Microsoft AB-731 Practice Test 1
Question 1
At Bridgewater Logistics, a multinational enterprise launches AI pilots in several departments without common oversight. Evidence: Different owners approve conflicting risk exceptions. The business goal is to create cross-functional governance accountability. The sponsor needs an evidence-based diagnosis rather than a guessed fix. Which initial investigation best identifies the controlling constraint before any solution is chosen?
- Inventory licenses, role requirements and expected daily tasks
- Identify executive sponsor, legal, security, HR and data stakeholders
- Catalog document owners, versions and approval status
- Assess data protection, residency, licensing and supplier assurance
Explanation
Correct Answer
B:
The conflicting AI-risk exceptions show that departments lack a shared accountability structure. Identify executive sponsors and responsible legal, security, HR, and data stakeholders, then establish who reviews, approves, and escalates exceptions before further rollout decisions.
Incorrect Answers
A:
Licensing determines which users can access Copilot capabilities, but it cannot resolve conflicting ownership of AI-risk exception decisions. Named sponsors and cross-functional approval roles are required.
C:
Document ownership and version provenance help explain which source is authoritative, not who may approve enterprise AI-risk exceptions across departments. Establish a decision and escalation structure.
D:
Supplier contract and model-risk due diligence assesses external vendors; it does not assign internal legal, security, HR, and data stakeholders authority over conflicting exception approvals.
Question 2
At Harborline Products, a client success leader asks for the source of AI-generated renewal totals. Evidence: Some underlying work records were revised after the first draft. The business goal is to make source identities and versions visible in answers. The proposed solution must respect documented authority and organizational controls. Which approach most directly meets the stated business objective within the approved authority?
- Limit graph queries to necessary users and approved entities
- Separate web and work findings with dated citations
- Request dated source-linked calculations and show missing data
- Extract text using suitable Vision tooling and validate key amounts
Explanation
Correct Answer
C:
Some underlying work records were revised after the first draft. Request dated source-linked calculations and show missing data. Dated source links let a reviewer trace each reported figure to the correct data version and identify missing inputs.
Incorrect Answers
A:
Graph permission scoping protects private collaboration relationships. It does not show which revised record version supplied an AI-generated renewal total; dated source-linked calculations do.
B:
Separating web information from workplace sources helps provenance, but a revised workbook still needs a dated record-level citation so the generated total can be traced to its current values.
D:
OCR extracts text from scanned images; it does not identify which updated source-record version contributed to a calculated renewal amount. Require linked, dated inputs and flag missing data.
Question 3
At Silverpine Group, a regional chain proposes an enterprise-wide copilot rollout before testing. Evidence: High deployment effort conceals whether one use case can succeed. The business goal is to use a bounded high-value pilot with explicit stop criteria. The sponsor needs an evidence-based diagnosis rather than a guessed fix. Which initial investigation best identifies the controlling constraint before any solution is chosen?
- Score candidate use cases for value, readiness, risk and feasibility
- Inventory integration, permission and approval requirements
- Check index refresh time against the policy effective date
- Inventory supported languages and representative customer language mixes
Explanation
Correct Answer
A:
High deployment effort conceals whether one use case can succeed. Score candidate use cases for value, readiness, risk and feasibility. A bounded, measured pilot tests value, feasibility and risk before resources are committed to enterprise-wide expansion.
Incorrect Answers
B:
Classifying integration and action boundaries is useful for technical scope, but it does not test whether one approved high-value use case delivers enough benefit and acceptable risk to justify expansion.
C:
RAG source freshness affects grounded answers; it is not the primary gate for whether a broad AI program should first run a bounded value/feasibility pilot.
D:
Multilingual coverage must be tested where relevant, but it does not replace prioritizing a pilot with measurable value, readiness, risk, and stop criteria.
Question 4
At Coral Bay Systems, a benefits team wants AI to approve employee exceptions automatically. Evidence: Decisions have legal consequences and policy ambiguity. The business goal is to limit automation to low-risk repeatable tasks and preserve human approvals. The sponsor needs an evidence-based diagnosis rather than a guessed fix. Which initial investigation best identifies the controlling constraint before any solution is chosen?
