AI-102 Retired: Choosing AI-103, AZ-104 and Top Cloud Certifications for 2026

A Q4 2026 career guide published by LiveChennai highlights how working professionals across Chennai—spanning tech hubs and neighborhoods such as Velachery, OMR, T. Nagar, Anna Nagar, Tambaram, and Thiruvanmiyur—are turning to vendor credentials from Microsoft Azure, Amazon Web Services (AWS), and Google Cloud to stay competitive as enterprises adopt cloud infrastructure and generative AI. Yet anyone who has ever stared at a sprawling vendor certification poster knows that exam codes shift faster than a Windows Insider build. Worse, local training brochures frequently blur the line between entry-level literacy, hardcore infrastructure administration, and Python-heavy AI agent engineering.

When you strip away the marketing gloss, the Q4 2026 certification landscape contains a major structural change for Microsoft-focused professionals: the retirement of the classic Azure AI Engineer Associate (AI-102) credential and the rise of Azure AI Apps and Agents Developer Associate (AI-103) alongside the mainstay Microsoft Azure Administrator Associate (AZ-104). Understanding how those Microsoft pathways actually work—and how they compare with competing credentials from AWS and Google Cloud—can save you months of misdirected study and hundreds of dollars in exam fees.

The Big Microsoft Shift: AI-102 Is Gone, and AI-103 Changes the Game​

For years, AI-102 (Microsoft Certified: Azure AI Engineer Associate) was the standard badge for builders plugging cognitive services into enterprise applications. As noted in LiveChennai’s Q4 2026 roundup and confirmed by Microsoft Learn’s official credential retirement schedule, Microsoft retired the Azure AI Engineer Associate certification on June 30, 2026.

What happens if you already earned or renewed AI-102 before that June 30 cutoff? According to Microsoft Learn’s credential retirement policy, a retired certification does not simply vanish from your record overnight. Instead, it remains in the Active Certifications section of your Microsoft Learn transcript until its normal one-year expiration date arrives, after which it moves to the Historical Certifications section. However, no new candidates can sit for the retired exam, and it cannot be renewed once retired.

What AI-103 Actually Tests (And Why It Isn't Just a Rebadged AI-102)​

LiveChennai points Azure-focused developers toward Azure AI Apps and Agents Developer Associate (Exam AI-103) as the replacement pathway. That is the right direction for builders, but candidates need to read the fine print on Microsoft Learn before assuming AI-103 is a casual conceptual test or an automatic drop-in renewal.

According to Microsoft Learn’s official AI-103 certification page and study guide (with skills measured as of April 16, 2026), AI-103 is an intermediate-level developer and AI engineer credential centered on Python and Microsoft Foundry. Microsoft explicitly states that candidates should have hands-on experience developing applications using Python, alongside familiarity with general AI, generative AI, and Azure services.

The proctored exam runs 120 minutes, requires a passing score of 700 or greater, and is offered in English, Chinese (Simplified and Traditional), French, German, Japanese, Korean, Italian, Portuguese (Brazil), and Spanish, with pricing based on the country or region where the exam is proctored.

Here is how Microsoft Learn weights the AI-103 exam domains as of April 16, 2026:

• Plan and manage an Azure AI solution (25–30%): Selecting appropriate models (large language models, small language models, multimodal models, and Foundry Tools) and Foundry services; designing Azure infrastructure and CI/CD pipelines for AI apps and agents; managing quotas, rate limits, and costs; configuring security (managed identity, private networking, keyless credentials, role policies); and enforcing responsible AI guardrails, safety evaluations, and trace logging.
• Implement generative AI and agentic solutions (30–35%): Building generative apps with Foundry SDKs and connectors; implementing retrieval-augmented generation (RAG); defining agent roles, conversation memory, and function-calling tool schemas; orchestrating multi-agent workflows with human approval safeguards; and setting up observability, token analytics, and reflection/self-critique loops.
• Implement computer vision solutions (10–15%): Designing image- and video-generation and editing workflows; building multimodal understanding pipelines using Azure Content Understanding in Foundry Tools; and mitigating visual risks such as indirect prompt injection hidden in images.
• Implement text analysis solutions (10–15%): Extracting entities, summaries, and structured JSON via generative prompting and Foundry Tools; translating text; and integrating speech-to-text, text-to-speech, and multimodal audio reasoning into agentic interactions.
• Implement information extraction solutions (10–15%): Building RAG ingestion and grounding pipelines with semantic, hybrid, and vector search; and combining optical character recognition (OCR), layout analysis, and field extraction via Content Understanding.

