Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 103

Question 1020

An enterprise team is making a platform design decision. Business context: Third-party documents include sentences telling the model to ignore rules. Present failure: Answers adopt an instruction embedded in a source document. Which approach resolves the mismatch?

  1. Synchronize an indexed source on an explicit freshness schedule
  2. Separate retrieved data from trusted system instructions
  3. Build a bespoke agent when differentiated orchestration is essential
  4. Version reusable prompts with owners and evaluation examples

Explanation

Correct Answer

B:

The defect is that answers adopt an instruction embedded in a source document; the expected business case is that third-party documents include sentences telling the model to ignore rules; the design fits because trust boundaries treat source text as evidence rather than authority.

Incorrect Answers

A:

When operational records update faster than the agent's indexed copy, incremental refresh and source-to-index lag monitoring reduce staleness; in the stated workflow, however, answers adopt an instruction embedded in a source document; the relevant architectural behavior is that trust boundaries treat source text as evidence rather than authority.

C:

A boxed product cannot express the execution and ownership rules; the alternate design can address that situation because custom engineering trades more control for more operating responsibility, whereas this question needs a solution to answers adopt an instruction embedded in a source document.

D:

This design would make sense if many teams copy slightly different instructions for the same task, particularly where a successful prompt change cannot be reproduced after rollout; the requested architecture instead needs the dependency that trust boundaries treat source text as evidence rather than authority because answers adopt an instruction embedded in a source document.

Question 1021

A product owner must measure whether an AI-enabled change actually improves this process: The solution demands its own model selection and hosting policies. The earlier problem was: Host-managed orchestration cannot satisfy a unique execution plan. Which outcome indicator is most relevant?

  1. Successful custom action calls from the prebuilt experience
  2. Completed role-specific workflows within supported host surfaces
  3. Interoperable tasks completed without custom point-to-point parsing
  4. Custom workflow completion against the required execution contract

Explanation

Correct Answer

D:

Because custom-engine architecture controls runtime, model calls and integrations, the outcome worth following is custom workflow completion against the required execution contract; the metric is tied to the reported issue—host-managed orchestration cannot satisfy a unique execution plan—for the scenario where the solution demands its own model selection and hosting policies.

Incorrect Answers

A:

Users need familiar Copilot behavior plus an internal status lookup; its intended measurement is successful custom action calls from the prebuilt experience; substituting that result here would not capture whether host-managed orchestration cannot satisfy a unique execution plan has been corrected.

B:

Completed role-specific workflows within supported host surfaces; that indicator describes a distinct workflow where staff need a role-specific conversation within Copilot and Teams; it does not quantify custom workflow completion against the required execution contract, which is needed for the stated case.

C:

For the workflow where independently hosted agents must exchange tasks and results, interoperable tasks completed without custom point-to-point parsing is an actionable measure; it does not establish whether host-managed orchestration cannot satisfy a unique execution plan in the situation under review.

Question 1022

An agent rollout is intended for this situation: New behavior must be evaluated without exposing the full workforce. The known problem is that a prompt update changes decisions for all users at once. Which success measure connects to that specific outcome?

  1. Successful custom action calls from the prebuilt experience
  2. Error-rate delta between canary and control cohorts
  3. Irreversible changes with verified approver attribution
  4. High-impact actions executed without required approval

Explanation

Correct Answer

B:

Error-rate delta between canary and control cohorts; this outcome measure belongs to the workflow where new behavior must be evaluated without exposing the full workforce; a change in it can help detect whether the defect that a prompt update changes decisions for all users at once is resolved.

Incorrect Answers

A:

That scorecard concerns the activity in which users need familiar Copilot behavior plus an internal status lookup and its result of successful custom action calls from the prebuilt experience; the reported shortcoming here is a prompt update changes decisions for all users at once.

C:

The premise of this indicator is confirmation and human approval separate advice from execution authority; the architecture being evaluated instead serves new behavior must be evaluated without exposing the full workforce, and its success is measured by error-rate delta between canary and control cohorts.

D:

A change in high-impact actions executed without required approval is informative for teams propose that an agent perform both low-risk lookups and financial postings; it does not establish whether the current process achieves error-rate delta between canary and control cohorts.

Question 1023

A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: External knowledge contains embedded role-change commands. Known defect: The agent obeys text retrieved from a supplier document. Which KPI should be selected?

  1. Incorrect tool selections identified before side effects
  2. Duplicate side effects per one thousand actions
  3. Successful indirect prompt-injection attacks
  4. Unapproved external destinations reached through agent tools

Explanation

Correct Answer

C:

For the use case where external knowledge contains embedded role-change commands, the indicator successful indirect prompt-injection attacks is worth baselining against the earlier issue where the agent obeys text retrieved from a supplier document.

Incorrect Answers

A:

A change in incorrect tool selections identified before side effects is informative for the planner calls an order-cancel tool for status requests; it does not establish whether the current process achieves successful indirect prompt-injection attacks.

B:

Although duplicate side effects per one thousand actions is observable, the affected process involves external knowledge contains embedded role-change commands; its acceptance outcome is successful indirect prompt-injection attacks, not the alternative measurement.

