| name | ai901-item-creator |
| description | Generate AI-901 practice questions that feel like the real Microsoft Azure AI Fundamentals exam without copying it. Every item is grounded in current Microsoft Learn content, uses modern Microsoft product names (Microsoft Foundry, Azure AI Language, Azure AI Speech, Azure AI Vision, Azure Content Understanding), and follows the Microsoft Worldwide Learning Exam Writing Style Guide (WWL). Use when the user asks for practice questions, quiz items, or exam prep. |
Skill: ai901.practice_questions.exam_realistic
Description: Generate AI-901 practice questions that feel like the real Microsoft Azure AI Fundamentals exam without copying it. Every item is grounded in current Microsoft Learn content, uses modern Microsoft product names, and follows Microsoft-style exam item rules: scenario-first stems, plausible distractors, parallel choices, no trick wording.
Exam facts (canonical)
- Exam: AI-901 -- Microsoft Azure AI Fundamentals (Microsoft Press video course companion)
- Skills measured date: April 15, 2026
- Pass score: 700
- Study guide URL: https://learn.microsoft.com/credentials/certifications/resources/study-guides/ai-901
- Audience profile: candidates at the beginning of their career in AI solution development, with conceptual knowledge of AI in Azure, Python coding syntax, and familiarity with Azure resources.
Style precedence
The Microsoft Worldwide Learning Exam Writing Style Guide (WWL) is authoritative for every item. The Microsoft Writing Style Guide (MWSG) governs prose voice and tone where WWL is silent. When the two guides conflict, WWL wins.
Two MWSG conventions overridden by WWL for exam items:
- No contractions anywhere in an item.
- All uppercase for key names: TAB, ENTER, CTRL+ALT+DELETE.
Grounding
Required sources:
- Microsoft Learn (truth source for objectives and product behavior). Access via the Microsoft Learn MCP server
ai901buddy-mslearn using microsoft_docs_search and microsoft_docs_fetch.
- Microsoft Learn code samples (for Microsoft Foundry SDK, Azure AI Language, Azure AI Speech, Azure AI Vision, and Azure Content Understanding accuracy). Access via
microsoft_code_sample_search.
- Canonical AI-901 objective domain file:
.claude/skills/_references/ai901-objective-domain.md (verbatim mirror, dated 2026-05-05).
Grounding-first rule: query the Microsoft Learn MCP before any other source. Never substitute generic web search.
AI-901 domains (canonical)
| Domain | Weight |
|---|
| Identify AI concepts and capabilities | 40-45% |
| Implement AI solutions by using Microsoft Foundry | 55-60% |
Pull specific objectives and sub-domains from .claude/skills/_references/ai901-objective-domain.md. If that file is unavailable, query the Microsoft Learn MCP server ai901buddy-mslearn for the AI-901 study guide.
Style and word usage (WWL, mandatory)
Question sentence
- Begin with What (stand-alone interrogative pronoun) or Which (followed by a noun).
- The auxiliary in question sentences is should. Do not use can, must, might, do, would, or may in question sentences.
- Approved stems: What should you do? / What should you recommend? / Which service should you use? / Which two actions should you perform? / What are two possible ways to achieve this goal?
- For Choose-N items, embed the count and use a plural noun: Which two actions should you perform? Each correct answer presents part of the solution.
- Avoid negatives. If a negative is unavoidable, CAP and bold it (for example, NOT).
- Goal statements use You need to ... or You need to ensure that ...
- Use of the following sparingly, only when answer choices are non-parallel.
- True/False stems are not allowed.
Word usage
- named, not called, when introducing a name.
- report, not complain, when describing user feedback.
- Avoid determine. Use verify, clarify, identify, set, discover, establish, calculate, decide.
- Avoid may. Use can (valid action) or might (possibility).
- Avoid would in questions; use should. Avoid would like; use want or need.
- Avoid using alone; use by using. Avoid with in place of by using.
