Question bank
Microsoft Certified: AI Business Professional
40 Microsoft Certified: AI Business Professional questions with the correct answer and a written explanation, organised across 3 exam domains.
- 40 questions
- 3 domains
- Foundation level
certilumQuestion index
Jump straight to any question. Answers and explanations are inline.
Draft and analyze business content by using AI
12 questions in this domain
- Question 1Single Choice
In the Microsoft ecosystem, what is the primary role of Microsoft 365 Copilot when drafting emails, reports, and presentations?
- AReplaces professional review and produces final documents
- BActs as augmentation layer for human-reviewed draftsCorrect answer
- CAutomatically sends emails and reports without user interaction
- DProvides grammar and spelling corrections only
Why this is the answer
Copilot is an assistive workflow, not a replacement for professionals; every draft must be reviewed and validated by a human.
- Question 2Multiple Select
Which elements should be included in an effective prompt when using artificial intelligence to draft a business email? Select all that apply.
- ATarget audienceCorrect answer
- BDesired toneCorrect answer
- CWord count or format constraintsCorrect answer
- DVersion number of the AI model
Why this is the answer
Effective business prompts include role, goal, audience, tone, and constraints; the AI model version is not relevant.
- Question 3Interactive
The prompt shown in the code block was sent to Copilot in Word. What important specification is missing from this prompt?
- AA defined role or target audienceCorrect answer
- BAn instruction to add emojis
- CA request for automatic sending
- DA request for AI model version
Why this is the answer
A strong business prompt should specify role, audience, tone, and constraints; here the audience and role are missing.
- Question 4Single Choice
After Microsoft 365 Copilot generates a draft report, what should a business professional do first before using the content?
- AVerify facts against trusted sourcesCorrect answer
- BChange font and color scheme
- CSend it to manager immediately
- DAsk Copilot to add more paragraphs
Why this is the answer
AI content may contain plausible but false data; factual verification is the first revision step.
- Question 5Multiple Select
Which dimensions should be evaluated when revising AI-generated content for professional business use? Select all that apply.
- AFactual accuracyCorrect answer
- BClarity and concisenessCorrect answer
- CBrand voice consistencyCorrect answer
- DColor of company logo
Why this is the answer
Revising AI output includes verifying facts, improving clarity, and aligning with brand voice; visual logo color is not part of text revision.
- Question 6Interactive
What is the main concern with the AI-generated sentence shown in the code block? Choose the correct answer.
- AStatistic needs verification and a sourceCorrect answer
- BSentence is too short
- CSentence contains a spelling error
- DTone is too formal
Why this is the answer
A specific statistic without a citation is a sign of possible hallucination; verify it before relying on it.
- Question 7Single Choice
Which Microsoft tool is designed to analyze structured data in a spreadsheet and create insights such as trends and outliers?
- ACopilot in ExcelCorrect answer
- BCopilot in Word
- CCopilot in PowerPoint
- DBing Chat Enterprise
Why this is the answer
Copilot in Excel works with tabular data to generate formulas, pivot tables, and insights.
- Question 8Multiple Select
Which tasks can Microsoft 365 Copilot support when analyzing business documents in Word and Outlook? Select all that apply.
- ASummarizing long reportsCorrect answer
- BIdentifying action items in email threadsCorrect answer
- CGenerating executive summariesCorrect answer
- DWriting code for external mobile apps
Why this is the answer
Copilot in Word summarizes documents and creates executive summaries; Copilot in Outlook extracts action items and sentiment from threads.
- Question 9Interactive
The analysis request shown in the code block is best handled by which Microsoft AI-assisted tool? Choose one answer.
- AMicrosoft Excel with CopilotCorrect answer
- BMicrosoft Word with Copilot
- CMicrosoft PowerPoint with Copilot
- DAzure AI Content Safety
Why this is the answer
The request asks to analyze structured data and create comparisons, which is the role of Copilot in Excel.
- Question 10Single Choice
What does a human-in-the-loop approach mean when artificial intelligence is used to create business content?
- AHuman reviews, validates, and corrects AI outputCorrect answer
- BAI automatically publishes content
- CHuman only checks word count
- DAI retrains itself after every draft
Why this is the answer
HITL keeps a human accountable for reviewing, approving, and correcting AI outputs.
- Question 11Multiple Select
Which criteria are commonly used to evaluate the quality of AI-generated business content? Select all that apply.
- AAccuracyCorrect answer
- BRelevanceCorrect answer
- CCompletenessCorrect answer
- DAI model processing time
Why this is the answer
Quality criteria include accuracy, relevance, completeness, tone, and readability; processing time is not a quality criterion.
- Question 12Interactive
What is missing from the AI-assisted workflow shown in the code block? Select the correct answer.
- AHuman review and approval stepCorrect answer
- BRequest for more emojis
- CAutomatic email attachment
- DA second AI-generated draft
Why this is the answer
The workflow skips the essential HITL step: a human must review and approve AI content before sending.
Manage prompts and conversations by using AI
16 questions in this domain
- Question 13Single Choice
Which statement best describes an outcome-focused prompt for a business task in Microsoft Copilot?