D:

The premise of this indicator is execution-time destination and permission checks block exfiltration; the architecture being evaluated instead serves external knowledge contains embedded role-change commands, and its success is measured by successful indirect prompt-injection attacks.

Question 1024

An enterprise team is making a platform design decision. Business context: Sales and Finance roles have different record entitlements. Present failure: A Copilot suggestion includes restricted invoice details. Which approach resolves the mismatch?

  1. Use delegated user identity for user-scoped data actions
  2. Test Sales-to-Service customer context and case handoff
  3. Test role-dependent visibility through Dynamics cross-app flows
  4. Test tool-permission escalation attempts at the execution layer

Explanation

Correct Answer

C:

Because cross-role tests verify that authorization survives integration, this approach corrects the reported breakdown where a Copilot suggestion includes restricted invoice details; the affected situation is that sales and Finance roles have different record entitlements.

Incorrect Answers

A:

Shared admin credentials return restricted cases to ordinary users; the alternate design can address that situation because delegated authorization evaluates each caller's effective permissions, whereas this question needs a solution to a Copilot suggestion includes restricted invoice details.

B:

For a won opportunity should create a properly owned onboarding case, this other approach relies on the fact that end-to-end tests verify identity, mapping and assignment across apps; the issue here is a Copilot suggestion includes restricted invoice details; that mechanism does not establish the required cross-role tests verify that authorization survives integration.

D:

For a user asks the agent to read an unrelated executive record, this other approach relies on the fact that authorization evaluations verify server-side deny behavior; the issue here is a Copilot suggestion includes restricted invoice details; that mechanism does not establish the required cross-role tests verify that authorization survives integration.

Question 1025

Security and business owners want to keep this capability: Support documents arrive in several customer languages. They must also prevent the following: A translated legal clause might lose meaning without access to the original text. Which guardrail should the design incorporate?

  1. Require explicit consent for consequential create or update operations
  2. Cap tool depth and require evidence for consequential conclusions
  3. Store source text and flag low-confidence legal translations
  4. Keep approved alternate processing for endpoint outages

Explanation

Correct Answer

C:

The business setting is support documents arrive in several customer languages, with the exposure that a translated legal clause might lose meaning without access to the original text; the enforceable protection must correspond to the path where translation provides standardized text input while preserving original evidence.

Incorrect Answers

A:

The alternate boundary mitigates a plugin might create business records without user consent or approval, which is relevant to a Teams user should query and update a supported external service; it leaves the current workflow vulnerable to a translated legal clause might lose meaning without access to the original text.

B:

The safeguard for procurement exceptions require comparing several policies and records deals with a different threat: a reasoning agent might continue invoking tools without evidence or budget limits; even after it is enforced, the exposure that a translated legal clause might lose meaning without access to the original text remains possible in the target workflow where translation provides standardized text input while preserving original evidence.

D:

Another process could need this guardrail because an inference outage might leave customers without any approved fallback; the design being assessed supports support documents arrive in several customer languages and has the explicit exposure that a translated legal clause might lose meaning without access to the original text.

Question 1026

A solution architecture is being reviewed for this business situation: Customers ask for current delivery or payment status. The observed defect is that a cached knowledge answer is several hours old. Which architectural decision addresses the underlying requirement?

  1. Check source availability and fallback when grounding is unavailable
  2. Use a supervisor agent that routes to specialized child agents
  3. Use MCP for an agent's access to external tools and data
  4. Call a live action for volatile transactional information
  5. Ground conversational answers in SharePoint knowledge with permissions

Explanation

Correct Answer

D:

The defect is that a cached knowledge answer is several hours old; the expected business case is that customers ask for current delivery or payment status; the design fits because authorized actions query the system of record at request time.

Incorrect Answers

A:

There is a valid place for the mechanism that health checks and explicit no-evidence responses handle missing context safely, namely required knowledge stores are intermittently unreachable; it does not remove the specific defect that a cached knowledge answer is several hours old from this case.

B:

For a request spans procurement, finance and policy expertise, this other approach relies on the fact that a coordinating agent selects bounded specialists and merges their outputs; the issue here is a cached knowledge answer is several hours old; that mechanism does not establish the required authorized actions query the system of record at request time.

C:

Every agent reimplements the same proprietary tool signatures; the alternate design can address that situation because MCP presents supported tools and resources through a common integration contract, whereas this question needs a solution to a cached knowledge answer is several hours old.

E:

Supported SharePoint grounding retrieves organizational content within access controls; its main purpose is employee policies reside in governed Microsoft 365 documents, not correction of the reported defect where a cached knowledge answer is several hours old in the target process.

Question 1027

A product owner must measure whether an AI-enabled change actually improves this process: Model retraining consumes periodically revised business records. The earlier problem was: No one can identify which dataset produced a model. Which outcome indicator is most relevant?

