Microsoft AB-100 Practice Test 69

Question 680

A support agent in Dynamics 365 in-app help sees a manager-only salary adjustment instruction when a frontline worker opens the same form. What is the correct correction?

Grant frontline workers temporary manager roles whenever help search fails.
Bind help retrieval to role and feature permissions and test negative access for the frontline user.
Let the worker view the manager instructions but disable the edit button only.
Hide the manager section with a client-side CSS class.

Explanation

Correct Answer

Context-sensitive help must respect the same authorization boundary as the underlying business action.

Incorrect Answers

Privilege elevation is not a valid fallback for a misconfigured knowledge policy.

Restricted procedure details may still expose confidential information and are not the worker’s authorized scope.

Client-side concealment does not remove the sensitive help text from fetched content.

Question 681

A payment assistant reports “refund succeeded,” but Finance has no matching credit. The only log is the final chat message. Which trace design would isolate the missing transaction?

Log full card details in every prompt to reconstruct transactions.
Count the number of refund-related tokens to infer commit status.
Capture correlation ID, validated tool inputs, status code and authoritative refund receipt while masking protected payment fields.
Keep only the assistant’s final prose because it contains the success claim.

Explanation

Correct Answer

An end-to-end trace separates model intent from tool submission and successful downstream commit.

Incorrect Answers

Unnecessary sensitive payment data exposes customers and is not required for action correlation.

Token usage measures inference consumption, not whether Finance posted a credit.

Conversational output cannot establish whether an external financial write occurred.

Question 682

An agent attempts a CRM update and receives HTTP 403. Its dashboard groups the outcome under “model hallucinations.” What should the telemetry capture to diagnose correctly?

Increase generation temperature and retry the CRM call.
Label the request successful because the model found a contact.
Use a global CRM admin token so error metrics remain green.
Classify the tool response as an authorization denial with caller role and denied operation, excluding sensitive content.

Explanation

Correct Answer

403 indicates a server refusal, commonly because the caller lacks authorization, rather than model prose quality.

Incorrect Answers

Text randomness cannot change backend permissions for the same identity.

Record retrieval success does not prove the separate update action was authorized.

Bypassing access control hides a serious authorization defect and creates broader risk.

Question 683

A customer agent calls shipment lookup, order status and a returns service. It says “return accepted” despite a returns API timeout. What observability improvement is necessary?

Associate each tool invocation and result with the conversation correlation ID and distinguish timeout from committed return.
Record only the total conversation duration.
Count the number of successful knowledge searches as evidence that the return was accepted.
Retry the returns submission with a new key whenever a timeout appears.

Explanation

Correct Answer

Per-hop traces show the exact dependency that failed and stop the final answer from masking it.

Incorrect Answers

Latency alone cannot identify which business API failed or what state it left.

Read-only search success does not confirm a returns mutation.

Uncoordinated retry IDs may duplicate the return when the initial request was actually committed.

Question 684

A chat agent’s answer error rate rises after an unannounced model routing change. Logs have timestamps but omit which model deployment served each response. What field is needed?

Switch every query back and forth among models at random.
Record the model deployment/version and routing decision alongside task outcome and trace ID.
Log only the user’s browser version.
Report a single lifetime answer count for the entire tenant.

Explanation

Correct Answer

Without route provenance the team cannot attribute a regression to a specific serving configuration.

Incorrect Answers

Untracked routing variation makes root-cause comparison harder rather than proving a fix.

Client software identity does not reveal which model produced the incorrect answer.

Aggregated volume hides before/after behavior of particular deployment versions.

Question 685

A claims agent retries the same decision tool three times after two network errors, then succeeds once. The dashboard reports three completed claims. What metric should be corrected?

Suppress all retry events from traces to force the total to one.
Generate a new claim ID for each attempt and display them as independent cases.
Use a stable business task ID to count one completed claim while retaining three distinct attempt records.
Count every tool request as a successful claim.

Explanation

Correct Answer

An attempt count diagnoses reliability, but the business outcome denominator must deduplicate retries.

Incorrect Answers

Retry evidence remains necessary to investigate network failures and API cost.

New IDs hide the shared underlying intent and create false volume.

Failed attempts did not produce completed case outcomes and inflate performance.

Question 686

An agent’s p95 latency jumps from two to ten seconds. Operators see only total call duration and suspect the model. What trace spans should be added?

Count prompt words and assume each additional word adds one second.
Remove all tool traces so model calls are easier to compare.
Increase every timeout without locating the slow component.
Instrument retrieval, model inference, action calls and queue waits separately with timestamps.

