examifyr

Readiness test · Microsoft · Free · No sign-up required

Microsoft AI-901 Readiness Test — Azure AI Fundamentals

AI-901 replaced AI-900 on 30 June 2026, and it is not the same exam for the same person: it asks for Python syntax and is built around Microsoft Foundry. This free readiness test samples across both skills-measured domains, weighted by the published blueprint, and tells you where you stand before you book.

~30 questions sampled across every exam domain, weighted by the official blueprint · no sign-up

Instant resultsTopic breakdownWeak area detectionPass likelihood

What this test covers

AI-901 leads to the Microsoft Certified: Azure AI Fundamentals credential, and its larger domain by some distance is Implement AI solutions by using Microsoft Foundry at 55-60% — deploying a model, prompting it, building a lightweight chat client and a single-agent solution, and wiring up text, speech, vision and information extraction. The other 40-45%, Identify AI concepts and capabilities, covers responsible AI, how generative models work, choosing a model, and recognising which workload a problem actually is. That split is what changed: AI-900 asked you to describe these capabilities and required no coding, while AI-901 asks you to implement them and assumes Python. Unlike a flat question dump, this readiness test mirrors the blueprint — questions are drawn across both domains in proportion to their published weight, so your score reflects what the exam emphasises. When you finish you get a readiness score from 0-100, a pass-likelihood band, and a per-domain breakdown showing which of the two to study next. Fresh questions every attempt. No sign-up, no email required.

  • Implement AI solutions by using Microsoft Foundry — 55-60%
  • Identify AI concepts and capabilities — 40-45%

How it works

01

Choose a level

Easy, Medium, or Hard for the full readiness test. Every level is sampled across the exam domains in proportion to their published weight.

02

Answer questions

One question at a time. Click an answer and get instant feedback with a full explanation.

03

Get your score

See your readiness score (0–100), pass likelihood, and a per-domain breakdown of your weakest exam areas.

What the questions look like

6 worked examples, with the reasoning spelled out. These are written for this page rather than taken from the live bank — showing a real question would spend it as a measurement for anyone who read this first.

Identify AI concepts and capabilities — 40-45%

Which of these is an example of a classification workload?

  1. APredicting next month's sales revenue for each store
  2. BPredicting whether an incoming email is spamCorrect
  3. CGrouping customers by purchasing behaviour without any labels
  4. DGenerating a product description from a list of bullet points
WhyClassification predicts which category something belongs to — spam or not spam. Predicting a number such as revenue is regression, grouping unlabelled data is clustering, and producing new text is generative AI.

Identify AI concepts and capabilities — 40-45%

A loan-approval model approves applicants from one region far less often than applicants with similar finances from elsewhere. Which Microsoft Responsible AI principle is most directly at stake?

  1. AReliability and safety
  2. BPrivacy and security
  3. CInclusiveness
  4. DFairnessCorrect
WhyFairness is about AI systems treating similar people similarly. Here, people with comparable finances get different outcomes because of where they live. Reliability, privacy and inclusiveness are real principles too, but none of them names this harm as directly.

Identify AI concepts and capabilities — 40-45%

What does a language model's context window limit?

  1. AHow many tokens the model can take into account in a single requestCorrect
  2. BHow many requests the deployment can serve each minute
  3. CHow many languages the model was trained to understand and produce
  4. DHow long the model remembers earlier conversations across separate sessions and days
WhyThe context window is the amount of text, measured in tokens, the model can consider at once in one request. Request rate is a deployment quota, training languages are a property of the model, and a model keeps no memory between separate sessions unless the application supplies it again.

Implement AI solutions by using Microsoft Foundry — 55-60%

In a Python app that calls a chat model deployed in Microsoft Foundry, what is the role of the system message?

  1. AIt stores the credential the app uses to authenticate
  2. BIt selects which model deployment handles the request
  3. CIt sets the assistant's behaviour, tone and constraintsCorrect
  4. DIt holds the model's earlier replies so it has a memory of the conversation
WhyThe system message instructs the model how to behave for the whole conversation — role, tone, what to refuse. Credentials and the deployment are part of the client and request configuration, and earlier replies are carried as assistant messages in the history, not in the system message.

Implement AI solutions by using Microsoft Foundry — 55-60%

Before releasing a Foundry chat app grounded in your product documentation, you want evidence that its answers stay faithful to that documentation. What should you use?

  1. AA higher temperature setting, so that answers vary more between runs
  2. BAn evaluation that scores groundedness on a test datasetCorrect
  3. CThe deployment's tokens-per-minute quota, raised to its maximum
  4. DThe model's training cut-off date, checked against the documentation date
WhyA groundedness evaluation checks each answer against the source content it was given and scores how well it is supported, across a dataset rather than a handful of spot checks. Temperature and quota change behaviour and capacity, not faithfulness, and the cut-off date says nothing about your own documentation.

Implement AI solutions by using Microsoft Foundry — 55-60%

You are creating a single agent in the Foundry portal that must look up the current status of a customer's order in your company's order system. What must you give the agent?

  1. AA tool it can call to fetch order dataCorrect
  2. BA much longer system prompt that lists every current order and its status
  3. CA larger model deployment, so that the agent has more general knowledge
  4. DA higher max tokens limit, so that the agent can write longer answers
WhyLive order status exists only in your system, so the agent needs a tool — an action or function it can call — to fetch it at the time of the question. A prompt listing orders is stale as soon as it is written, and model size or output length adds no access to data the model has never seen.

Frequently asked questions

Did AI-901 replace AI-900?
Yes. AI-900 retired on 30 June 2026 and AI-901 replaced it, under the same certification name — Microsoft Certified: Azure AI Fundamentals. They are not interchangeable to prepare for: AI-900 was a describe-it exam that required no coding, and AI-901 asks for Python syntax and familiarity with Azure resources, with implementation in Microsoft Foundry as its largest domain.
What is the passing score for AI-901?
A scaled score of 700 out of 1000. That is not 70% of questions answered correctly — Microsoft scales scores so a given number means the same level of competence across exam versions, questions are not weighted equally, and some may be unscored. You cannot convert it into a number of questions you need right, which is why a readiness score is best read as a per-domain diagnosis.
How many questions are on AI-901, and how long is it?
Microsoft publishes neither a question count nor a duration for AI-901 on its exam page, so any specific figure you find elsewhere is an estimate rather than an official number. We would rather tell you a figure does not exist than repeat one that does not.
Does the Azure AI Fundamentals certification expire?
No. Microsoft expires its associate, expert and specialty certifications annually, with a free online renewal assessment — but fundamentals certifications do not expire and need no renewal. Once you pass AI-901 the credential is permanent.
Do I need coding experience for AI-901?
Some. Microsoft’s audience profile asks for knowledge of Python coding syntax and programming techniques, plus familiarity with Azure resources. This is the clearest break from AI-900, whose page said data science and software engineering experience were not required — so if you were told this exam is non-technical, that advice predates June 2026.
Is AI-901 enough to work as an AI engineer on Azure?
It is the entry point rather than the destination. AI-901 is fundamentals-level and proves you understand the capabilities and can build lightweight solutions with them; AI-103, Developing AI Apps and Agents on Azure, is the associate-level exam for the engineering role itself, and its heaviest domain is generative AI and agentic solutions at 30-35%.
Is it free? Do I need to sign up?
Completely free, no account needed. Start the test, answer the questions, and get your domain-weighted readiness score instantly.

Related