Section Summary: AI-102 retired on June 30, 2026. Its successor in the Azure AI portfolio, AI-103, is a rigorous 120-minute intermediate exam focused heavily on Python development, Microsoft Foundry, RAG pipelines, and multi-agent orchestration.

AZ-104 vs. AI-103: Infrastructure Ops vs. Python AI Development​

One of the most common pitfalls in regional career guides is treating certifications like rungs on a single linear ladder. For instance, LiveChennai suggests that Azure-focused learners can pursue Microsoft Azure Administrator Associate (AZ-104) before moving into AI application development.

Can an ambitious engineer take AZ-104 and then tackle AI-103? Certainly—understanding virtual networks, managed identities, and role-based access control makes you a much safer AI architect. However, from a strictly technical and vendor-policy standpoint, AZ-104 is not a prerequisite for AI-103, and the two exams target fundamentally different daily job roles.

Inside Exam AZ-104 (Microsoft Azure Administrator Associate)​

If your background is in Windows Server, Active Directory, hybrid networking, or enterprise IT support, AZ-104 is your natural home turf. According to Microsoft Learn’s official AZ-104 certification page and study guide (updated for skills measured as of April 17, 2026), this intermediate-level exam gives candidates 100 minutes to prove they can implement, manage, and monitor an organization’s Microsoft Azure environment.

Microsoft expects AZ-104 candidates to understand operating systems, networking, servers, and virtualization, and to have practical command of PowerShell, Azure CLI, the Azure portal, Azure Resource Manager (ARM) templates or Bicep files, and Microsoft Entra ID.

The April 17, 2026 AZ-104 blueprint breaks down across five core operational pillars:

• Manage Azure identities and governance (20–25%): Managing Microsoft Entra users, groups, external users, licenses, and self-service password reset (SSPR); managing built-in Azure roles and access scopes; implementing Azure Policy, resource locks, tags, management groups, and cost alerts/budgets.
• Implement and manage storage (15–20%): Securing storage accounts with firewalls, virtual networks, shared access signature (SAS) tokens, and identity-based access for Azure Files; configuring redundancy, object replication, AzCopy, Azure Storage Explorer, lifecycle management, soft delete, and versioning.
• Deploy and manage Azure compute resources (20–25%): Automating deployments with ARM templates and Bicep files; managing virtual machines, encryption at host, availability zones, and Virtual Machine Scale Sets; provisioning containers with Azure Container Registry, Azure Container Instances, and Azure Container Apps; and configuring Azure App Service plans, TLS certificates, custom DNS, backup, and deployment slots.
• Implement and manage virtual networking (15–20%): Configuring virtual networks, subnets, peering, user-defined routes, Network Security Groups (NSGs), Azure Bastion, service and private endpoints for PaaS, Azure DNS, and internal/public load balancers.
• Monitor and maintain Azure resources (10–15%): Analyzing metrics and logs in Azure Monitor, configuring alert rules and Network Watcher/Connection monitor, and executing backup and failover operations using Recovery Services vaults, Azure Backup, and Azure Site Recovery.

