We earn a commission if you buy through links on this page. How we're funded.
This certification has been retired. Microsoft Certified: Azure AI Engineer Associate (AI-102) is no longer available to take. Microsoft reports the retirement date as 2026-06-30. The replacement is Microsoft Certified: Azure AI Apps and Agents Developer Associate (AI-103). We have left the material below in place because it is still useful for understanding what the exam covered and how the successor differs, but do not book it — you cannot.
Quick answer
AI-102 (Azure AI Engineer Associate) was retired on 30 June 2026 and can no longer be booked; Microsoft names Azure AI Apps and Agents Developer Associate (AI-103) as its replacement. AI-102 assumed you could write Python or C#, and it tested whether you could implement Azure AI services, including generative AI solutions. For a working developer with some Azure exposure, about six weeks of part-time preparation was realistic. For AI-103, as for AI-102, spend your hours building in a sandbox: candidates who only watch course videos meet scenario questions they have never touched. If you want a structured course in the app-and-agent patterns first, DataCamp’s Associate AI Engineer for Developers builds them in Python in about 29 hours, against the OpenAI API rather than Azure.
See Associate AI Engineer for Developers on DataCamp →
The table below compares 4 certifications on provider, level, realistic time, coding needed and best for.
| Certification | Provider | Level | Realistic time | Coding needed | Best for |
|---|---|---|---|---|---|
| Azure AI Engineer Associate (AI-102) | Microsoft | Intermediate (associate) | ~6 weeks part-time for working devs | Yes (Python or C#) | Retired 30 June 2026 — replaced by AI-103 |
| Microsoft Learn AI-102 learning path | Microsoft | Intermediate | Included in the six weeks | Yes | The free official prep spine |
| Azure AI Fundamentals (AI-901) | Microsoft | Foundational | ~2 weeks part-time | Basic Python | The optional on-ramp for Azure newcomers |
| AWS ML Engineer Associate (MLA-C01 / MLA-C02) | AWS | Intermediate (associate) | ~2–3 months of prep | Yes (Python) | The AWS-side counterpart; moving to MLA-C02 |
How hard was AI-102 really?
Harder than its fundamentals sibling by a full category. Where AI-900 asked what Azure's AI services are, AI-102 asked how you would configure, secure, deploy and troubleshoot them in scenarios with more than one defensible-looking answer — the full gap is mapped in our AI-900 vs AI-102 comparison. Working developers with Azure exposure generally found it demanding but fair. Candidates who had never deployed an Azure resource found it brutal, because the exam's difficulty lives in implementation detail that videos cannot install in you.
Exam AI-900: Microsoft Azure AI Fundamentals is retired. The replacement is Exam AI-901: Microsoft Azure AI Fundamentals, which earns the same Azure AI Fundamentals certification but expects Python and familiarity with REST APIs and SDKs.
If you cannot yet write working Python, you are not ready for this exam's successor or for Azure's entry exam — build that foundation first. Azure AI Fundamentals is still the entry point to the Azure AI stack, but its exam is now AI-901, which replaced AI-900 on 30 June 2026 and expects basic Python; our AI-900 study guide explains the change.
What's actually on the exam?
The official skills outline was the exam's contract. At a high level, it covered planning and managing an Azure AI solution; implementing generative AI solutions; and implementing agentic, computer-vision, natural-language and knowledge-mining or document-intelligence solutions.
Question formats mixed standard multiple choice with scenario sets and case studies. The scenario questions were where unprepared candidates failed: they presented a business requirement — cost ceiling, latency constraint, data-residency rule — and four implementations that all sound plausible until you have actually built one.
What should you have before starting?
Three things make the six-week plan realistic rather than optimistic: working Python or C# (you will read and complete SDK code, not admire it), basic Azure portal fluency (resource groups, keys, endpoints, cost management), and comfort calling a REST API. Missing the language: fix that first — the exam cannot be talked past. Missing only the Azure familiarity: add a week of portal time at the front of the plan. Data engineers and developers coming from AWS or GCP usually adapt quickly; the concepts transfer, the service names do not — and if your employer runs on those clouds anyway, prep for AWS's Machine Learning Engineer – Associate or the GCP ML Engineer exam instead of learning Azure for its own sake.
