Microsoft added two new exam codes to its certification catalog in September 2026: AB-100 and AB-620. Neither is a rename of an existing exam. Together, they mark Microsoft’s decision to stop treating agentic AI as a chapter inside its older AI credentials and instead give it a certification track of its own, split cleanly between deciding where AI agents belong in a business and actually building them.
Why Microsoft split this into two exams instead of one
Designing an agentic AI strategy and configuring an agent in Copilot Studio are different skill sets, and Microsoft’s new track treats them as two separate jobs rather than two halves of the same exam.
- AB-100 sits at the architecture and governance level, planning where agents fit an organization’s processes, designing guardrails, and making the case for AI-powered transformation.
- AB-620 sits at the build level, configuring, integrating, and shipping agents inside Copilot Studio, connected to real enterprise data and systems.
A solutions architect who never opens Copilot Studio can still sit AB-100. A developer who never writes a business case can still sit AB-620. That’s the actual signal here: Microsoft expects two distinct job profiles to form around agentic AI, not one generalist role trying to cover both.
Inside AB-100: Agentic AI Business Solutions Architect
AB-100 leads to the Microsoft Certified: Agentic AI Business Solutions Architect Expert credential, and Expert is the correct word for it this sits at the top tier of Microsoft’s Fundamentals/Associate/Expert ladder, above associate-level prerequisites.
- Duration: 100 minutes, delivered via Pearson VUE
- Format: Multiple-choice and scenario-based questions
- Passing score: 700/1000
- Prerequisite: At least one active associate-level Microsoft certification; this is the detail worth getting right. Microsoft’s own certification page specifies an associate-level credential, and that list includes things like AB-620, AB-410, or AI-103, along with qualifying Dynamics 365 or Power Platform associate certifications. AB-730 and AB-731 are fundamentals-level exams, not associate-level, so they don’t satisfy this prerequisite on their own a distinction that’s easy to get wrong if you’re going off secondhand summaries rather than Microsoft’s own exam page.
Content-wise, AB-100 leans on Microsoft Copilot Studio, Microsoft Foundry, multi-agent orchestration design, application lifecycle management for agentic-first solutions, and responsible AI governance aligned to Microsoft’s Responsible AI Standard. It’s built for solutions architects, enterprise architects, and senior consultants who are deciding where agentic AI belongs in an organization, not necessarily the people writing the integrations themselves.
Inside AB-620: AI Agent Builder Associate
AB-620, formally Designing and Building Integrated AI Agent Solutions in Copilot Studio, leads to the Microsoft Certified: AI Agent Builder Associate credential. Microsoft describes the target candidate plainly: a professional developer or advanced builder who builds, extends, and integrates custom agents for enterprise-grade solutions.
- Format: Proctored exam with interactive components, delivered via Pearson VUE
- Passing score: 700/1000
- No standalone prerequisite exam required, unlike AB-100
Despite that, “no prerequisite” doesn’t mean no real barrier to entry. The exam assumes working familiarity with Power Fx, Microsoft Dataverse, the Power Platform, Microsoft 365 Copilot, and Microsoft Foundry, plus comfort discussing retrieval-augmented generation and agent-to-agent protocol behavior. Reviewers who’ve sat the AB-620 exam describe the domain covering integrating and extending agents as the largest and most decisive section the part that rewards actual hands-on Copilot Studio build time over memorized definitions. Someone who’s read the documentation but never shipped an agent tends to get caught out specifically there.
AB-620 on p2p exam is the right target if your actual job is building and shipping agents IT application developers, consultants, and ISV partners working hands-on in Copilot Studio rather than deciding organizational AI strategy.