Explanation

Correct Answer

Segmented spans identify whether delay is caused by search, inference, downstream tools or orchestration.

Incorrect Answers

Token count alone does not establish the actual latency source under concurrent workload.

Absent downstream timing hides potential queue and connector bottlenecks.

Longer limits do not correct the cause and can worsen resource backlog.

Question 687

A payroll copilot trace captures full employee bank details in plaintext as agent tool arguments. Engineers need error evidence but not bank secrets. What is the required change?

Retain structured tool status and correlation while redacting or tokenizing protected argument fields before logs are stored.
Keep the complete arguments and share tracing access with every maker.
Convert banking data into Base64 and treat it as encrypted.
Delete all error timestamps and identifiers instead of protecting sensitive fields.

Explanation

Correct Answer

Diagnostic traceability must be balanced with data minimization and access-controlled log retention.

Incorrect Answers

Broadly readable payroll secrets create a new exposure surface unrelated to troubleshooting.

Base64 is reversible encoding and provides no confidentiality boundary.

Removing operational evidence prevents diagnosis without solving the need for scoped retention.

Question 688

A support workflow starts after an Event Grid-style case update, but identical events sometimes produce two agent sessions. Which trace correlation is required?

Disable event logging so sessions cannot be compared.
Carry the source event ID into agent run, action attempts and downstream committed case state.
Use the generated conversation title as the only event identifier.
Count all agent sessions as distinct incoming complaints.

Explanation

Correct Answer

Event-to-run lineage exposes duplicate delivery and supports idempotent handling of one intended change.

Incorrect Answers

Without source IDs the team loses the ability to diagnose duplicate triggers.

Conversation labels may differ on retries and do not prove two runs share one source event.

Redelivery can inflate customer workload estimates and create duplicate service actions.

Question 689

An agent backlog has 200 complaints about response tone and three verified incidents that exposed restricted contract details. Which triage rule is appropriate?

Classify every comment as equally urgent because it involves AI.
Ignore the privacy incidents until more users complain.
Prioritize confirmed confidentiality incidents by severity and affected scope before tone improvements.
Rank only by complaint volume so tone fixes go first.

Explanation

Correct Answer

Incident impact and risk outweigh raw feedback count when restricted business data was disclosed.

Incorrect Answers

Uniform priority prevents a meaningful risk-based response.

A verified information exposure warrants action regardless of its small occurrence count.

Popular feedback can conceal low-frequency high-impact security failures.

Question 690

Users give a shipment assistant low satisfaction scores during a carrier outage, although it correctly reports unavailable status. What diagnostic distinction matters?

Force the agent to invent tracking updates so customer satisfaction rises.
Call all low ratings model hallucinations without examining dependency status.
Delete negative ratings that coincide with outages.
Separate user sentiment, factual correctness and upstream service availability in feedback classification.

Explanation

Correct Answer

Low satisfaction may reflect unavoidable service unavailability rather than an inaccurate agent answer.

Incorrect Answers

False delivery facts harm reliability even when they temporarily reassure users.

The error could be in a carrier outage or expectation mismatch, not generated facts.

Feedback remains useful for designing clearer fallback and recovery experiences.

Agentic AI Business Solutions Architect (AB-100) Practice Tests

Microsoft AB-100 Practice Test 67 - 10 Questions (Last Reviewed 04-Nov-2026)
Microsoft AB-100 Practice Test 68 - 10 Questions (Last Reviewed 03-Nov-2026)
Microsoft AB-100 Practice Test 69 - 10 Questions (Last Reviewed 02-Nov-2026)
Microsoft AB-100 Practice Test 70 - 10 Questions (Last Reviewed 01-Nov-2026)
Microsoft AB-100 Practice Test 71 - 10 Questions (Last Reviewed 31-Oct-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 680

Explanation

Correct Answer

Incorrect Answers

Question 681

Explanation

Correct Answer

Incorrect Answers

Question 682

Explanation

Correct Answer

Incorrect Answers

Question 683

Explanation

Correct Answer

Incorrect Answers

Question 684

Explanation

Correct Answer

Incorrect Answers

Question 685

Explanation

Correct Answer

Incorrect Answers

Question 686

Explanation

Correct Answer

Incorrect Answers

Question 687

Explanation

Correct Answer

Incorrect Answers

Question 688

Explanation

Correct Answer

Incorrect Answers

Question 689

Explanation

Correct Answer

Incorrect Answers

Question 690

Explanation

Correct Answer

Incorrect Answers

Agentic AI Business Solutions Architect (AB-100) Practice Tests