Quick answer
AI-102 (Azure AI Engineer Associate) is a genuine practitioner exam: it assumes you can write Python or C#, and it tests whether you can implement Azure AI services — including generative AI solutions — not whether you can define them. For a working developer with some Azure exposure, about six weeks of part-time preparation is realistic. The single most important prep decision is where your hours go: candidates who build in a sandbox pass; candidates who only watch course videos meet scenario questions they have never touched. Exam fee: whatever the provider currently lists.
| 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#) | The exam this guide prepares you for |
| Microsoft Learn AI-102 learning path | Microsoft | Intermediate | Included in the six weeks | Yes | The free official prep spine |
| Azure AI Fundamentals (AI-900) | Microsoft | Foundational | ~2 weeks part-time | No | The optional on-ramp for Azure newcomers |
| AWS ML Engineer Associate (MLA-C01) | AWS | Intermediate (associate) | ~2–3 months of prep | Yes (Python) | The AWS-side counterpart |
How hard is AI-102 really?
Harder than its fundamentals sibling by a full category. Where AI-900 asks what Azure's AI services are, AI-102 asks 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 find it demanding but fair. Candidates who have never deployed an Azure resource find it brutal, because the exam's difficulty lives in implementation detail that videos cannot install in you.
If you cannot yet write working Python or C#, this is not your exam — build that foundation first, and treat the AI-900 route as your entry point to the Azure AI stack in the meantime.
What's actually on the exam?
Work from the official skills outline — it is the exam's contract. At a high level, the outline covers 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 mix standard multiple choice with scenario sets and case studies. The scenario questions are where unprepared candidates fail: they present 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 MLA-C01 or the GCP ML Engineer exam instead of learning Azure for its own sake.
The six-week study plan
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. Book the exam at the start of the week — a fixed date converts studying into finishing.
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.
Registration, cost and renewal
Register through Microsoft's exam delivery partner for online proctored or test-centre sittings. Unlike the fundamentals tier, associate certifications carry a renewal requirement — a free online renewal assessment on a recurring cycle — so passing is a subscription to staying current, not a one-off stamp.
Who should take AI-102 — and who shouldn't?
Take it if you build on Azure: developers in Microsoft-stack organisations, data engineers wiring AI services into pipelines, and consultants implementing Azure AI for clients. The wider credential landscape for your role is mapped in our software engineers guide.
Skip it if you do not code — AI-900 is the honest ceiling for non-developers, and that is fine — or if your platform is AWS or GCP, where the counterpart exams cover the same ground natively. And skip it for now if you are mid-way through learning Python: the exam will still be there when your code is.
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: treat AI-102 prep as six weeks of supervised portfolio-building with an exam at the end. That framing costs nothing, changes where the hours go, and leaves you with working artefacts plus a credential — instead of a credential and a watch history.
Verdict
For working developers on Azure, AI-102 is one of the few AI certifications whose preparation is directly the job: six weeks of building against Azure AI services, verified by a proctored associate credential that appears by name in job specs. Prepare hands-on, keep the mini-app you build, and book the exam before you feel fully ready — week-six readiness is a myth candidates use to postpone. 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.
Frequently asked questions
How long does it take to prepare for AI-102?
About six weeks part-time for a working developer with some Azure exposure — longer if Azure is new to you, and add the time to learn Python or C# first if you cannot yet code. Candidates coming straight from AI-900 with no coding background should not book until the programming foundation is real.
Do I need to take AI-900 before AI-102?
No — Microsoft sets no prerequisite, and working developers routinely skip the fundamentals exam. AI-900 earns its place only if Azure's AI stack is entirely new to you and you want a low-stakes first pass at the vocabulary before committing to associate-level prep.
Does AI-102 require coding?
Yes. The exam assumes you can read and complete code that calls Azure AI services — Python and C# are the standard paths — and its scenario questions presume implementation experience. It is the defining difference between AI-102 and the no-code AI-900 beneath it.
What is 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.
Does the AI-102 certification expire?
Associate-level Microsoft certifications require periodic renewal via a free online assessment — unlike fundamentals certificates, which do not expire. The renewal is unproctored and considerably lighter than the original exam.
Keeping this current. Course formats, prices, and certification exam fees change and vary by region. We review our guides regularly — this one was last updated in July 2026 — and we always recommend confirming the specifics on the provider's official page before you enrol.
Still deciding which certification to take?
Answer a few quick questions and get a personalized recommendation in under a minute.
Try the AI Certification Picker →