- Inspect input, output and retrieval tokens per request type
- Review ranking failures with labeled relevance judgments
- Classify requested outputs as generated text or forecast values
- Separate routine information requests from discretionary decisions
Explanation
Correct Answer
D:
Decisions have legal consequences and policy ambiguity. Separate routine information requests from discretionary decisions. Routine policy retrieval can be automated, but discretionary legal or benefits exceptions require an authorized human decision.
Incorrect Answers
A:
Token accounting measures usage costs, not authority to approve an employee benefits exception. Discretionary decisions with legal consequences need a human reviewer.
B:
Ranking improves retrieval relevance; it cannot authorize AI to decide ambiguous policy exceptions with legal effects. Separate routine policy guidance from approvals.
C:
Classifying AI as text generation or prediction does not solve the decision-rights boundary. The core issue is preserving authorized human judgment for discretionary exceptions.
Question 5
At Northlake Services, an account director wants a polite reply to a customer renewal thread. Evidence: A proposed discount is mentioned but not yet approved. The business goal is to keep Copilot email work within confirmed thread facts and human send review. The project owner requires a measurable acceptance criterion before signoff. Which evidence would most convincingly demonstrate that the requirement has been met?
- Chosen model meets quality and cost limits for its intended task
- The draft omits unsupported discounts and remains unsent
- A quality score includes accuracy and safe escalation, not clicks alone
- The restricted user does not receive protected excerpts
Explanation
Correct Answer
B:
A proposed discount is mentioned but not yet approved. The draft omits unsupported discounts and remains unsent. A drafted response must distinguish proposed terms from approved commitments; human review is required before sending a binding offer.
Incorrect Answers
A:
Choosing a suitable model does not make a proposed discount legally or commercially approved. Drafted customer terms must be checked against actual approval before sending.
C:
Usage and adoption analytics describe how a Copilot solution performs; they cannot authorize an unapproved promotion included in an external reply.
D:
Graph access permissions govern which data an app may read or act on, not whether a draft email can promise an unapproved discount. Human approval is still needed.
Question 6
At Cedar Analytics, an organization offers one technical AI course to all employees. Evidence: Nontechnical teams need practical role-specific workflows. The business goal is to tailor AI learning paths to user tasks and literacy. The project owner requires a measurable acceptance criterion before signoff. Which measured outcome proves that employees can perform their actual role-specific work safely, beyond merely recognizing permitted inputs and escalation routes?
- Administrators can detect and investigate cost deviations
- The deployment is blocked until required signoff exists
- Learners demonstrate safe task proficiency rather than attendance alone
- Employees accurately identify allowed inputs and escalation routes
Explanation
Correct Answer
C:
Nontechnical teams need practical role-specific workflows. Learners demonstrate safe task proficiency rather than attendance alone. Task-based proficiency checks demonstrate that employees can apply approved AI workflows safely; attendance alone does not.
Incorrect Answers
A:
Consumption billing controls AI service costs; it does not show whether employees can perform safe, role-specific work. Validate learning with representative tasks and proficiency criteria.
B:
Separate approvers reduce model-release risk, but do not tailor employee training to the actual workflows of HR, support and other nontechnical roles. Use task-based proficiency checks.
D:
Knowing safe-input classifications and whom to contact is valuable security literacy, but it does not demonstrate successful execution of the role-specific business workflows. A practical task-based proficiency assessment is the fuller acceptance criterion for the learning-path objective.
Question 7
At Northlake Services, a training deck promises an agent capability not yet enabled in the tenant. Evidence: Microsoft features can be region or license-dependent. The business goal is to qualify feature claims against actual rollout state. The project owner requires a measurable acceptance criterion before signoff. Which evidence would most convincingly demonstrate that the requirement has been met?
- The result matches approved arithmetic and labels units correctly
- Each conclusion is sourced to the correct evidence stream
- Responses use current clauses and follow the required format
- Training instructions do not promise unavailable controls
Explanation
Correct Answer
D:
Microsoft features can be region or license-dependent. Training instructions do not promise unavailable controls. Feature availability depends on entitlement, tenant settings and rollout location; verify eligibility before promising a capability.