Crucial Microsoft Exam Rules Every Candidate Should Know​

Whether you choose AZ-104 or AI-103, Microsoft Learn enforces several practical policies that every test-taker should keep in mind:

1. Use a Personal Microsoft Account (MSA) for Registration: Microsoft strongly recommends registering for Pearson VUE exams with a personal MSA rather than an organizational work or school Microsoft Entra account. As Microsoft warns on its certification pages, if you bind your exam records to a company account and later leave that employer, your exam records will be lost and are unrecoverable.
2. Annual Free Renewal on Microsoft Learn: Microsoft associate, expert, and specialty certifications expire every 12 months, but you do not have to pay for a full proctored exam again if you stay current. You can renew role-based certifications at no cost by passing an unproctored online renewal assessment on Microsoft Learn before your credential expires.
3. Practice Tools and Retake Windows: Both AZ-104 and AI-103 offer an interactive Exam Sandbox to preview the user interface. AZ-104 includes a free Practice Assessment on Microsoft Learn, while AI-103 provides its Practice Assessment through AI Skills Navigator (requiring account sign-in). If you do not pass a proctored exam on your first try, Microsoft’s retake policy allows a second attempt after 24 hours, with waiting periods varying for subsequent retakes.

Section Summary: Choose AZ-104 (100 minutes) if you administer cloud compute, storage, Entra ID, networking, and Bicep/ARM deployments. Choose AI-103 (120 minutes) if you write Python code to build and govern generative AI apps and agents in Microsoft Foundry. Always register with a personal Microsoft account so you never lose your transcript when changing jobs.

Cross-Cloud Context: How AWS, Google Cloud, and Kubernetes Fit In​

Because enterprise environments rarely live in a single-vendor bubble, LiveChennai’s Q4 2026 guide also highlights certifications across AWS, Google Cloud, and Kubernetes. However, comparing these credentials side by side requires separating foundational literacy tests from senior engineering exams.

AWS Certified AI Practitioner vs. Architecture Pathways​

At the entry point of the AI market, LiveChennai highlights the AWS Certified AI Practitioner, noting its 90-minute duration, 65 questions, and US$100 cost. AWS’s official certification documentation confirms those exact specifications and classifies the credential as Foundational—valid for three years.

Who is it actually for? AWS states that the ideal candidate is familiar with AI, machine learning, and generative AI concepts and AWS service use cases, and "uses, but does not necessarily build AI/ML solutions on AWS." That makes it an accessible starting point for business analysts, product managers, IT staff, and developers looking to validate broad AI literacy. (One localization detail worth noting from AWS’s official page: the Italian and German language versions of the AWS Certified AI Practitioner exam are retired after October 15, 2026, while English, Arabic, French, Japanese, Korean, Portuguese, Spanish, and Simplified/Traditional Chinese continue.)

Once candidates finish that foundational step, both LiveChennai and AWS’s official career guidance note a natural fork in the road: professionals targeting cloud infrastructure design can pursue AWS Certified Solutions Architect – Associate, while AWS also points data and ML practitioners toward its Associate-level Data Engineer or Machine Learning Engineer tracks (earning AWS Certified Machine Learning Engineer – Associate automatically recertifies the AI Practitioner credential).

Google Cloud: From Associate Cloud Engineer to Professional ML Engineer​

On the Google Cloud side, LiveChennai pairs the entry-level Google Cloud Associate Cloud Engineer (aimed at professionals starting in Google Cloud and DevOps) with the senior-tier Google Cloud Professional Machine Learning Engineer.

Candidates should not underestimate the jump between those two credentials. Google Cloud’s official documentation confirms that the Professional Machine Learning Engineer exam lasts two hours, contains 50–60 multiple-choice and multiple-select questions, is offered in English and Japanese, and costs US$200 (plus applicable tax). While there are no formal prerequisites, Google explicitly recommends 3+ years of industry experience, including 1 or more years designing and managing solutions using Google Cloud, along with minimum proficiency in Python and SQL to interpret code snippets.

Crucially, Google Cloud’s official exam page notes that the Professional Machine Learning Engineer exam was updated to reflect the transition from Vertex AI to Gemini Enterprise Agent Platform, alongside updates to Google Cloud’s data and analytics stack. Anyone studying from older third-party prep books that only mention legacy Vertex AI workflows should immediately download the current exam guide from Google Cloud before scheduling their test.

