100% Latest & Updated Microsoft AB-100 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 20
A municipal chatbot must work for keyboard-only users and provide readable error and escalation paths. Which recommendation addresses the identified failure most directly?
- Make the agent decline unsupported legal determinations and route the case to qualified counsel.
- Enforce approved data-region placement and verify every cross-region data flow.
- Specify graceful degradation and a tested fallback path for the failed dependency.
- Define accessibility and user-experience acceptance criteria and test them with assistive workflows.
Explanation
Correct Answer
D:Specify keyboard-only operation, visible focus, accessible error descriptions and usable escalation routes; test them with representative assistive workflows before release.
Incorrect Answers
A:Escalation for unauthorized legal questions protects scope of advice but does not ensure keyboard navigation or readable error feedback in the municipal chatbot.
B:National data-processing location is a compliance decision, not a remedy for inaccessible keyboard controls or unusable error messages.
C:Dependency failover maintains a reduced service during outages, while the stated interface is inaccessible even during normal operation.
Question 21
A housing assistance agent ranks applicants but the policy requires staff to make every eligibility determination. Which option most directly closes the documented gap?
- Define one governed knowledge-and-decision service reused through approved channel adapters.
- Use AI to recommend with reasons while reserving eligibility decisions for authorized staff.
- Encode vendor allowlists, budget caps and transaction scopes in the execution policy.
- Record analyst dispositions and evidence in a reviewable feedback loop.
Explanation
Correct Answer
B:Allow the agent to rank and explain applications as decision support, while requiring an identified authorized staff member to approve every final eligibility outcome.
Incorrect Answers
A:Centralizing a policy engine may make recommendations consistent, but it does not enforce that authorized staff make the final housing eligibility decision.
C:Vendor lists and spending caps constrain automated purchases, not the legal and organizational responsibility for deciding housing eligibility.
D:Review dispositions create an audit trail after an analyst acts, but do not by themselves prevent automatic approval or denial of applicants.
Question 22
An inventory replenishment agent may auto-order standard low-value stock only within approved vendor and spending limits. Which approach best satisfies the actual business constraint?
- Record analyst dispositions and evidence in a reviewable feedback loop.
- Design a bounded agent with tools and explicit approval gates for consequential steps.
- Encode vendor allowlists, budget caps and transaction scopes in the execution policy.
- Make the agent decline unsupported legal determinations and route the case to qualified counsel.
Explanation
Correct Answer
C:Enforce approved vendor IDs, allowed stock categories, per-order and cumulative budgets, and transaction scopes at the ordering API before any automatic purchase.
Incorrect Answers
A:Fraud analyst feedback captures review outcomes, but it cannot bound purchases by preapproved vendor, item category, and spending limits.
B:A general tool agent with approval gates may be useful, but the organization authorizes automatic ordering within fixed limits that must be machine-enforced on each transaction.
D:Escalating out-of-scope legal advice is unrelated to applying inventory procurement limits to low-value replenishment transactions.
Question 23
A fraud triage model flags suspicious cases but needs analysts to capture why a flag was accepted or dismissed. Which specific remediation should the business adopt?
- Use AI to recommend with reasons while reserving eligibility decisions for authorized staff.
- Make the agent decline unsupported legal determinations and route the case to qualified counsel.
- Design an escalation payload containing evidence, action history and unresolved questions.
- Record analyst dispositions and evidence in a reviewable feedback loop.
Explanation
Correct Answer
D:Record the analyst's disposition, evidence links, reason code and timestamp in a governed review loop so later audit and model evaluation can distinguish true from dismissed alerts.
Incorrect Answers
A:Keeping eligibility decisions with staff is an important decision-rights rule, but does not record why this fraud analyst accepted or dismissed a particular alert.
B:Referring legal interpretation to counsel does not capture fraud-alert evidence, disposition reasons, or the analyst responsible for triage.
C:An escalation packet helps transfer unresolved cases, but the requirement is to preserve the accept/reject explanation for every reviewed fraud flag.
Question 24
An employee inquiry agent receives a request for a legal interpretation outside its approved knowledge scope. Which solution approach best meets the stated requirement?
- Make the agent decline unsupported legal determinations and route the case to qualified counsel.
- Encode vendor allowlists, budget caps and transaction scopes in the execution policy.
- Confirm legal availability, licensing and continuity of source access.
- Use AI to recommend with reasons while reserving eligibility decisions for authorized staff.
Explanation
Correct Answer
A:Decline to make binding legal interpretations beyond the approved knowledge scope and hand the request, relevant evidence and uncertainty to qualified counsel.
Incorrect Answers
B:Vendor and budget enforcement constrains inventory purchases, not an employee asking the assistant for legal interpretation it is not authorized to provide.
C:Content licensing verifies whether a source can be used, but a licensed document does not authorize the agent to issue a legal determination.
D:Human responsibility for benefit eligibility concerns an adjudication workflow; this scenario needs a qualified legal escalation rather than an AI legal conclusion.
Question 25
A merger introduces dozens of undocumented document repositories; agents are planned before teams know what data exists. Which design choice targets the observed problem?
- Document timestamp semantics and freshness service-level objectives in the data contract.
- Collect classifications and effective access policies for sensitive attributes.
- Inventory repositories, owners, classifications and access pathways before designing retrieval.
- Enforce approved data-region placement and verify every cross-region data flow.