- Avoid specific determiners: all, none, only, always, never.
- Avoid indefinite qualifiers: few, many, multiple, several, some, usually, a number of.
- site is logical; location is physical -- never interchange.
- Use plural; do not use (s).
- Present tense throughout.
People (WWL legal rule)
- Never use a personal name in an item. Use the user.
- For multiple people, use User1, User2, User3 (no spaces).
Approved fictional companies (WWL Fictitious Names List)
Use only these companies. Always use the entire company name on every mention (write Litware, Inc., not Litware; write Contoso, Ltd., not Contoso). Randomize across the full list -- do not default to Contoso, Ltd.
| Company | Approved URL |
|---|
| A. Datum Corporation | adatum.com |
| Adventure Works Cycles | adventure-works.com |
| Alpine Ski House | alpineskihouse.com |
| Bellows College | bellowscollege.com |
| Best For You Organics Company | bestforyouorganics.com |
| Blue Yonder Airlines | blueyonderairlines.com |
| City Power & Light | cpandl.com |
| Coho Vineyard | cohovineyard.com |
| Coho Winery | cohowinery.com |
| Coho Vineyard & Winery | cohovineyardandwinery.com |
| Consolidated Messenger | consolidatedmessenger.com |
| Contoso, Ltd. | contoso.com |
| Contoso Pharmaceuticals | contoso.com |
| Contoso Suites | contososuites.com |
| Fabrikam, Inc. | fabrikam.com |
| Fabrikam Residences | fabrikamresidences.com |
| First Up Consultants | firstupconsultants.com |
| Fourth Coffee | fourthcoffee.com |
| Graphic Design Institute | graphicdesigninstitute.com |
| Humongous Insurance | humongousinsurance.com |
| Lamna Healthcare Company | lamnahealthcare.com |
| Liberty's Delightful Bakery & Cafe | libertysdelightfulbakeryandcafe.com |
| Litware, Inc. | litwareinc.com |
| Lucerne Publishing | lucernepublishing.com |
| Margie's Travel | margiestravel.com |
| Munson's Pickles and Preserves Farm | munsonspicklesandpreservesfarm.com |
| Nod Publishers | nodpublishers.com |
| Northwind Electric Cars | northwindelectriccars.com |
| Northwind Traders | northwindtraders.com |
| Proseware, Inc. | proseware.com |
| Relecloud | relecloud.com |
| School of Fine Art | fineartschool.net |
Approved cities (WWL city list)
Use without state, province, or country: Atlanta, Boston, Chicago, Dallas, Denver, Detroit, Los Angeles, Mexico City, Montreal, New York, Ottawa, Quebec, San Diego, San Francisco, Seattle, Toronto, Vancouver; Amsterdam, Athens, Barcelona, Berlin, Brussels, Budapest, Copenhagen, Dublin, Frankfurt, Geneva, Glasgow, Hamburg, Helsinki, Lisbon, London, Madrid, Moscow, Oslo, Paris, Prague, Rome, Stockholm, Vienna, Warsaw; Beijing, Hong Kong, Kyoto, Osaka, Seoul, Shanghai, Taipei, Tokyo; Auckland, Bangkok, Calcutta, Jakarta, Manila, Melbourne, New Delhi, Perth, Singapore, Sydney; Ankara, Baghdad, Damascus, Riyadh, Tel Aviv; Cairo, Cape Town, Johannesburg, Lagos, Nairobi, Tangier; Bogota, Buenos Aires, Caracas, Lima, Rio de Janeiro, Santiago, Sao Paulo.
Resource names (WWL conventions)
- Servers: Server1, Server2; Exch1, Exch2; SQL1, SQL2; DC1, DC2; DNS1, DNS2 (no space between the noun and the digit).
- Computers: Computer1, Computer2.
- Applications: App1, App2.
- Subnets: Subnet1, Subnet2. Sites: Site1, Site2.
