Microsoft AI Certification Updates - March 2026

Microsoft AI Certification Updates - March 2026

In case you missed it, Microsoft published a blog in early March 2026 announcing a number of changes to AI focused certifications in what I consider to be Microsoft’s biggest certification update since the introduction of role-based certifications.

I’ll avoid repeating the blog post verbatim and would encourage you to read it in full, but the biggest updates include:

  • AI Engineer Associate (AI-102) is being retired in June in favour of the new AI App and Agent Developer (AI-103) certification.
  • Data scientist associate (DP-100) is being retired. Additionally, a new MLOps Engineer (AI-300) certification is being added.
  • AI Fundamentals (AI-900), Azure Developer (AZ-204), Azure Security Engineer (AZ-500), and Windows Server Hybrid Admin (AZ-801) are being replaced by updated versions - Azure AI Fundamentals (AI-901), AI Cloud Developer (AI-200), AI Security Engineer (SC-500), and Windows Server Hybrid Admin (AZ-802).
  • Two new Data focused certifications including SQL AI Developer (DP-800) and Azure Databricks Data Engineer (DP-750).
  • A new cert in the security space, Cybersecurity Business Professional (SC-730).

Note: Those with AI-900 will maintain their certification without undertaking AI-901. The same is true for AZ-801/2 and AZ-803.

My thoughts on the updates

Focusing on the more positive elements, I think it’s good to see a variety of AI certifications available for a range of different users / developers where only a couple previously existed. The DP-100 (data scientist associate) certification does feel a little outdated, not because of the role-based requirement changing, but because the content needed refreshed (e.g. lots of Azure ML, not foundry focused) so I think change was needed there. It’s also nice to see additional focus on AI security, data, operations, etc.

At a high level, updating 5 certifications (I'm excluding DP-100 -> AI-300 as I don't think it's 1:1), retiring 1 (DP-100), and adding 4 (AI-300, DP-750, DP-800, SC-730) seems like a reasonable set of updates.

However, I think the changes are mostly challenging - I’m going to temper some of the below by acknowledging that creating certification usually takes 12-18 months as far as I understand, but most of the difficulty is in the updates not giving new people to the certification ecosystem a straightforward path in the near-term, a lack of positioning for data science skills, and a current lack of clarity in what the differences are in both learning and exam content (i.e. what's different between AI-102 and AI-103). More thoughts below:

  • I think such a volume of changes is going to cause confusion. Figuring out what to do and when is not easy even with the blogs guidance on what to do if you’re prepping or considering certs and what replaces what. This is reinforced by the number of questions being asked by people in the blog post comments.
  • The beta periods will roll out starting now until June (each one will likely end at a different date), though some retirements come during that time. This is tricky for those certified with upcoming renewals as well as new prospects, but the bigger issue is that the beta periods usually last a few months meaning the exams probably won't be GA until at August to November (depending on beta dates in the timeline above), and for most users who won't undertake during the beta period, there's a bit of a gap between exam retirements and the replacements being GA.
  • People are constantly undertaking exams or looking to obtain certifications, including partner organisations that have pushed to undertake the likes of the AI engineer associate certification in the past 18-24 months. Microsoft even had a drive last summer for pushing the AI-102 certification. Anyone who wants to undertake replacement exams will now need not just to study for the new certifications, but pay to do the exam (50% discount for partner orgs). I think it would have been a good step to create some conversion course for those already certified, or offer vouchers for people to undertake the new exam at no cost if it's done within a certain timeframe. Of course there's still time for this to happen.
  • The certification poster has already been updated with the new beta certifications, but I think that there is a risk that some of the changes create more role-based overlap. For example, the AI-103 and AI-200 (AI App and Agent, AI Cloud Developer) certifications both have mention of generative AI apps and orchestration, the same applies for AI-200 and AI-300 (ML Ops Engineer) around automation, monitoring, and performance. I know we will see more through the upcoming learn pathways but it seems as though it's likely some of these could be targeted at multiple roles.
  • Beta exams are usually a little tricker for two reasons; inconsistent or relatively new question sets, and lack of available learning pathways for exam prep. In this case, it’s the exam prep I think is likely to be most challenging, and I think that unless you’re actively developing AI and agentic solutions on a weekly basis, I would suggest waiting a few months until this is a little better positioned.

I think the biggest challenge here is just about the unknowns, including not being clear on how this picture might change as time goes on - this makes it difficult for people who have multiple certifications now to jump in with confidence that their time and effort is well spent.

Conclusion

I gave feedback previously that the Microsoft AI training available was largely pretty good, but the certifications didn’t provide complete coverage and needed some updates, so I’m hopeful these updates are going to help resolve that and it’s good to see these focused AI updates.

That said, I was reasonably critical when AWS retired some well positioned specialty certifications on short notice and announced a set of new certs quite rapidly, each of which required current certification holders to sit new exams. Even though I can see the logic in the sense of the need for modernised AI training pathways and certifications, the same thing applies here - L&D takes time and dedicated effort alongside a day job and for the next number of months, I think this will make things difficult for people to navigate.

I will be undertaking probably around half of the beta exams, targeting MLOps and the updated AI/Agent Developer (AI-103) certs first, so will post thoughts here as soon as possible.