I Took All Three New Microsoft AI Exams - So You Know What To Expect

TL;DR

• I sat all three Microsoft AI exams in beta. Each one tests a different aspect of AI readiness across Microsoft 365 & Azure.
• AB-900 felt closer to a modernised MS-900, with far more Purview and M365 awareness than I expected.
• AB-730 focussed on practical Copilot usage, agent administration and understanding when to use each Copilot capability.
• AB-731 shifted fully into business and organisational AI adoption, with a heavy focus on Azure AI services, licensing and real-world transformation.

Introduction

Microsoft recently introduced a new set of AI-focussed certifications, covering fundamentals, practical AI usage and business transformation. With AI now embedded across Microsoft 365, these exams will become relevant skill paths for architects, consultants and anyone shaping AI adoption. (I do have an opinion on this later in the article).

Over the last couple of weeks, I have sat all three beta exams: AB-900, AB-730 and AB-731. The experience across them was quite different, as to be expected and there were areas that surprised me.

In this post, I will share my experience of sitting the exams, what stood out across each one and what you should expect when they release globally, and come out of beta. I'll also cover how beta exams work, and when results are typically released.

Let's go! 📜

What is a Microsoft Beta Exam?

Microsoft runs new exams in beta before they are released globally. These are typically full exams, but they are still being validated, and the process works slightly differently to normal. Beta exams are usually heavily discounted (80% off!) for the first 300-400 people that register and sit the exam in the required window.

• If you use the beta discount and complete the exam, you will typically receive a 25% discount voucher on your next exam.
• You do not receive a score at the end of the exam. Results are released after the exam goes live worldwide.
  • Exams are usually in beta for 10-12 weeks. Once they go GA, results are then released to beta takers around 2 weeks after that.
• Beta exams are intended for people with real experience in the technology area being tested. There may be no learning path, documentation or official study material available during this period.
• Passing a beta exam counts fully, like passing a normal Microsoft exam. If you don't pass, you can't re-take until the exam is GA (Even though you should only be getting your results when the exam is GA).

AB-900 - More Purview Than Expected!

Going into AB-900, I expected a light introduction to AI concepts with some Copilot and Agent basics. Instead, it felt much closer to a modernised MS-900 with a bit of SC-900 overlap. With MS-900 being deprecated, AB-900 seems to be picking up some of that ground, especially around the M365 landscape.

• The exam leaned far more towards Purview than Copilot or Agent administration.
• A lot of the challenges for me came from knowing where certain things live across the various portals. Brush up on where things are!
• Far less Agent "administration" than I expected - but I have spent a lot of my time recently in Copilot Studio. My expectation of Agent administration compared to this fundamentals exam may be different.

It wasn't difficult (I say w/o knowing my results!) but it was broader than I anticipated. Get to know where things live in what portal and definitely spend some time in the Purview portal.

AB-730 - Where It Starts To Happen

AB-730 felt like the point where the exams move forward from broad fundamentals into something more practical. Compared to AB-900, this one was much more aligned to real usage, business outcomes and the decisions required when introducing AI into an organisation.

For me, the focus areas were clear:

• An emphasis on Copilot Studio / Custom Agents and when to use it/them.
• More around Agent creation and how things fit together operationally.
• Not much Purview!
• Several questions that felt broad and open to interpretation, which is not unusual for a beta.
• A need to understand the differences between Copilot Chat, Microsoft 365 Copilot and Copilot Studio agents.
• Licensing also appeared, so it helps to be confident in which capabilities come from where.

Overall, the AB-730 felt more practical and more aligned to real customer scenarios. It tests judgement of AI adoption and not just feature knowledge. Given that it sits within the AI Business Professional track, the focus on aligning AI to business value made complete sense.

AB-731 - Business First, with a Heavy Azure AI Focus

AB-731 was very different to the first two exams. It felt far more like a business AI exam, but you still needed a solid understanding of Azure AI services to have a good go at the answers. This exam is the AI Transformation Leader exam, and that sentiment was seen throughout.

My experience:

• The exam was heavily focussed on Azure AI services and Azure AI Foundry. Remembering which service for what scenario was the most challenging part for me.
  • The exam still referenced Azure AI Foundry; I'd imagine this would change to Microsoft Foundry by the time the exam is GA.
• There were clear questions around Copilot licensing and organisational AI adoption, which aligned well with the exams' purpose.
• I liked how it tested higher level AI understanding without going into technical deep dives.
• Very little portal / configuration content, which makes sense given the first two exams test this criteria.
• Whilst it was suggested Microsoft Learn is available during this exam, for me, it wasn't.

It felt like this was the exam to tie everything together. AB-900 being the fundamentals, AB-730 with some more practical work, and AB-731 which allows you to step back and look at AI from a business-perspective.

How Do They Compare?

Innovation Beats Regulation & Education

I like the phrase "Innovation beats Regulation". In this case, innovation will beat both regulation and education, which is a concern. By the time these exams reach GA, some terminology, service names or portal locations will almost certainly have shifted. That isn’t a criticism of the exams, but the reality of working in a space where the technology seems to change daily.

The long-term value comes from understanding the concepts, business impact and the way AI fits into organisational processes. Certifications are a point-in-time, and with these so far not requiring renewal; people will have to refresh their knowledge to keep up.

Wrap Up

These three exams give a solid foundation in how Microsoft approaches AI across Microsoft 365 and Azure, but they each test something different; awareness, practical usage and business adoption.

If you're working with AI day-to-day, they are worth taking. However, wait for them to leave beta, use the official learn paths and focus on the study guide for each exam.