  1. Unauthorized accesses to model-training datasets
  2. Model releases with reproducible data lineage
  3. Time needed to recover a known-good model/data pair
  4. Candidate promotions meeting multimetric acceptance criteria

Explanation

Correct Answer

B:

The observed shortcoming is that no one can identify which dataset produced a model; measuring model releases with reproducible data lineage gives business owners a directly relevant indicator for the process where model retraining consumes periodically revised business records.

Incorrect Answers

A:

A change in unauthorized accesses to model-training datasets is informative for proprietary records are copied to an open test workspace; it does not establish whether the current process achieves model releases with reproducible data lineage.

C:

The proposed metric concerns a preprocessing release changes category labels incorrectly, whereas the failure in this case is that no one can identify which dataset produced a model; the indicator needed here is model releases with reproducible data lineage.

D:

The premise of this indicator is candidate comparison evaluates quality, latency and safety under matched load; the architecture being evaluated instead serves model retraining consumes periodically revised business records, and its success is measured by model releases with reproducible data lineage.

Question 1028

A policy review covers an agent for which multiple low-code apps need a common extraction prompt. Its most relevant exposure is that a shared AI prompt update might spread to live Power Platform apps without change control. What safeguard should be required before release?

  1. Reject generated fields outside the approved schema
  2. Require solution lifecycle controls before production propagation
  3. Enforce source-system entitlements in the integration
  4. Keep authorization at the tool server, not just in agent instructions

Explanation

Correct Answer

B:

In the context where multiple low-code apps need a common extraction prompt, the exposure is that a shared AI prompt update might spread to live Power Platform apps without change control; containing it requires a guardrail in the operation where centralized reusable AI capabilities promote consistency and controlled updates.

Incorrect Answers

A:

Another process could need this guardrail because generated response keys might be trusted even when they violate the schema; the design being assessed supports multiple low-code apps need a common extraction prompt and has the explicit exposure that a shared AI prompt update might spread to live Power Platform apps without change control.

C:

The alternate guard addresses sales enrichment might query external records beyond the seller's entitled scope for sellers need current credit exposure held outside CRM; the present threat, a shared AI prompt update might spread to live Power Platform apps without change control, needs a different enforcement boundary.

D:

This restriction is associated with several agents need the same inventory query operation and addresses the exposure that a shared MCP tool might trust the agent prompt rather than checking actual authorization; it does not specifically protect the target workflow against a shared AI prompt update might spread to live Power Platform apps without change control.

Question 1029

The fault affects an AI-powered workflow in which employee policies reside in governed Microsoft 365 documents. Specifically, answers omit current employee handbook changes. Which diagnostic investigation best distinguishes the relevant failure path?

  1. Audit samples for missing owner, classification and source identifiers
  2. Observe navigation abandonment and contextual data transfer
  3. Compare retrieval results for users with different permissions
  4. Compare orchestration requirements with host extensibility limits

Explanation

Correct Answer

C:

In the setting where employee policies reside in governed Microsoft 365 documents, comparing retrieval results for users with different permissions can isolate the failure in which answers omit current employee handbook changes, because supported SharePoint grounding retrieves organizational content within access controls.

Incorrect Answers

A:

Auditing samples for missing owner, classification and source identifiers explores a separate connection, where metadata enables filtering, review, retention and traceable answers; the diagnosed requirement is employee policies reside in governed Microsoft 365 documents, and the actual defect is answers omit current employee handbook changes.

B:

Observing navigation abandonment and contextual data transfer explores a separate connection, where app-hosted integration brings relevant AI actions into an owned user interface; the diagnosed requirement is employee policies reside in governed Microsoft 365 documents, and the actual defect is answers omit current employee handbook changes.

D:

The proposed probe concerns the solution demands its own model selection and hosting policies, where custom-engine architecture controls runtime, model calls and integrations; this incident instead requires evidence about supported SharePoint grounding retrieves organizational content within access controls and the failure that answers omit current employee handbook changes.

Question 1030

A product owner must measure whether an AI-enabled change actually improves this process: Executives request AI before defining an operational improvement target. The earlier problem was: Prototype activity increases but no business value is measured. Which outcome indicator is most relevant?

  1. Portion of pilots tied to measurable business objectives
  2. High-value concessions with complete approval trace
  3. Improvement in domain error rate versus approved baseline
  4. Median elapsed time from design decision to approval

Explanation

Correct Answer

A:

Portion of pilots tied to measurable business objectives; this outcome measure belongs to the workflow where executives request AI before defining an operational improvement target; a change in it can help detect whether the defect that prototype activity increases but no business value is measured is resolved.

Incorrect Answers

B:

That scorecard concerns the activity in which service, sales and finance jointly decide high-value concessions and its result of high-value concessions with complete approval trace; the reported shortcoming here is prototype activity increases but no business value is measured.

C:

For the situation where a unique domain vocabulary produces persistent measured errors, improvement in domain error rate versus approved baseline is a sensible indicator; the problem that prototype activity increases but no business value is measured requires measurement of portion of pilots tied to measurable business objectives.

D:

The proposed metric concerns business, platform, security and legal teams make conflicting AI decisions, whereas the failure in this case is that prototype activity increases but no business value is measured; the indicator needed here is portion of pilots tied to measurable business objectives.

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