Microsoft officially retired the Azure AI Engineer Associate certification and exam AI-102 on June 30, 2026. The move closes a popular credentialing path and forces Windows administrators, developers, and IT generalists to rethink how they prove AI skills. As of mid‑July, any training plan that still lists AI-102 as earnable is out of date.

The news comes just as workplace AI responsibility is being distributed across roles that didn’t exist three years ago. A Microsoft 365 administrator now manages Copilot agents; a senior developer builds retrieval-augmented generation (RAG) pipelines on AWS; a line‑of‑business leader is expected to evaluate generative AI adoption. No single certificate covers all of that, and the one exam that used to bridge Azure AI engineering—AI‑102—is now history.

Here is what changed, why it matters for your next career move, and which credentials you should consider instead.

What Actually Changed

Microsoft retired the Azure AI Engineer Associate certification and its associated exam AI-102: Designing and Implementing a Microsoft Azure AI Solution on June 30, 2026. The credential is no longer available to earn, and the exam cannot be scheduled. The Microsoft Learn pages that once promoted AI-102 now redirect to newer certification paths or general information about the retirement.

For anyone who earned the certification before the cutoff, Microsoft’s transcript will still list it, and its historical relevance remains intact. But for new candidates, the path is closed. The retirement follows a pattern of consolidation as Microsoft aligns its credentials with real‑world job functions rather than broad technology silos.

Meanwhile, new and updated certifications are stepping in to fill the gap:

  • Microsoft 365 Certified: Copilot and Agent Administration Fundamentals (exam AB-900) targets administrators who manage AI within the Microsoft 365 environment. Its English‑language exam will be updated on July 22, 2026, which means preparation materials must align with the revised objectives.
  • AWS Certified Generative AI Developer – Professional (exam AIP-C01), designed for experienced developers with at least two years of production application experience and one year of hands-on generative AI implementation, debuted earlier in 2026. It covers RAG, agents, security, evaluation, and cost optimization on AWS.
  • Google Cloud’s Generative AI Leader certification remains the vendor‑neutral lightweight option for business decision-makers. It does not require technical experience, costs $99, and lasts three years.
  • Microsoft Applied Skills credentials have grown in number and specificity, offering lab‑based proof of capability for targeted scenarios like “Build a copilot with Azure OpenAI Service” or “Administer Microsoft 365 Copilot and agents.”

What It Means for You

The retirement of AI‑102 forces a role‑first reassessment of AI certification plans. The new landscape breaks down into distinct audiences:

For Microsoft 365 Administrators and IT Generalists

If your daily work lives inside the Microsoft 365 admin center—managing data protection, identity, governance, and the rollout of Copilot—the immediate replacement is AB‑900. It’s a fundamentals‑level credential that validates understanding of Microsoft 365, identity, data protection, compliance, and the administrative tasks around Copilot and agents. It is not a development exam; it won’t prove you can train a model or design a vector database. But for the administrator who is suddenly expected to “own” AI governance, it signals relevant readiness.

Timing is critical. The AB‑900 English exam update on July 22 means that any study plan started now must reference the updated exam objectives available on Microsoft Learn. Relying on older guides or practice questions risks missing scope changes. After passing AB‑900, a natural next step is to add a scenario‑specific Applied Skills credential—for example, “Administer Microsoft 365 Copilot and agents”—which proves you can actually perform the tasks in a live environment.

For Developers and Engineers Moving into Generative AI

AI-102 was often the second certification for Azure‑focused developers after they earned Azure AI Fundamentals (AI‑900). With its removal, the nearest equivalent for production AI engineering no longer lives in Microsoft’s catalog unless you qualify for the advanced Azure Solutions Architect Expert or Azure AI Engineer successor paths, which are still evolving. For developers who already have AWS experience or are willing to cross that platform, AIP‑C01 now presents the strongest signal of hands‑on generative AI competence.

AIP‑C01 explicitly requires production experience. AWS states that candidates should have at least two years of application development and one year of hands‑on generative AI implementation. The exam goes deep into RAG architectures, agent design, security governance, model evaluation, monitoring, and cost optimization. It’s not an entry‑level credential. For a junior developer who has only completed tutorials, attempting AIP‑C01 by memorizing AWS documentation will likely fail in both the exam and the job market. But for the experienced engineer already deploying LLM‑based systems, it validates exactly the skills employers need to maintain production AI services.

For Business Leaders and Non-Technical Roles

AI‑102 was never the right exam for this group, yet many managers and consultants felt pressure to earn a technical AI badge. The retirement is a chance to realign with credentials that actually match decision‑making authority. Google Cloud’s Generative AI Leader is explicitly designed for any job role—no hands‑on experience required—and covers the business implications, adoption strategies, and governance of generative AI. At $99 and valid for three years, it’s a focused business‑adoption credential that won’t demand weeks of technical study.

It should not be presented as a technical proficiency signal. A hiring manager evaluating a data scientist or AI engineer will not be impressed by a leadership certificate. But for a product owner, department head, or consultant who needs to speak credibly about AI initiatives, it closes a significant communication gap without pretending to be an engineering exam.

