Free AIF I Practice Exam: AI Fluency for Product Team Members

A 40-question practice run

Allow 60 minutes, answer before opening the explanations, and mark uncertain choices for review. These are original independent PM Mastery questions, not official Scrum.org questions, copied live-exam content, protected course exercises or exam dumps.

The count and time match the published assessment for pacing. Our four-option single-select construction and equal category allocation are editorial choices. The exact official single/multiple-answer proportions are not established, so this set is not an exact reproduction or a validated predictor of the exam result.

Use the goal, evidence, authority and constraints supplied in each situation. A confident model response, another model’s agreement or a disclosure statement does not by itself establish accuracy or responsible use. Conversely, adequate checks can justify proceeding; extra process is not automatically the better answer.

As optional pacing checkpoints, aim to reach questions 10, 20 and 30 near 15, 30 and 45 minutes. These are practice aids, not official milestones.

Questions 1-25

Question 1

Topic: AI Foundations

A product team uses an approved AI tool to draft release summaries for support staff from approved release notes. Its working agreement requires the author to verify the final summary against those notes before sharing it.

Across several release cycles, a teammate routinely:

• Adjusts the audience and scope when drafts are vague, without reminders, producing more relevant summaries.
• Explains which observed shortcoming each prompt change addresses and seeks increasingly task-specific improvements.
• Checks revised text against the release notes. Recently, she removed an unsupported capability claim and rechecked the final summary before sharing.

Which interpretation of this pattern is best supported by the AI Fluency Journey Map?

• A. The pattern supports next-level proficiency: independent prompt revision and final evidence checks demonstrate capability beyond foundational practice.
• B. The pattern supports foundational habits still developing: repeated prompt revisions indicate that effective Direction has not yet become routine.
• C. The pattern supports uncontrolled experimentation: remaining factual errors indicate that the prompt revisions are not producing dependable improvement.
• D. The pattern supports Functioning: purposeful prompt revision and evidence checks are routine, with signals worth discussing for further development.

Best answer: D

Functioning describes foundational habits becoming dependable routine, not AI responses becoming error-free. The teammate responds to specific shortcomings, obtains more relevant drafts, and verifies the finished work against approved evidence. This is purposeful iteration rather than uncontrolled retrying.

The pattern connects the four lenses: Understanding recognizes that fluent output can still be wrong; Direction improves instructions using observed gaps; Evaluation checks the final text against release notes; Responsibility keeps the human author accountable for sharing it.

Natural iteration and dissatisfaction with generic output can be Migrating State signals worth exploring in a development discussion. They do not automatically establish next-level proficiency, confer certification, or justify removing the required evidence checks.

Answer choices explained:

A — Incorrect: Independent iteration and checking demonstrate strong foundational habits, but these observations alone do not establish proficiency at a further level.

B — Incorrect: Revising prompts in response to observed shortcomings demonstrates effective Direction; routine practice does not require using an unchanged prompt.

C — Incorrect: Her revisions address identifiable shortcomings and improve relevance; detecting a remaining factual error shows why evidence checks remain necessary.

D — Correct: Targeted revisions and final verification are dependable habits; increasingly natural, task-specific iteration also supports a conversation about further development.

Question 2

Topic: AI Accountability

A product practitioner uses an AI assistant approved for processing the team’s interview notes. Her weekly summaries help the team choose what to investigate next.

Self-assessment:

• Drafting is faster, but two shared summaries overstated the interview evidence.
• One conclusion cited a passage that did not support it; another omitted feedback challenging a recommendation.

For next week’s summary, she has ten minutes to review its three main conclusions. She wants to test one personal habit and keep a short record of what she checked.

Which review habit best addresses the weakness shown by her self-assessment?

• A. Compare each main conclusion with the interview notes, log missing or contrary evidence, and revise the conclusions to reflect that evidence.
• B. Ask the assistant to rate confidence in each main conclusion, log low ratings, and revise conclusions it identifies as uncertain.
• C. Generate another draft in a fresh chat, log disagreements between the main conclusions, and qualify conclusions where the drafts differ.
• D. Check that each main conclusion has a source reference, log missing references, and add uncertainty labels where references are absent.