Incorrect Answers
A:
Excel Copilot features may depend on licensing and deployment eligibility; assuming workbook support without checking tenant, license and regional availability does not establish the required feature is usable.
B:
Choosing Researcher instead of Analyst addresses task fit, but cannot override whether the capability is licensed and available in the tenant’s region.
C:
Grounding settings affect how a model uses references. They do not prove that the requested Microsoft feature is offered for this customer’s region and subscription.
Question 8
At Redwood Consulting, a support assistant cites documents but invents unsupported exceptions. Evidence: Citations exist but conclusions are not entailed by the passages. The business goal is to measure groundedness and factual relevance with human spot checks. Only non-sensitive synthetic data and approved sandbox systems may be used. Which bounded trial with synthetic records best tests the proposed approach?
- Try a simulated inventory update with denied and allowed mock roles
- Model simulated daily peak and overnight idle usage
- Test a bounded fictional adversarial input set
- Test mock policies with absent exception clauses
Explanation
Correct Answer
D:
Citations exist but conclusions are not entailed by the passages. Test mock policies with absent exception clauses. A citation is only valid evidence when its source actually supports the claim; attaching a reference cannot justify an unsupported conclusion.
Incorrect Answers
A:
Connectors and actions grant an assistant ways to retrieve or change information; they cannot turn a citation into support for a conclusion the cited passage does not entail.
B:
Dedicated hosting affects capacity and predictable cost. It does not test whether an answer’s claim logically follows from the supplied source excerpts.
C:
Content safety screens prohibited or harmful output patterns. A plausible but unsupported factual conclusion still requires a source-entailment check, not only content moderation.
Question 9
At Elmwood Banking, an AI outreach draft consistently favors one demographic group. Evidence: Training examples omit important customer contexts. The business goal is to evaluate fairness across representative groups before use. The sponsor needs an evidence-based diagnosis rather than a guessed fix. Which initial investigation best identifies the controlling constraint before any solution is chosen?
- Inspect time-series completeness, seasonality and business constraints
- Record baseline cost per resolved case, volume and quality
- Locate the approved forecasting method and uncertainty inputs
- Break down outcomes by relevant cohorts and missing data
Explanation
Correct Answer
D:
Training examples omit important customer contexts. Break down outcomes by relevant cohorts and missing data. Representative test cases are needed to expose subgroup performance gaps and avoid unfair conclusions from narrow training examples.
Incorrect Answers
A:
Demand-forecasting benefits do not fix a training sample that omits customers and edge cases. Evaluation must include the missing customer contexts before generalizing performance.
B:
A return-on-investment comparison measures business outcomes, not whether the model’s examples represent the customer populations it will encounter.
C:
Uncertainty presentation prevents false confidence in a forecast, but cannot compensate for missing training scenarios that produce systematic failures.
Question 10
At Tamarack Logistics, a sales lead wants generated customer contracts sent without inspection. Evidence: Generated text could contain binding promises beyond approval. The business goal is to insert accountable review before external actions. The project owner requires a measurable acceptance criterion before signoff. Which evidence would most convincingly demonstrate that the requirement has been met?
- The external message is sent only after approved corrections
- Answers stay grounded and preserve source version context
- Only accepted decisions are reported as final
- No unauthorized tool call or sensitive disclosure occurs
Explanation
Correct Answer
A:
Generated text could contain binding promises beyond approval. The external message is sent only after approved corrections. Draft generation is not final authorization; a human approval step prevents unsupported contractual commitments from being sent.
Incorrect Answers
B:
A notebook helps maintain background source context; it does not authorize AI-drafted contractual commitments. Have an accountable approver validate external promises.
C:
A meeting recap summarizes approved discussion content. It does not make every AI-generated contractual promise approved for release.
D:
Prompt-injection defenses mitigate malicious instructions in retrieved content, but cannot substitute for commercial or legal approval of the promises in a proposed response.