For the MLOps and Platform Operations Specialist

Professionals who own deployment reliability, monitoring, governance, and cost control of AI systems fall into a separate track that no single certification fully covers. AI‑102 touched on some Azure‑specific operations topics, but it wasn’t an MLOps exam. With its retirement, candidates must evaluate credentials based on the platform they actually operate.

For those running AI on AWS, AIP‑C01 includes relevant operations domains—monitoring, security, cost optimization—and may serve as a complementary credential alongside platform‑specific operations certifications. For those on Azure, the Applied Skills path currently offers the most targeted operations proofs (e.g., “Monitor and optimize generative AI solutions on Microsoft Azure”). The key is to begin with the job description and only choose a credential that directly assesses your operational environment.

How We Got Here

The Microsoft AI certification journey began with AI-900: Azure AI Fundamentals in 2020, which remains available and unchanged in retirement discussions. It was followed by the associate‑level AI-102, which quickly became the standard for Azure AI engineering. Over the years, the exam objectives evolved to include Copilot, content safety, and responsible AI, but the underlying role—building and implementing AI solutions on Azure—stayed consistent.

In parallel, the industry’s definition of “AI engineer” fractured. The arrival of large language models and generative AI made it clear that a single associate exam couldn’t cover the breadth now required: from model deployment to prompt engineering, agent orchestration, and production governance. Microsoft responded by splitting the path: one for administrators inside Microsoft 365 (AB‑900) and one for platform‑specific builders. The Applied Skills program, launched in 2023, filled the gap for scenario‑based proof without a full certification.

AWS and Google also accelerated their credentials. AWS released its Machine Learning Engineer Associate in 2025 and Generative AI Developer – Professional (AIP‑C01) soon after. Google added the Generative AI Leader certification in early 2026. The result is a market where “AI certification” no longer means a single ladder from fundamentals to associate to expert, but a constellation of role‑ and platform‑specific options.

The retirement of AI-102 on June 30, 2026, is thus the culmination of a multi‑year shift. It isn’t just a product sunset; it’s a signal that the era of a general Azure AI engineer credential is over. The same forces also explain why AB‑900’s exam will be updated on July 22—Microsoft is refining the Copilot administration scope in response to early feedback.

What to Do Now

Your next step depends entirely on your current role and next career goal. Here is a decision sequence that starts with your daily work:

  1. If you administer Microsoft 365 and Copilot or are being asked to govern AI in that environment:
    - Register for the AB‑900 exam. Begin studying immediately using the updated study guide that Microsoft will publish for the July 22 revision. Don’t rely on old practice tests.
    - After AB‑900, add an Applied Skills lab in a task you perform regularly—for example, “Administer Microsoft 365 Copilot and agents”—to have a practical proof alongside the foundational certification.

  2. If you’re a senior developer building generative AI applications (with AWS experience or willingness to get it):
    - Assess your readiness against the AIP‑C01 exam guide. If you have the required production experience, schedule the exam. If you don’t, first build that experience through real projects; the credential won’t compensate for missing years of hands‑on work.
    - For Azure‑exclusive developers, keep an eye on Microsoft’s evolving advanced certifications, and consider an Applied Skills credential in “Build a copilot with Azure OpenAI Service” or “Implement a retrieval augmented generation (RAG) pattern” as an interim signal.

  3. If you’re a business leader or manager who needs AI literacy, not a technical badge:
    - Prepare for the Google Cloud Generative AI Leader exam. It can be completed with a few days of focused study and does not require a technical background.
    - Do not chase a developer exam just to look more “advanced.” An AB‑900 or AIP‑C01 without the supporting skills will be exposed quickly in a technical interview.

  4. If you have an old AI‑102 study plan, course enrollment, or training materials:
    - Cancel any purchase tied exclusively to the retired exam. Check course descriptions: many providers have already updated their catalogs to reflect the new paths. If not, request a switch to AB‑900 or AIP‑C01 preparation.

  5. If you’re early‑career or transitioning into AI:
    - Start with a fundamentals credential appropriate to your platform (AI‑900 for Azure basics, or the AWS Cloud Practitioner for broader cloud literacy). Then choose a specialist path: AB‑900 for Microsoft 365 administration, or an Applied Skills credential if you need to prove a specific implementation skill immediately.

Key deadline: July 22, 2026. That’s when the English AB‑900 exam updates. After that date, any study material not aligned with the new objectives may be outdated. Check the official Microsoft Learn page for AB‑900 on that day to download the latest skills outline.

Outlook

The retirement of AI‑102 is a clear marker that AI credentials are becoming more granular and role‑specific. Expect Microsoft to continue evolving its certification portfolio around Copilot and agent administration, with possible new applied skills for security, governance, and enterprise agent design. For AWS developers, AIP‑C01 will likely become a prerequisite for future professional‑level machine learning certifications. And Google’s leadership credential may spawn a more advanced business architect path.

For Windows users and IT professionals, the takeaway is straightforward: stop searching for a single “best” AI certification. Instead, choose the credential that maps directly to what your employer or target job expects you to administer, build, approve, or operate. That approach will survive the next retirement.