Best answer: A

A useful personal adoption plan connects an observed weakness to a behavior that can be practiced and reviewed on the next relevant task. Here, the gap is not drafting speed or tool approval: it is checking whether conclusions are supported and contrary observations remain visible.

Her review should compare the three main conclusions with the interview notes, record gaps or contradictions, and narrow, qualify, or remove claims as needed. She should then check the revised wording. The short review log provides evidence for her next self-assessment, including whether the habit helped her catch overstatements. Completing this practice supports personal development; it is not a certification threshold or a reason to transfer responsibility to the assistant.

Answer choices explained:

A — Correct: Comparing conclusions with source evidence exposes unsupported certainty and contrary observations, while the log makes her practice observable.

B — Incorrect: The assistant’s confidence ratings do not verify support; a confidently overstated conclusion could escape the review she records.

C — Incorrect: Agreement between drafts does not verify the interview evidence, and both drafts can repeat the same unsupported interpretation.

D — Incorrect: A reference can exist without supporting the conclusion, as her earlier summary showed; checking its presence would miss that failure.

Question 3

Topic: AI Prompting & Iteration

A Scrum Team is updating a delivery-address editor. The Developers want an AI assistant to draft a gap report against their unchanged Definition of Done. They will review the report and remain responsible for verifying quality.

Definition of Done:

• Save and postcode-validation acceptance checks pass on the current build.
• Keyboard-only checks pass on the current build.
• Support notes for the change are reviewed by a support representative.

Available evidence:

• Build 18: Save and postcode-validation checks passed using a mouse.
• Build 17: Keyboard-only checks passed. Build 18 changed the editor’s dialog, and no keyboard-check result is recorded for it.
• Build 18 support notes: Drafted; support representative review is pending.

Which prompt best directs the assistant’s comparison?

• A. Compare each check with the most recent result available. Carry forward prior-build passes for checks with no current result; cite supporting records and identify outstanding verification.
• B. Compare each check with current-build evidence. Use passed acceptance checks as confirmation of keyboard usability; cite supporting records and identify any outstanding support-note review.
• C. Compare each check with current-build evidence. Treat every check lacking a passing record as a failed test; cite supporting records and identify the defects needing correction.
• D. Compare each check with current-build evidence. Distinguish confirmed passes, confirmed unmet checks, and checks lacking evidence; cite the records and identify what remains to be verified.

Best answer: D

A Definition of Done provides the quality standard for comparison; the assistant should not adjust it to fit available results. Each check needs applicable evidence, and missing evidence should not be converted into an invented pass or failure.

For build 18, the save and validation checks have passing evidence. Keyboard-only usability remains unverified because the only recorded pass predates the changed dialog. The support-note review is a confirmed outstanding activity because it is explicitly pending. The supplied evidence therefore does not demonstrate that all checks are met.

The assistant can organize these distinctions and identify gaps. The Developers remain responsible for completing the necessary verification and ensuring the work meets the Definition of Done.

Answer choices explained:

A — Incorrect: The keyboard pass from build 17 does not establish that build 18 meets the current-build requirement, particularly after the dialog changed.

B — Incorrect: Mouse-based save and validation checks do not demonstrate keyboard-only usability, which is a separate requirement with no current-build result.

C — Incorrect: A missing passing record establishes an evidence gap, not an observed test failure or a product defect requiring correction.

D — Correct: This distinguishes the pending support review from the missing keyboard result while preserving the supplied quality checks and their evidence requirements.

Question 4

Topic: AI Quality & Confidence

A product team is reviewing an AI-drafted support reply on 6 June 2026. The reply must explain how this customer should transfer workspace ownership. All dates below are in 2026.

Case record:

• 29 May: The customer opens a support ticket about billing.
• 4 June: The customer, who is the current workspace owner, first requests an ownership transfer in that ticket.