- AStructured instruction with desired result and formatCorrect answer
- BDetailed but no clear deliverable
- CBroad topic without specified outcome
- DMessage that leaves format to the AI
Why this is the answer
An outcome-focused prompt clearly defines the business result, context, and format so the output is measurable and decision-ready.
- Question 14Multiple Select
Which components should be included in an outcome-focused prompt for a business task?
- ABusiness objectiveCorrect answer
- BContext and constraintsCorrect answer
- COutput format and audienceCorrect answer
- DIrrelevant supporting data
Why this is the answer
Outcome-focused prompts include objective, context, constraints, audience, and format. Irrelevant data dilutes focus and should be avoided.
- Question 15Interactive
Review the prompt in the code block. Which key element is missing?
- ASpecific desired outcome or output formatCorrect answer
- BA polite greeting
- CA list of competitors
- DA longer sentence
Why this is the answer
The prompt names a topic but does not state the deliverable, making it vague rather than outcome-focused.
- Question 16Single Choice
What is the main benefit of specifying an output format in a prompt?
- ATurns the prompt into a measurable deliverableCorrect answer
- BIncreases the model's creativity
- CReplaces the need for a business objective
- DMakes the prompt longer and more complex
Why this is the answer
Specifying format, such as a table, list, or email, defines the deliverable's shape and makes the output measurable and reviewable.
- Question 17Multiple Select
Which prompt techniques should a business professional use to guide AI responses?
- AInstructions, examples, and formattingCorrect answer
- BSystem messagesCorrect answer
- CChanging the model temperature
- DAdding irrelevant context
Why this is the answer
Instructions, examples, formatting, and system messages are core prompt techniques; temperature is not a language-based prompt control.
- Question 18Interactive
The text in the code block is configured before a session starts. What is this prompt element called?
- ASystem messageCorrect answer
- BOne-time instruction
- CFew-shot example
- DOutput format specifier
Why this is the answer
This persistent session-level directive defines the persona and behavioral boundaries, so it is a system message.
- Question 19Single Choice
What is the main purpose of including examples in a prompt?
- ADemonstrate the expected pattern, tone, or structureCorrect answer
- BExpand the token limit
- CReplace the system message
- DEliminate all need for instructions
Why this is the answer
Few-shot examples show the model the desired pattern, tone, or structure, especially when output must match a template.
- Question 20Multiple Select
Which actions are part of iterative prompt refinement?
- AUsing follow-up turns to narrow scopeCorrect answer
- BIncorporating user feedbackCorrect answer
- CRestarting from scratch after every minor error
- DEvaluating output against the original goalCorrect answer
Why this is the answer
Iterative refinement uses follow-up turns, feedback, and evaluation. Restarting after every minor error discards valuable context.
- Question 21Interactive
The follow-up prompt in the code block is an example of what kind of instruction?
- ASpecific, actionable follow-upCorrect answer
- BVague user complaint
- CContradictory instruction
- DConversation reset command
Why this is the answer
This follow-up gives precise corrective details, making it an effective refinement rather than vague feedback.
- Question 22Single Choice
What is the scope of a system message in an AI conversation?
- AApplies to the entire sessionCorrect answer
- BApplies only to the next user turn
- CApplies only to formatting
- DApplies only to one example
Why this is the answer
A system message is persistent and applies to all turns, unlike single-turn instructions.
- Question 23Multiple Select
Which communication parameters can be controlled through natural language prompts?
- AToneCorrect answer
- BStyleCorrect answer
- CAudience and output formatCorrect answer
- DAPI model temperature
Why this is the answer
Tone, style, audience, and output format are controlled in prompt language. Temperature is a technical model setting.
- Question 24Interactive
The prompt in the code block contains a role, an instruction, and a format. What technique is being applied?
- ARole-based prompt controlCorrect answer
- BConversation reset
- CGrounding data selection
- DFine-tuning the model
Why this is the answer
Assigning a role and specifying style, audience, and format is role-based prompt control.
- Question 25Single Choice
What is conversation context in an AI assistant?
- AInformation the model uses to interpret the current requestCorrect answer
- BThe model's pre-training dataset
- CThe user's local file system
- DThe AI's internal model weights
Why this is the answer
Conversation context includes the system prompt, prior messages, user goal, and uploads used to interpret the current request.
- Question 26Multiple Select
When is a full conversation reset appropriate?
- AThe business goal changes completelyCorrect answer
- BThe conversation has drifted and context is corruptedCorrect answer
- CThe context window is nearly fullCorrect answer
- DThe AI makes one small formatting mistake
Why this is the answer
Reset is for major changes, corrupted context, or token limits. Minor errors should be corrected with follow-up prompts.
- Question 27Interactive
After a supplier negotiation conversation, the user sends the prompt in the code block. What is this an example of?
- AConversation driftCorrect answer
- BGrounding
- CSystem message
- DIterative refinement
Why this is the answer
The new topic is unrelated to the original negotiation, so the conversation has drifted away from intent.
- Question 28Single Choice
What is grounding data in AI prompt management?
