syntax, understand Azure resources, and be familiar with REST APIs, SDKs, and command-line tools. That is a noticeable shift from the older assumption that a nontechnical candidate could stay almost entirely at the terminology level. Older material covering AI-900 fundamentals<\/a> can help with foundational vocabulary, but it should not be treated as a current objective map.<\/p>\n AI-901 divides its scope between identifying AI concepts and implementing lightweight AI solutions with Microsoft Foundry. That means candidates need to recognize model types, deployment choices, responsible-AI concerns, and common workloads, but also understand prompts, model deployment, simple client applications, and basic agent creation.<\/p>\n The practical takeaway is that modern AI literacy is no longer only \u201cknow the difference between classification and regression.\u201d It includes understanding how an application interacts with a model, how an agent can use tools, how information is extracted from multimodal content, and why security, safety, and accountability must be designed into the workflow.<\/p>\n The current fundamentals scope also combines previously separate concepts. A single exercise can involve a prompt, a multimodal model, an extraction step, and an agent action. This encourages candidates to see AI as a composed application rather than a menu of isolated services. That systems view becomes essential when you later move to a developer or operations credential.<\/p>\n Responsible AI is still foundational, but it should be connected to implementation. Fairness, reliability, privacy, safety, transparency, inclusiveness, and accountability influence model choice, data handling, human review, logging, and user experience. They are not just definitions to recite at the beginning of a course.<\/p>\n If you already passed AI-900, you usually do not need AI-901 simply to prove that you understand Azure AI basics. The certification name remains Azure AI Fundamentals, and the better use of your time may be to close the technical gaps introduced by the new platform rather than repeating the same credential level.<\/p>\n Use the current AI-901 study guide as a skills checklist. If Foundry, agents, multimodal models, or the lightweight Python expectations are unfamiliar, study those areas directly. The older AI-900<\/a> preparation frame is useful for comparison, but the next step should be driven by the role you want rather than by a desire to replace an old badge with a new code.<\/p>\n A practical signal that AI-103 is the right next step is that you want to write code, connect models to data, use retrieval, call tools, orchestrate agents, and operate an application rather than only explain AI concepts. The transition from fundamentals to developer-level work therefore requires more than memorizing a larger service catalog. It requires basic Python fluency, comfort with APIs and SDKs, and the ability to reason about identity, data flow, evaluation, monitoring, and failure modes inside an AI application.<\/p>\n Microsoft replaced the retired Azure AI Engineer Associate exam AI-102 with AI-103 and the Azure AI Apps and Agents Developer Associate certification. This is the most direct role-based path for developers who want to build with Microsoft Foundry. AI-103 expects Python development experience and covers planning AI solutions, generative and agentic systems, computer vision, text analysis, and information extraction.<\/p>\n The change is important for AI-900 graduates because it shows what \u201cintermediate\u201d now looks like. You move from recognizing workloads to designing deployments, using retrieval and grounding, integrating tools, monitoring agents, applying responsible-AI controls, and operating AI solutions. The retired AI-102<\/a> page can still help explain the previous role, but it should not be mistaken for the current certification target.<\/p>\n Not every AI professional spends the day building end-user AI applications. Some own machine-learning platforms, evaluation pipelines, model deployment, observability, governance, and operational reliability. That is where AI-300<\/a> is more relevant.<\/p>\n Think of the distinction as product engineering versus operational engineering. AI-103 centers on building AI apps and agents. AI-300 centers on the systems and practices that allow models and AI workloads to be released, monitored, governed, and improved in production. There is overlap, but the day-to-day decisions are different.<\/p>\n Some people move from technical implementation into business architecture: deciding where agents belong in a process, how human approvals work, which systems an agent may call, how value is measured, and how governance applies across departments. Microsoft\u2019s AB-100<\/a> path is closer to that role than a developer-focused credential.<\/p>\n This is especially relevant for candidates who entered AI through a fundamentals exam but work in consulting, transformation, automation, or enterprise architecture. The deeper skill is not writing every component yourself. It is designing a governed system that uses AI where it creates measurable value without turning every workflow into an uncontrolled automation experiment.<\/p>\nAI-901 is the current Azure AI Fundamentals exam<\/h3>\n
Foundry, generative AI, and agents now define the fundamentals layer<\/h3>\n
Do not collect another fundamentals credential unless it serves a purpose<\/h3>\n
Move to AI-103 if you want to build AI apps and agents<\/h3>\n
Choose AI-300 when the problem is operationalizing models and AI systems<\/h3>\n
Use AB-100 when AI becomes a business-solution architecture problem<\/h3>\n
Add security and governance before you scale AI adoption<\/h3>\n