Archived guide, retired 1 June:

The current owner emails support. Support verifies the request and initiates the ownership transfer.

Current guide, effective 1 June, last reviewed 5 June:

For ownership-transfer requests received on or after 1 June, the current owner initiates the transfer in workspace settings. Support cannot initiate these transfers.

AI draft, citing the archived guide:

Email support to initiate the ownership transfer. Support will verify your request and start the transfer.

Which assessment of the draft is best supported by this evidence?

• A. The draft needs procedural correction because the transfer request arrived after the current guide took effect, regardless of when the support ticket began.
• B. The draft can be used because the support ticket began before the current guide took effect, so the archived procedure governs this transfer request.
• C. The draft can be used because the current guide was reviewed after the transfer request arrived, so its procedure did not yet apply to that request.
• D. The draft needs only a citation update because the archived guide documents its instructions, so citing the current guide instead would make the reply usable.

Best answer: A

Confidence in a support reply depends on whether its evidence supports the instructions for the specific case. Source currency involves both dates and applicability, not simply whether a citation exists or a page was recently reviewed.

The current guide took effect on 1 June and covers ownership-transfer requests received from that date. This request arrived on 4 June; the earlier billing conversation does not change its applicability. The 5 June review date is not the procedure’s start date.

The reply should therefore direct the current owner to initiate the transfer in workspace settings and cite the current guide. Updating the citation alone would leave the outdated instructions intact.

Answer choices explained:

A — Correct: The transfer was first requested on 4 June, so the procedure effective 1 June requires the owner to initiate it in workspace settings.

B — Incorrect: The current guide applies according to when the ownership transfer was requested, not when the earlier billing ticket began.

C — Incorrect: The 5 June review date does not override the stated 1 June effective date, which precedes the transfer request.

D — Incorrect: Replacing the citation would leave instructions that conflict with the applicable procedure; the substantive guidance also needs correction.

Question 5

Topic: AI Prompting & Iteration

A product team is preparing customer interviews to understand repeated entry in a quote-building tool. Its discovery brief contains all the evidence currently available:

Three support messages mention entering a quote again. We suspect this is slowing quote completion.

The team has five interview places next week and needs participant screening questions. The AI assistant can access only this brief, and its proposed explanations will be treated as assumptions.

Which request would best use AI to expose the missing context for this research step?

• A. Identify unanswered questions about likely workflow failures and their causes, then draft screening questions to find people who match the leading explanation.
• B. Identify unanswered questions about customer priorities and expected improvements, then draft screening questions to find people most interested in a redesign.
• C. Identify unanswered questions about typical customer profiles and product usage, then draft screening questions to reflect the overall customer population.
• D. Identify unanswered questions about which users experience re-entry and when it occurs, then draft screening questions to find people with relevant experiences.

Best answer: D

AI can strengthen a discovery brief by exposing what its evidence does not establish. The messages report re-entry, but do not reveal which users experienced it, what they were doing, or under what circumstances it happened. Slower completion is also an interpretation rather than a demonstrated outcome.

For five exploratory interviews, these gaps should inform screening questions about participants’ roles and actual re-entry experiences. AI can help draft those questions, but suggested user groups, causes and impacts remain hypotheses. People’s responses provide evidence for recruitment and further discovery; the AI’s plausible suggestions do not.

Answer choices explained:

A — Incorrect: Using an assumed cause as a recruitment filter could exclude people whose re-entry occurred under different circumstances.

B — Incorrect: Interest in a redesign does not establish experience of re-entry and shifts recruitment toward a solution before the problem is understood.

C — Incorrect: Reflecting the overall customer population may recruit people who have never experienced re-entry, leaving the reported problem unexplored.

D — Correct: The messages suggest a problem but do not identify affected users or situations, which are needed to recruit participants with relevant firsthand experience.