How AB-100 and AB-620 fit Microsoft’s wider AI certification reset
These two exams didn’t launch in isolation. Microsoft has been rebuilding its AI certification catalog across 2026, and the older AI-102 (Azure AI Engineer Associate) was retired mid-year in favor of AI-103, Developing AI Apps and Agents on Azure, built around Microsoft Foundry rather than individual pre-built AI services. That’s the same directional shift AB-100 and AB-620 represent from a different angle toward generative AI application development and multi-agent orchestration as the baseline expectation, not a specialty.
| Exam | Level | Focus |
| AI-900 | Fundamentals | General AI literacy |
| AB-730 / AB-731 | Fundamentals | Personal Copilot productivity / organizational AI adoption and governance |
| AI-103 | Associate | Azure AI app and agent development on Microsoft Foundry |
| AB-620 | Associate | AI agent building in Copilot Studio |
| AB-410 | Associate | Power Platform AI app building (successor to the retired PL-200) |
| AB-100 | Expert | Agentic AI business solutions architecture |
Worth noting: AB-620 and AI-103 sit at the same associate level but target different tooling. AI-103 is code-first Foundry development; AB-620 is Copilot Studio and Power Platform-first. Microsoft isn’t asking developers to pick one over the other so much as expecting different developer profiles to land on whichever platform their employer has already standardized on.
Why now the labor market context
Microsoft doesn’t usually build out a certification track this quickly without data behind it. Stanford’s 2026 AI Index reported that agentic AI’s share of US job postings grew sharply year over year between 2024 and 2025 a jump large enough that a category barely present in listings two years ago now represents a meaningful share of postings referencing AI skills specifically. That kind of rapid demand shift is consistent with what’s driving certification bodies generally to move faster than their usual multi-year revision cycles: Anthropic built a Claude certification program from scratch in 2026, and AWS rewrote its own Machine Learning Engineer Associate exam around generative AI and agentic workflows the same year. Agentic AI is becoming its own certification lane across vendors, not a chapter tucked inside older exams.
Should you sit either exam right now?
A brand-new exam code isn’t automatically worth booking the day it opens. A few things worth weighing:
- Practice assessment lag: Microsoft’s own practice assessments for a newly launched exam typically take time to catch up to general availability. Early candidates are often studying from the official skills outline and community write-ups rather than an official practice test.
- Content still settling: third-party practice question sets for both exams have already gone through multiple revisions within weeks of launch, a normal sign that Microsoft is still tightening the official skills outline. Expect it to keep shifting for a few months.
- Prerequisite planning for AB-100: confirm which of your existing certifications actually qualifies as the associate-level prerequisite before you commit a study plan to AB-100 specifically; don’t assume any Microsoft credential you hold automatically counts.
- Match the exam to the job, not the hype. Check whether the roles you’re targeting actually name Copilot Studio and hands-on agent building (lean AB-620) or AI strategy, planning, and governance (lean AB-100). Don’t assume either credential covers both halves.
If your organization has already committed to Copilot Studio for internal automation, sitting AB-620 now puts you ahead of a genuinely thin talent pool. If you’re further from the keyboard and closer to deciding where agents belong across the business, AB-100 is the better use of study time once your prerequisite certification is actually sorted.
FAQ
Does AB-100 or AB-620 replace the retired AI-102 exam?
No. AI-102’s direct successor is AI-103, Developing AI Apps and Agents on Azure. AB-100 and AB-620 are a separate, newer track focused specifically on agentic AI business strategy and Copilot Studio agent building.
Do AB-730 and AB-731 count as the associate-level prerequisite for AB-100?
No, Microsoft classifies AB-730 and AB-731 as fundamentals-level exams, not associate-level, so they don’t satisfy AB-100’s prerequisite on their own. You’ll need an actual associate-level credential such as AB-620, AB-410, or AI-103.
Is AB-620 hard to pass without hands-on Copilot Studio experience?
Candidates without real build time consistently report struggling most on the integrating-and-extending-agents domain, which carries the heaviest weight on the exam. Reading alone tends to leave gaps specifically there.
Are these exams likely to change soon after launch?
Very likely. Both launched in September 2026, and third-party practice materials tracking them were already being revised within weeks a normal early-life-cycle pattern. Check Microsoft’s own exam page for the current skills outline close to your test date rather than relying on notes from a few months back.