Explanation
Correct Answer
C:Inventory every repository, accountable owner, access pathway, content class and permitted use before choosing any retrieval connectors for the merged organization.
Incorrect Answers
A:Timestamp freshness rules help once sources are identified, but no one yet knows the inventory of merger repositories or who owns their records.
B:Field classification is an important later control, but cannot substitute for discovering the unknown repositories and documenting their owners and access paths.
D:Residency verification constrains where known data may be processed; it does not discover which source systems exist or whether they can be accessed.
Question 26
The architecture team knows a source has customer records but not who is permitted to use each column. Which concrete solution decision should the team make?
- Normalize entity identifiers and deduplicate conflicting source records before indexing.
- Document timestamp semantics and freshness service-level objectives in the data contract.
- Confirm legal availability, licensing and continuity of source access.
- Collect classifications and effective access policies for sensitive attributes.
Explanation
Correct Answer
D:Classify each sensitive column and document its effective access policy by user or workload before building retrieval paths that could expose regulated customer attributes.
Incorrect Answers
A:Deduplicating customer identifiers reduces conflicting entities but does not determine which sensitive columns each caller may lawfully retrieve.
B:Data freshness definitions distinguish event time and processing time, not the field-level authorization the architect has not documented.
C:Source licenses establish legal availability of a dataset, but do not specify different privileges for a customer address, account number, or regulated field.
Question 27
Two CRM exports have similarly named timestamps, but one is an event time and the other a nightly refresh time. Which design decision should the architect prioritize?
- Document timestamp semantics and freshness service-level objectives in the data contract.
- Inventory repositories, owners, classifications and access pathways before designing retrieval.
- Extract validated delivery events into a governed structured event dataset.
- Confirm legal availability, licensing and continuity of source access.
Explanation
Correct Answer
A:Define event occurrence time, ingestion or refresh time, timezone, and freshness objectives in the CRM data contract so agents do not treat a nightly load as the customer event date.
Incorrect Answers
B:Repository inventories record where the exports live but do not resolve the conflicting meanings of similarly named timestamp fields.
C:Structured delivery-event extraction targets logistics notes, not the different time semantics already present in the two CRM exports.
D:Licensing checks confirm permission to use a source, but cannot tell an agent whether a timestamp means event occurrence or nightly synchronization.
Question 28
A knowledge base is accurate but external suppliers will lose their licenses to access it after a vendor contract expires. Which controlled change is most appropriate for this situation?
- Inventory repositories, owners, classifications and access pathways before designing retrieval.
- Confirm legal availability, licensing and continuity of source access.
- Collect classifications and effective access policies for sensitive attributes.
- Apply caller-scoped access permissions when retrieving protected knowledge documents.
Explanation
Correct Answer
B:Review the licensing terms and vendor dependency, then establish a lawful continuity or replacement-access plan before suppliers lose the right to retrieve the knowledge base.
Incorrect Answers
A:An inventory reveals which knowledge base exists, but the problem is that access rights will expire under a vendor contract, not that the source is unknown.
C:Field-level classification may limit confidential attributes, but cannot extend a supplier's expired license to use the accurate corpus.
D:Per-user access filters enforce current entitlements, yet an expired underlying content license could invalidate use even for an otherwise authorized employee.
Question 29
The team has twenty AI ideas but wants to prioritize those with measurable hours saved and low implementation risk. What is the most defensible intervention for this case?
- Use AI to recommend with reasons while reserving eligibility decisions for authorized staff.
- Agree on a primary business outcome, target metric and accountable owner before building.
- Measure baseline-to-pilot outcomes, adoption, unit costs and failure modes before scaling.
- Score use cases by business value, feasibility, data readiness and risk.
Explanation
Correct Answer
D:Score each candidate against measurable hours saved, feasibility, source readiness and risk, documenting assumptions so high-value practical ideas outrank flashy demos.
Incorrect Answers
A:Human review of eligibility decisions protects final outcomes in public services, but is not the portfolio-level method for ranking twenty unrelated AI ideas.
B:Agreeing on one project's outcome helps later evaluation; it does not compare twenty proposals across value, data readiness, complexity and implementation risk.
C:Pilot measurement is appropriate after a shortlisted use case is tested, but the team first needs a defensible way to select which ideas enter pilots.
Question 30
A pilot handles only 2% of tickets, and sponsors ask whether it is worth expanding to ten departments. Which architecture response most directly addresses the risk?
- Measure baseline-to-pilot outcomes, adoption, unit costs and failure modes before scaling.
- Agree on a primary business outcome, target metric and accountable owner before building.
- Defer the use case until data access and legal feasibility are demonstrated.
- Give the summarization component read-only data and no outbound communication tool.
Explanation
Correct Answer
A:Compare the pilot's actual resolution time, adoption, costs, exception handling and failure rates against baseline before expanding from two percent of tickets to ten departments.
Incorrect Answers
B:A named KPI clarifies what success means, but the existing pilot needs observed outcomes and operating costs to justify broader deployment.
C:Deferring for legal data access applies to a proposal with unusable evidence; here the pilot exists and must be judged using measurable results.
D:Removing outbound messaging from a summarizer is a safety permission choice, not evidence that this ticket pilot delivers enough benefit to scale.
Agentic AI Business Solutions Architect (AB-100) Practice Tests