Question 6

Topic: AI Quality & Confidence

A product team is reviewing an AI-drafted incident brief. The receiving coordinator will see only the brief and must justify initial routing using these rules:

• File Generation investigates invoice files that were not created.
• Download Access investigates download failures for invoice files that exist.

Source notes:

• Six customers report invoice download failures after today’s release.
• A pre-existing test invoice downloads successfully.
• Issuing new links does not restore the affected downloads.
• File-creation checks for all six affected invoices fail: no files were created.

AI draft:

Six customers cannot download invoices following today’s release. A pre-existing test invoice downloads successfully; new links do not restore affected downloads. Suspected owner: Download Access.

An automated format check confirms that all required sections are present. Which revision would provide the decisive evidence needed for initial routing?

• A. Route to File Generation, citing the failed file-creation checks for the affected invoices.
• B. Route to Download Access, citing the successful download test with a pre-existing invoice.
• C. Route to Download Access, citing the failed downloads after issuing new links for affected invoices.
• D. Route to File Generation, citing the start of all reported failures after today’s release.

Best answer: A

Human review must check whether an AI-drafted handoff preserves the observations needed for its intended decision, not merely whether it contains the required sections. Here, the failed file-creation checks establish that the affected invoices have no files. The routing rule therefore points to File Generation.

The reviewer should restore that finding, update the suspected owner, and compare the revised brief with the source notes before sending it. A successful download of a pre-existing file tests a different condition; it does not establish that the affected files were created. Passing a format check does not validate the brief’s evidence or conclusion.

Answer choices explained:

A — Correct: The failed creation checks establish that the affected files do not exist, which is the condition assigned to File Generation.

B — Incorrect: Downloading a pre-existing invoice shows that one existing file is accessible, not that the affected invoice files were created.

C — Incorrect: A replacement link failing does not distinguish an access fault from a missing file; the source notes contain a more decisive check.

D — Incorrect: The release timing identifies when failures began, but it does not establish whether file creation or download access failed.

Question 7

Topic: AI Foundations

A product team member has used AI to turn interview notes into problem statements for four backlog discussions. The drafts read well, and she feels more capable. Her own synthesis skill has not been assessed.

The team has 20 minutes for a learning check using approved interview notes. AI remains available for routine work. A teammate will compare the resulting statement with its source notes using these success criteria:

• Factual claims are traceable to the notes.
• Contradictory feedback is retained.
• Assumptions are labeled.

Which exercise would provide the strongest evidence that her confidence is backed by her own synthesis capability?

• A. Select between two AI-written statements from unfamiliar notes, and explain how the selected statement meets the success criteria.
• B. Write a statement from unfamiliar notes without AI, and explain how it meets the success criteria.
• C. Recreate a previously reviewed statement from familiar notes without AI, and explain how it meets the success criteria.
• D. Revise an AI-written statement from unfamiliar notes, and explain how the revisions meet the success criteria.

Best answer: B

Confidence is a useful self-report, but it does not establish independent capability. Polished AI-assisted drafts show what a person and tool produced together, not what the person can do alone.

A small transfer exercise provides more direct evidence: produce a statement from unfamiliar notes without AI, then have a teammate check it against the sources. Traceable claims, retained contradictory feedback, and labeled assumptions make success observable rather than a matter of polish.

A successful attempt supports a bounded claim about current capability. Without an earlier independent baseline or repeated checks, it does not establish improvement, lasting mastery, or that AI caused learning. AI can still support routine work.

Answer choices explained:

A — Incorrect: Selecting and justifying an existing statement tests recognition and evaluation more directly than the ability to construct a statement independently.

B — Correct: An unaided attempt with unfamiliar notes directly tests independent synthesis, while source-based review checks whether the resulting work meets observable criteria.

C — Incorrect: Recreating a familiar statement may rely on remembering the reviewed result, providing weaker evidence of applying synthesis skills to new feedback.

D — Incorrect: Revision can demonstrate editing and evaluation skills, but the AI supplies the initial synthesis, leaving independent construction less directly tested.