The six-week study plan we published for it
Built around building, at roughly an hour a day:
- Weeks 1–2: foundations and plumbing. Work the Microsoft Learn AI-102 path's opening modules, deploy your first Azure AI resources, and get comfortable with authentication, keys and endpoints. Everything later sits on this.
- Weeks 3–4: the big domains. Generative AI first — deploy a model, build a small chat application, add retrieval over your own documents — then computer vision and language services, calling each from code rather than only the portal.
- Week 5: knowledge mining and document intelligence, plus a review pass over whatever felt thinnest. Wire one end-to-end mini-app that touches three services — this is both revision and a portfolio artefact.
- Week 6: practice assessments, wrong-answer review, and booking. The plan put the booking at the start of the week, on the reasoning that a fixed date converts studying into finishing. That step is the one part of it you cannot repeat — AI-102 can no longer be booked.
Not sure this is the right one for you?
Tell the picker about your background and what you want the certificate to do, and it narrows the list to the one or two courses we would start with. It suggests only our affiliate partners’ courses, and says so before it suggests anything.
Try the AI Certification Picker →Why hands-on prep beats video courses
Because the exam was written by people who build with these services, and its distractor answers are exactly the mistakes you make the first time you build — wrong resource tier, wrong authentication pattern, a service that almost fits the requirement. One real deployment inoculates you against a whole family of trick options. Azure's free tiers and trial credit cover most of what prep requires; the sandbox time costs far less than a second video course and teaches far more.
There is a career bonus hiding in this: the mini-app you build in week 5 is interview material. A candidate who can demo retrieval over their own documents has something no exam badge provides.
If you want structured practice in those patterns before the Azure-specific work, the course we would use is DataCamp’s Associate AI Engineer for Developers: about 29 hours of hands-on Python exercises on the OpenAI API, embeddings, a Pinecone vector database, LangChain and the Model Context Protocol. It is not prep for AI-103 — nothing in it is Azure, and we have not reviewed AI-103’s syllabus — so treat it as the retrieval-and-agents groundwork and do the Foundry building in your own sandbox. It is part of a DataCamp subscription and ends in a certificate of completion, not a Microsoft certification.
Registration is closed — what to do instead
AI-102 can no longer be registered for or sat. Microsoft retired both the certification and its renewal assessment, and the credential page no longer lists a fee. If you already passed it, it simply cannot be earned or renewed any more — check your Microsoft Learn transcript for how Microsoft now displays it, which we have not independently verified.
The table below shows the fee and format of AI-103, as Microsoft publishes them. Every other exam we track is compared on our AI certification exams page.
| Credential | Vendor | Level | Fee | Exam format | Enrol |
|---|---|---|---|---|---|
| Microsoft Certified: Azure AI Apps and Agents Developer Associate (AI-103) | Microsoft | Intermediate | $165 USD (varies by country/region) | 120 minutes | Microsoft → |
If you were preparing for it, the credential Microsoft names as the replacement is Azure AI Apps and Agents Developer Associate (AI-103), listed at $165 USD with pricing that varies by region. Check its own exam guide before reusing any of the plan below: the scope moved toward AI apps and agents, and we have not yet reviewed the new syllabus, so we are not going to pretend the six weeks map across unchanged.
Who this exam was for — and who it wasn't
It was for people who build on Azure: developers in Microsoft-stack organisations, data engineers wiring AI services into pipelines, and consultants implementing Azure AI for clients. That is the same audience AI-103 now addresses. The wider credential landscape for your role is mapped in our software engineers guide.
It was not for people who do not code — Azure AI Fundamentals is the honest ceiling for non-developers, though its exam is now AI-901, which expects basic Python — nor for anyone whose platform is AWS or GCP, where the counterpart exams cover the same ground natively. Those two boundaries are the part of this page that still holds: the successor exam assumes code in the same way, so if you are mid-way through learning Python, the programming foundation comes first either way.
Where most AI-102 advice gets it wrong
The dominant prep pattern — buy a video course, watch it end to end, book the exam — produces exactly the candidates the scenario questions were designed to filter out. Completion of content is not competence with services, and AI-102 is unusually good at telling them apart. The advice industry keeps selling watching because watching is what it can sell; the exam keeps rewarding building because building is what the job is.