The DevOps and Platform Engineering Layer: Kubernetes (CKA and CKAD)​

For DevOps engineers, site reliability engineers (SREs), and platform engineers, LiveChennai notes that cloud credentials pair effectively with Kubernetes certifications such as CKA (Certified Kubernetes Administrator) and CKAD (Certified Kubernetes Application Developer), backed by practical skills in Docker, Linux, Git, and basic Python. In modern hybrid environments—including Azure Kubernetes Service (AKS) or container workloads managed alongside AZ-104—combining core cloud administration with container orchestration remains one of the most resilient career strategies in enterprise IT.

Section Summary: Not all AI and cloud badges carry the same weight. AWS Certified AI Practitioner (US$100, 90 minutes) tests foundational concepts, whereas Google Cloud Professional Machine Learning Engineer (US$200, 2 hours) targets veterans with 3+ years of experience and now reflects Google’s shift toward the Gemini Enterprise Agent Platform.

A Buyer-Beware Checklist for Local Training Centers and Bootcamps​

Because Chennai hosts one of India’s largest technology workforces—anchored by major enterprise campuses along the OMR IT corridor, Siruseri SIPCOT, and Sholinganallur —the city supports hundreds of classroom and online coaching centers. LiveChennai notes that institutes across Velachery, OMR, T. Nagar, Anna Nagar, Tambaram, and Thiruvanmiyur, including providers such as FITA, Besant Technologies, ACTE, SLA Institute, AP2V, and 360DigiTMG, advertise training in cloud computing, AI, and DevOps.

While structured classroom or weekend instruction can help working professionals stay disciplined, LiveChennai includes a warning that every IT veteran will applaud: ensure that your certification is issued by the technology provider (Microsoft, AWS, or Google) rather than the training institute.

Before you hand over tuition money to any third-party coaching center—in Chennai or anywhere else—run through this five-point verification checklist:

1. Separate "Course Completion" from Vendor Certification: A certificate of completion printed by a local training institute only proves you sat through their classes. The credential employers actually verify on your resume is issued only when you pass the official proctored exam administered by Pearson VUE (for Microsoft and AWS) or Google Cloud’s official testing partner.
2. Audit the Syllabus Against Live Exam Codes: Ask the trainer for their exact module breakdown and compare it line-by-line against Microsoft Learn, AWS Skill Builder, or Google Cloud’s current exam guide. If a coaching center is still pitching AI-102 syllabus sheets after its June 30, 2026 retirement—or teaching Google ML courses that ignore the Gemini Enterprise Agent Platform update—walk away.
3. Demand Hands-On Sandbox and Lab Access: You cannot pass intermediate or professional cloud exams by memorizing slide decks. Confirm before paying that the program includes live cloud sandbox labs, mock assessments, and recorded sessions that fit around shift or on-call schedules.
4. Compare Free Official Learning Paths First: Before spending thousands of rupees or hundreds of dollars on private coaching, explore the free self-paced modules on Microsoft Learn, AWS Skill Builder, and the Google Cloud Free Tier. Often, the vendor’s own free study guides, exam sandboxes, and practice assessments cover 80% of what you need.
5. Align With Your Current Employer’s Stack: As LiveChennai rightly points out, the highest-return certification is the one you can apply at work on Monday morning. If your enterprise runs Microsoft Entra ID, Windows Server hybrids, and Azure VMs, AZ-104 will pay immediate dividends. If your development team is building Python RAG pipelines and enterprise copilots, AI-103 is the sharper tool.

Cloud and AI certifications will never replace real-world troubleshooting under pressure, nor will they write clean Python or Bicep code for you. What the right 2026 credential does provide is a structured, vendor-verified benchmark showing that your skills match current production architectures—not the cloud of three years ago. Check the official study guides, register with your personal account, build real labs, and make sure the badge you chase matches the job you actually want.