- AAuthoritative reference materials used to anchor responsesCorrect answer
- BUser feedback collected from ratings
- CThe AI model's parametric memory
- DA list of system messages
Why this is the answer
Grounding anchors output to selected authoritative sources during inference, improving reliability and auditability.
Understand generative AI fundamentals
12 questions in this domain
- Question 29Single Choice
What is the primary characteristic that differentiates generative AI from predictive AI?
- AIt creates new, previously unseen content by learning patterns from training dataCorrect answer
- BIt assigns inputs to a fixed set of predefined labels
- CIt retrieves the most relevant existing document from a database
- DIt guarantees that every response is factually accurate
Why this is the answer
Generative AI synthesizes original content based on statistical patterns in training data, rather than classifying, retrieving, or guaranteeing factual accuracy.
- Question 30Multiple Select
Which of the following are examples of foundation models? Select all that apply.
- AGPT-4Correct answer
- BClaudeCorrect answer
- CLLaMACorrect answer
- DA rules-based workflow engine
Why this is the answer
GPT-4, Claude, and LLaMA are large pre-trained foundation models. A rules-based workflow engine is traditional automation, not a foundation model.
- Question 31Interactive
The code block contains a Python function that estimates LLM token usage by splitting text on spaces. Why is this estimate unreliable?
- ALLMs split text into subword tokens, so a single word may be one token or several tokensCorrect answer
- BLLMs do not process text and instead use only images
- CToken counts depend only on the number of sentences in the text
- DWhitespace is the exact method used by Azure OpenAI to count tokens
Why this is the answer
Tokenization breaks text into subword units. Counting whitespace-separated words does not reflect how language models create tokens, which can lead to inaccurate cost estimates.
- Question 32Single Choice
What does a model's context window represent?
- AThe maximum number of tokens that can be used for input and output combinedCorrect answer
- BThe number of training documents the model was pre-trained on
- CThe physical storage capacity of the GPU cluster hosting the model
- DThe total size of the model's token vocabulary
Why this is the answer
The context window is a hard token limit that includes the system prompt, examples, conversation history, and the model's generated output.
- Question 33Multiple Select
Which of the following consume token capacity inside a model's context window? Select all that apply.
- ASystem promptCorrect answer
- BConversation historyCorrect answer
- CThe model-generated responseCorrect answer
- DThe number of active users in the tenant
Why this is the answer
The context window is a shared budget that includes the system prompt, chat history, few-shot examples, and the generated output. User count is not a token factor.
- Question 34Interactive
The code block shows a JSON definition for a workflow. Which category best describes this implementation?
- ATraditional business automationCorrect answer
- BGenerative AI
- CPredictive analytics
- DMultimodal modeling
Why this is the answer
The code uses an explicit trigger, a condition, and a predefined action. This is deterministic rule-based automation, not generative AI.
- Question 35Single Choice
Which Microsoft tool should you use to design a custom AI copilot that queries SharePoint and backend systems?
- AMicrosoft Copilot StudioCorrect answer
- BCopilot in Word
- CMicrosoft Copilot consumer chat
- DPower Automate
Why this is the answer
Microsoft Copilot Studio supports custom AI copilots that orchestrate multiple data sources and publish to channels like Teams, web portals, and internal applications.
- Question 36Multiple Select
Which of the following are valid generative AI business use cases? Select all that apply.
- ASummarizing meeting transcriptsCorrect answer
- BDrafting personalized customer response emailsCorrect answer
- CGenerating multiple product description variationsCorrect answer
- DForecasting next quarter's revenue from historical sales data
Why this is the answer
Summarization, drafting, and content variation are generative AI use cases. Revenue forecasting is a predictive analytics task, not a content generation task.
- Question 37Interactive
The code block shows an Azure OpenAI API configuration. Which parameter should be adjusted to make the model's responses more deterministic and less creative?
- AtemperatureCorrect answer
- Bmax_tokens
- Cmodel
- Dmessages
Why this is the answer
A lower temperature produces more deterministic responses, while max_tokens controls output length and messages contain the prompt context.
- Question 38Single Choice
What is hallucination in generative AI?
- AProducing confident, fluent content that is factually incorrectCorrect answer
- BGenerating output only from a verified knowledge base
- CA training method that uses labeled data
- DThe process of splitting text into tokens
Why this is the answer
Hallucination occurs when a generative model produces plausible but factually incorrect content because it is optimizing for likely language patterns rather than verified truth.
- Question 39Multiple Select
Which of the following are Microsoft responsible AI principles? Select all that apply.
- AFairnessCorrect answer
- BReliability and safetyCorrect answer
- CPrivacy and securityCorrect answer
- DMaximum revenue growth
Why this is the answer
Microsoft's responsible AI principles include fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.
- Question 40Interactive
The code block shows a Python function that assigns a label based on a keyword. Which category best describes this approach?
- ADeterministic rule-based classificationCorrect answer
- BGenerative AI
- CLarge language model
- DFoundation model
Why this is the answer
The code applies a fixed, predefined rule to produce a label. It does not generate new content, so it is not generative AI.
Frequently asked questions
Everything about the Microsoft Certified: AI Business Professional question bank and how to use it.