Question 8

Topic: AI Quality & Confidence

A product team uses AI to summarize onboarding interview notes for a Product Owner’s discovery decisions. Each note has a unique ID. The brief must remain three bullets and no more than 120 words. Reviewers will continue checking it against the original notes.

Current prompt excerpt:

Summarize the main customer needs and possible product responses using the supplied interview notes.

Latest draft excerpt:

Customers prefer a guided setup tour.

Source-check findings:

• Two successive briefs omitted links between individual conclusions and note IDs.
• The notes describe difficulty finding instructions but contain no stated preference for a tour.

Which refinement to the team’s prompting habit would most directly address these findings?

• A. Ask for a representative quotation beside each conclusion and a short statement identifying the most common interview themes.
• B. Ask for supporting note IDs beside each conclusion and labels distinguishing reported observations from inferred product responses.
• C. Ask for separate bullets for customer problems and product responses, followed by a shared list of interview-note IDs.
• D. Ask for a confidence rating beside each conclusion and a shared list of the interview-note IDs used.

Best answer: B

Source-check findings should inform the instructions used for similar tasks. Here, the recurring gaps are missing conclusion-level references and an inferred solution presented as a customer preference. Requesting supporting note IDs makes each conclusion easier to inspect within the short format. Distinguishing observations from inferred responses preserves the difference between difficulty finding instructions and a possible guided-tour solution.

These additions support review; they do not guarantee accuracy. Reviewers must still confirm that the cited notes support the associated statements and that interpretations are labeled appropriately. The team can reuse the refined instruction and inspect subsequent briefs to see whether traceability consistently improves.

Answer choices explained:

A — Incorrect: Representative quotations can illustrate themes, but this request does not require note-level attribution or distinguish proposed solutions from reported preferences.

B — Correct: Conclusion-level references make support traceable, while the labels distinguish the recorded difficulty from the inferred guided-tour response.

C — Incorrect: Separating problems from responses clarifies the synthesis, but a shared reference list still leaves individual conclusions disconnected from their supporting notes.

D — Incorrect: Confidence ratings and a shared source list do not show which note supports a specific conclusion or whether it is an interpretation.

Question 9

Topic: AI Foundations

A financial-services product team wants to reduce avoidable support contacts after customers submit a request.

Compliance rule:

The regulatory notice must remain verbatim and fully visible. Separate help and navigation may be added beside it.

Observed evidence:

• Of 100 recent contacts, 70 requested confirmation that the customer’s own submission had been recorded, 20 concerned the notice’s meaning, and 10 concerned withdrawal.
• An existing authenticated receipt page shows each customer’s recorded submissions. Public help describes this page, but the notice screen does not link to it.

Available AI: The approved tool can draft copy and provide help chat using public guidance and aggregate contact counts. It cannot query customer records, and policy prohibits entering customer records into it.

Which AI-suggested addition is best supported for an experiment targeting the largest source of contacts?

• A. Add a plain-language explanation of the notice, drafted from the approved public guidance.
• B. Offer an AI help chat that answers individual receipt questions from the approved public guidance.
• C. Link to the authenticated receipt page so customers can view their recorded submission status.
• D. Add a withdrawal checklist for submitted requests, drafted from the approved public guidance.

Best answer: C

An AI-generated alternative is useful when it fits the observed problem, available evidence, and operating constraints. Here, most contacts seek confirmation of an individual submission, not a general explanation of the process. Linking to the existing authenticated receipt page gives customers access to the relevant recorded information without changing the regulatory notice or exposing customer records to AI.

The AI tool can help with general guidance, but public information cannot establish a customer’s live receipt status. The proposed support reduction remains a hypothesis: test the link and observe whether receipt-related contacts decrease and customers can find their submission records.

Answer choices explained:

A — Incorrect: Separate guidance is permitted, but explaining the notice targets the 20 meaning-related contacts rather than the 70 requests for individual receipt confirmation.