Our position at the time: treat AI-102 prep as six weeks of supervised portfolio-building with an exam at the end. That framing cost nothing, changed where the hours went, and left you with working artefacts plus a credential — instead of a credential and a watch history. Nothing about the retirement makes it less true of the exam that replaced it.
Verdict
For working developers on Azure, AI-102 was one of the few AI certifications whose preparation was directly the job: six weeks of building against Azure AI services, verified by a proctored associate credential that appeared by name in job specs. It cannot be sat now. The action this page points at is AI-103: read its exam guide, then prepare hands-on and keep the mini-app you build, because that is the part of the method the retirement did not change. What we have not done is re-time the six weeks against the new syllabus, so treat the plan above as a shape rather than a schedule. If you are still choosing between cloud exams, our 2026 rankings and analysis of which certifications are worth it set the field, the staged path lives in the AI certification roadmap, and our free Picker matches an exam to your stack in two minutes.
Ready to start?
Included in a DataCamp subscription rather than bought outright. DataCamp's pricing page shows the plans and the price for your country, and one subscription covers the rest of its catalogue too.
Frequently asked questions
How long did AI-102 take to prepare for?
About six weeks part-time for a working developer with some Azure exposure — longer if Azure was new to you, and add the time to learn Python or C# first if you could not yet code. Candidates coming straight from AI-900 with no coding background should not have booked until the programming foundation was real.
That figure is recorded here because the shape of the preparation is the useful part, not the exam itself. AI-103 replaced this credential at the associate tier and is a different syllabus, but the same constraint governs it: the weeks go into hands-on service configuration rather than into video, and a developer who has not built against the platform will not shorten that with more reading. Read the new exam guide before reusing any of this as a plan.
Was AI-900 required before AI-102?
No — Microsoft set no prerequisite, and working developers routinely skipped the fundamentals exam. AI-900 earned its place only if Azure's AI stack was entirely new to you and you wanted a low-stakes first pass at the vocabulary before committing to associate-level prep.
Both halves of that question are now history. AI-900 was retired on 30 June 2026 as well; Azure AI Fundamentals now requires exam AI-901, which expects basic Python. If you are planning the same route today it runs AI-901, then AI-103 — and you should check AI-103's own stated prerequisites rather than assuming they carried over.
Did AI-102 require coding?
Yes. The exam assumed you could read and complete code that calls Azure AI services — Python and C# were the standard paths — and its scenario questions presumed implementation experience. It was the defining difference between AI-102 and the no-code AI-900 that sat beneath it.
That divide is the one structural fact worth carrying forward from this page. Every vendor's AI ladder has a rung where the exam stops testing recognition and starts assuming you have built something, and crossing it unprepared is the most expensive mistake available in certification. Whether the associate exam in front of you is AI-103 or another vendor's, find out which side of that line it sits on before you book.
What was the passing score for AI-102?
Microsoft scores its role-based exams on a scaled system with a published passing threshold. Treat consistent comfortable passes on the official practice assessment as your booking signal rather than aiming to scrape the line.
Scaled scoring is worth understanding because it makes the raw number of correct answers a poor guide: questions are not worth equal marks and the reported figure is not a percentage of the paper. The practical consequence is that you cannot compute how close you are from a practice run, which is why a comfortable margin on unseen questions is the only sensible readiness test — and that holds for AI-103 and the other role-based exams too.
Does an AI-102 certification I already hold still expire?
There is nothing left to renew. Microsoft's credential page states that the certification AND its renewal assessment are both retired, so the annual free assessment that associate-level credentials normally require no longer exists for this one.
That is a change from how the exam worked while it was live. Associate certifications carried a recurring renewal — unproctored, free, and considerably lighter than the original exam — while fundamentals certifications such as Azure AI Fundamentals do not expire at all.
What that means for a credential you already passed is a question for Microsoft rather than for us: check your Microsoft Learn transcript for how a retired certification is displayed. We have not verified that behaviour and are not going to guess at it. If you need a current Azure AI credential on your profile, the replacement is AI-103.
Keeping this current. Course formats, prices, and certification exam fees change and vary by region. We review our guides regularly, and we always recommend confirming the specifics on the provider's official page before you enrol.