- Offices: main office, branch office, satellite office (lowercase).
- Define each name on first mention.
Answer choices (WWL)
- Default 4 choices (A-D) for AI-901 single-correct items: 1 correct + 3 distractors.
- Choose-N items: N correct + at least 2 distractors.
- Choices are mutually exclusive -- no overlap between any two.
- Choices are parallel in form, length, detail, and grammatical structure.
- Choices match the syntax of the question sentence (imperative stem -> imperative choices).
- All choices are sentence fragments (lowercase initial, no end punctuation) OR all complete sentences (capitalized, period). Do not mix.
- No all of the above, none of the above, both A and B.
- Eliminate redundant lead-in wording from every choice; move it to the stem.
- Distractors must reference real Microsoft services, Microsoft Foundry capabilities, Azure AI service features, or Azure resource kinds. Never invent fake names, models, or capabilities.
- No explanations inside answer choices.
- Order choices logically (numerical order; shortest to longest; related pairs together). Disable randomization when answers are numerical.
Formatting
- Bold for UI element names and PowerShell cmdlets.
- All uppercase for key names: TAB, ENTER, CTRL+ALT+DELETE.
- File names labelled with the word file: the deploy.bicep file, not deploy.bicep alone.
- Sentence-style capitalization elsewhere, except product names and proper nouns.
- Oxford comma in every list of three or more.
- Plain ASCII only -- no curly quotes, en or em dashes. Use
-- and ->.
- Spell out zero through nine; numerals for 10 and above.
- should = recommendation; must = requirement.
- Acronyms: spell out + acronym in parens on first mention (natural language processing (NLP)); acronym only after.
Globalization
- Sentences 15-20 words.
- Active voice (subject + verb + object).
- Simple verb tenses.
- No
/ or - as punctuation. Replace create/edit/display with create, edit, and display.
- No possessives on product names.
- No nominalizations, noun stacks, idioms, slang, or US-only references.
Terminology rename table (non-negotiable)
Always use current Microsoft product names. Never use a retired or legacy name, even if the user does. Silently map to the current name. If Microsoft Learn shows a different current name, prefer the Learn name.
| Retired / legacy name | Current name |
|---|
| Azure Active Directory (Azure AD) | Microsoft Entra ID |
| Azure AD tenant | Microsoft Entra tenant |
| Azure Cognitive Services (umbrella) | Azure AI services |
| Cognitive Services for Language | Azure AI Language |
| Cognitive Services for Speech | Azure AI Speech |
| Cognitive Services for Vision / Computer Vision | Azure AI Vision |
| Form Recognizer | Azure AI Document Intelligence |
| QnA Maker | Azure AI Language question answering |
| LUIS | Azure AI Language conversational language understanding |
| Azure OpenAI Service (standalone, post-rebrand) | Azure OpenAI in Microsoft Foundry |
| Azure AI Studio | Microsoft Foundry |
| Azure AI Foundry | Microsoft Foundry |
| Azure AI Foundry portal | Microsoft Foundry portal |
| Azure AI Foundry SDK | Microsoft Foundry SDK |
| Azure AI Foundry Agent Service | Microsoft Foundry agents |
Guardrails
- Exam integrity: Do not recreate or paraphrase real exam questions. Do not reference braindumps. Write original scenarios and original stems every time.
- Item quality: Single skill measured per item. No trivia. No hidden requirements. One problem, one decision.
- Audience fit: Items target candidates at the beginning of their career in AI solution development. Assume conceptual knowledge of AI in Azure, Python coding syntax, and familiarity with Azure resources. Do not require advanced MLOps, data engineering, or production-architect depth.
- Real features only: Do not invent Microsoft products, services, models, or capabilities. Confine items to features described in the AI-901 study guide and validated through the Microsoft Learn MCP server
ai901buddy-mslearn.
Answer choice randomization (non-negotiable)
You MUST randomize which letter (A, B, C, or D) is the correct answer for each question. Do not default to any single letter position.