B — Incorrect: Public guidance can describe submission procedures, but the tool lacks the customer records needed to establish whether an individual request was recorded.

C — Correct: The link addresses the dominant contact reason using existing customer-specific records while preserving the notice and keeping those records outside the AI tool.

D — Incorrect: A withdrawal checklist could address the 10 withdrawal-related contacts, but it does not address the largest observed source of support demand.

Question 10

Topic: AI Prompting & Iteration

A Product Owner is preparing an internal Support briefing and gives an AI assistant this request:

Draft a two-sentence rollout update from the release note. Preserve the activation condition and distinguish planned timing from commitments.

Release note:

Enable pilot access only after Support approves the troubleshooting guide. Friday is a target; activation moves if approval is delayed.

AI draft:

Pilot access will start on Friday. Support will approve the troubleshooting guide during the pilot.

Which feedback best corrects the draft before the Product Owner reviews it for the briefing?

• A. Revise the update so Support approval remains a task during the pilot, while Friday becomes a tentative rather than committed date.
• B. Revise the update so Support approval precedes pilot access, while Friday remains a tentative rather than committed activation date.
• C. Revise the update so Support approval precedes pilot access, while Friday becomes the deadline for obtaining approval and activating access.
• D. Revise the update so Support approval gates wider release rather than pilot access, while Friday remains a tentative pilot date.

Best answer: B

Corrective feedback should identify the observed mismatch and request a revision grounded in the supplied evidence. Here, Support approval is a prerequisite for pilot access, and Friday is a target that may move.

The draft changes both meanings: it promises Friday and places approval after activation. Focused feedback should restore the approval gate and tentative timing while retaining the requested short Support update. There is no need to replace the task with a longer, more general request or ask the assistant to sound more certain.

The Product Owner should check the revised draft against the release note before using it. Clearer direction supports correction, but does not guarantee an accurate result.

Answer choices explained:

A — Incorrect: Making Friday tentative fixes the timing claim, but leaving approval during the pilot still contradicts the required approval before access.

B — Correct: This restores the approval-before-access condition and preserves Friday as planning information rather than a promised activation date.

C — Incorrect: A target that moves when approval is delayed does not establish a deadline for approval or activation.

D — Incorrect: Moving the approval gate to wider release incorrectly permits pilot access before the approval explicitly required by the note.

Question 11

Topic: AI Foundations

A product team is preparing its first customer interviews for a proposed shift-swapping feature. It has collected no customer evidence. The team gives an AI assistant only a feature sketch and its own assumptions about coordination difficulties.

AI output:

Recurring interview themes: unclear approval status and concern about last-minute staffing gaps.

The team must finish an interview guide today; the interviews begin next week. Which use of these themes is supported?

• A. Treat the themes as independent support for team assumptions and use interviews to explore potential solutions.
• B. Treat the themes as indirect customer feedback and use interviews to identify which problem has the highest priority.
• C. Treat the themes as hypotheses for investigation and use interviews to learn whether customers experience these problems.
• D. Treat the themes as preliminary customer findings and use interviews to assess how widespread each problem is.

Best answer: C

AI can help turn a feature sketch and team assumptions into useful interview topics. It cannot establish that customers experience those problems simply by producing a plausible summary. Here, no customer observations support the themes, and the word “recurring” does not make them research findings.

The team can use the output to prepare neutral questions about how managers handle shift swaps, while remaining open to different or unexpected problems. The discovery brief should label these themes as hypotheses. Actual customer observations should be recorded separately and used to support, revise or reject those hypotheses.

Answer choices explained:

A — Incorrect: Reworking the team’s assumptions into fluent themes does not independently corroborate those assumptions or establish which problems need solutions.

B — Incorrect: No customer feedback was supplied, so the assistant’s themes cannot stand in for customer input about problems or priorities.

C — Correct: The supplied material supports drafting possible topics, while whether those problems occur for customers remains to be investigated.

D — Incorrect: No customer observations have been collected, so the themes cannot yet be called findings, even if interviews later investigate prevalence.