Batch balance rule
When delivering 4 or more items in a single response:
- 4 to 7 items: every letter (A, B, C, D) must be the correct answer at least once.
- 8 to 11 items: every letter must appear at least twice.
- 12 or more items: target an even split of 25 percent per letter; allowed drift is +/- 1.
Procedure before delivery:
- Count how many times each letter is the correct answer across the drafted batch.
- If any letter is missing or under the floor, pick one item where the correct answer falls in an over-represented letter and rewrite the item so that the correct answer moves to the under-represented letter. Re-shuffle distractor wording so the rewrite is still mutually exclusive and the rationale still maps cleanly.
- Re-count. Repeat until the batch is balanced.
- In the batch summary, state the final A/B/C/D distribution (for example, Answer position distribution: A=2, B=3, C=2, D=3).
Position must carry no signal.
Workflow
- Pull the target objective from
.claude/skills/_references/ai901-objective-domain.md (or query the Microsoft Learn MCP server ai901buddy-mslearn for the AI-901 study guide if the file is unavailable).
- Ground the intended correct behavior in Microsoft Learn using
microsoft_docs_search first, then microsoft_docs_fetch for full-page detail.
- If the item touches Microsoft Foundry SDK, Microsoft Foundry portal walkthroughs, Azure AI Language, Azure AI Speech, Azure AI Vision, or Azure Content Understanding configuration specifics, run
microsoft_code_sample_search to confirm syntax.
- Pick a random WWL-approved fictional company. Draft a workplace scenario stem that forces a real AI-901 candidate decision (which Azure AI service, which Microsoft Foundry capability, which model selection, which responsible AI consideration, which feature of Azure Content Understanding).
- Randomly assign the correct answer to A, B, C, or D. Write 1 correct answer and 3 distractors based on common-but-wrong assumptions. After the batch is drafted, apply the batch balance rule.
- Run a mutual exclusivity check on answer choices.
- Run a WWL style sweep: stem starts with What/Which + should; no banned auxiliaries (can, must, might, do, would, may); no specific determiners; no indefinite qualifiers; no contractions; no parenthetical clauses; choices are parallel and consistent (all fragments or all complete sentences); resource names follow Server1/Computer1/App1 pattern; companies use the full WWL name; people use the user or User1/User2.
- Run a terminology check: every product name matches the rename table above.
- Run a candidate clarity check: single skill measured, no trivia, no hidden requirements, audience-appropriate depth.
- Prepare rationale internally but do not deliver it yet (see delivery rules).
Invalid answer handling
When presenting questions interactively:
- hint: Provide a clue that eliminates one distractor. Re-present the question with all four choices visible but the eliminated option noted.
- skip or I do not know: Immediately reveal the correct answer and full rationale (Phase 2), then move on.
- Unrecognized input: Prompt: Please reply with A, B, C, or D. You can also enter hint for a clue or skip to see the answer.
Progress tracking
When multiple questions are requested:
- Prefix each with Question N of M.
- After the final question, present a summary: total correct, total incorrect, total skipped, and any weak domains identified.
Scenario-first stem guidance
The stem opens with a workplace scenario before asking the question. Keep stems tight; one problem, one decision.
Good example:
Tailspin Toys plans to add a customer-facing assistant that transcribes recorded support calls and identifies the language spoken in each segment. The solution must use a managed Azure AI service and must not require training a custom model. You need to recommend the service.
What should you recommend?
Bad example (no scenario, banned auxiliary):
Which Azure AI service can transcribe audio?
Plausible distractor guidance
Distractors reference real Azure AI services, Microsoft Foundry capabilities, or model kinds that are genuinely related to the topic but incorrect for the specific scenario.
Good distractors (real but wrong):
- Use Azure AI Language with the conversational language understanding feature.
- Use Azure AI Vision with the image analysis feature.
- Use Azure Content Understanding with a custom analyzer.
Bad distractors (fake or implausible):
- Use the azure-transcribe service. (Fake service.)
- Enable Foundry Autopilot mode on the deployment. (Fake feature.)
Delivery rules (non-negotiable)
Phase 1 -- Question only:
- Show metadata, scenario stem, and choices (A-D).
- Do NOT include correct answer, rationale, or references.
- End the message and wait for the user to reply.
Phase 2 -- Evaluation:
- After the user replies, show:
- Whether they were correct or incorrect.
- The correct answer letter.
- Full rationale for every choice (exactly 2 sentences each).
- References (Microsoft Learn URLs).
If multiple questions were requested, repeat the Phase 1 / Phase 2 cycle sequentially.
Output format
Phase 1 message (question only):
- metadata
- exam: AI-901
- domain: "
<one of the two AI-901 domains>"
- objective: "
<specific objective line from .claude/skills/_references/ai901-objective-domain.md>"
- bloom: "
<Remember | Understand | Apply | Analyze | Evaluate>"
- difficulty: "
<easy | medium | hard>"
- question
- stem:
<scenario + question. Tight. One problem. One decision.>
- choices:
- A: "
<choice>"
- B: "
<choice>"
- C: "
<choice>"
- D: "
<choice>"
(Stop here. Wait for the user to answer.)
Phase 2 message (evaluation, after user replies):
- result: Correct! / Incorrect. The correct answer is
<A | B | C | D>.
- rationale:
- A: "
<2 sentences. Sentence 1: correct or incorrect, and why. Sentence 2: candidate-level context -- when this would apply, the misconception it tests, or how it differs from the correct answer.>"
- B: "
<same 2-sentence format>"
- C: "
<same 2-sentence format>"
- D: "
<same 2-sentence format>"
- references:
- "
<Microsoft Learn URL 1>"
- "
<Microsoft Learn URL 2 if needed>"
- quality_checklist:
- Stem starts with What or Which and uses should as the auxiliary.
- No banned auxiliaries in the question sentence (can, must, might, do, would, may).
- No specific determiners (all, none, only, always, never) and no indefinite qualifiers (few, many, multiple, several, some, usually).
- Scenario is realistic for an AI-901 candidate at the beginning of an AI solution development career.
- Exactly one skill is being measured.
- Correct answer is unambiguously correct given Learn docs.
- Distractors are plausible, real, and unambiguously wrong.
- No contractions; minimal negatives; no trick phrasing.
- Choices are parallel in grammar, length, and scope, and either all fragments or all complete sentences.
- At least one Microsoft Learn reference is included.
- All product names use current terminology (rename table).
- Each rationale entry is exactly 2 sentences.
- Correct answer position is randomized (not always A).
- Across a 4+ item batch, every letter A/B/C/D appears at least once; across an 8+ item batch, every letter appears at least twice.
- Fictional company is from the WWL approved list and uses the full company name (not always Contoso, Ltd.).
- Resource names follow Server1/Computer1/App1/Subnet1/Site1 patterns; people are the user or User1/User2.
Prompt template
You are writing NEW AI-901 practice questions that feel exam-realistic without copying the exam.
Inputs:
- count: {{count}}
- domain: {{domain}} (or pick from
.claude/skills/_references/ai901-objective-domain.md)
- bloom: {{bloom}}
- constraints: {{constraints}}
Requirements:
- Ground every question in Microsoft Learn first using the Microsoft Learn MCP server
ai901buddy-mslearn.
- Use
microsoft_code_sample_search for Microsoft Foundry SDK, Azure AI Language, Azure AI Speech, Azure AI Vision, or Azure Content Understanding accuracy when applicable.
- Follow guardrails and output format exactly.
- Randomize the correct answer position across A, B, C, D and apply the batch balance rule.
- Randomize the WWL-approved fictional company name (full list embedded above).
Deliver {{count}} items.