Question 12

Topic: AI Accountability

Noor is reviewing her personal AI adoption plan before coaching becomes less frequent. She wants a habit to become her normal way of working: give AI task-specific context, check the final output against evidence, and own the decision to use it.

Review records:

• Coaching notes document all three behaviors during repeated weekly sessions, after reminders.
• A practice log documents all three behaviors across several exercises completed independently in one afternoon.
• Saved discovery summaries, delivery notes, and support drafts document all three behaviors across several weeks, without reminders.
• Weekly self-assessments consistently report using all three behaviors in everyday work.

Which record most strongly supports treating Noor’s planned habit as routine?

• A. The weekly self-assessments reporting consistent use during everyday work.
• B. The coaching notes documenting the habit during repeated guided sessions.
• C. The saved work documenting the habit across several weeks without reminders.
• D. The practice log documenting the habit across several independent exercises.

Best answer: C

A routine habit is supported by repeated, observable use in ordinary work without reminders. Noor’s saved work spans several weeks and different product tasks. It demonstrates actual context-setting, evaluation of final outputs, and human decisions about using those outputs, rather than intention alone.

This pattern supports treating the habit as routine within the sampled work and choosing a personal improvement target that builds on it. There is no universal number of tasks or weeks that creates a formal pass threshold. Routine practice also does not guarantee that future AI outputs will be accurate or remove the need for checking and human accountability.

Answer choices explained:

A — Incorrect: Consistent self-reports support reflection, but provide less direct evidence of actual behavior than saved work documenting the checks and decisions.

B — Incorrect: Repeated success after coaching reminders demonstrates supported performance, not that Noor initiates the habit independently during everyday work.

C — Correct: Repeated, unprompted work samples directly show the planned behaviors occurring over time in ordinary product tasks.

D — Incorrect: Several successful exercises in one afternoon demonstrate capability, but do not show that the behavior persists across separate work occasions.

Question 13

Topic: AI Foundations

A Scrum Master is discussing further AI development with Amira, who prepares customer-facing incident updates. Updates must accurately reflect the current incident status. Her approved assistant can draft from supplied text but cannot access the live incident tracker. The tracker contains the response team’s verified status, and authors must review the final update before publication.

Observed practice:

• Amira uses the assistant most teammates prefer and revises requests when drafts are too technical for customers.
• For an incident update, she obtains current status from the tracker and supplies it to the assistant for drafting. She explains why each tool fits its part of the task.
• She checks the revised update against the verified status.

Which observation most strongly demonstrates independent tool judgment as a readiness signal for further development?

• A. She uses the tracker to establish current status and the assistant to draft the customer-facing wording.
• B. She revises her requests when drafts are too technical for the customers who will read the update.
• C. She uses the assistant that most teammates prefer for preparing similar customer-facing incident updates.
• D. She checks the revised customer update against the verified incident status before it is published.

Best answer: A

Independent tool judgment means choosing AI use according to the task, available information and ability to evaluate the result. The assistant can help communicate a verified incident status, but cannot independently establish live status without access to the tracker. Amira explains this distinction and assigns the work accordingly. Choosing the team’s popular assistant does not undermine independent judgment when the choice has a sound basis.

In the AI Fluency Journey Map, dependable prompting and checking support Functioning. Reasoned decisions about when an assistant fits, and when another source is needed, can support a discussion about further development. These are readiness signals, not automatic certification or permission to relax controls. Understanding capabilities, directing the work, evaluating the result and retaining responsibility remain connected habits.

Answer choices explained:

A — Correct: Her justified division of the work matches each tool’s capabilities to the task, demonstrating judgment beyond following a familiar default.

B — Incorrect: Audience-focused iteration demonstrates effective Direction, but does not directly show how she decides when the assistant is suitable.

C — Incorrect: Using a popular assistant may be appropriate, but popularity itself does not demonstrate independent assessment of its suitability.

D — Incorrect: Final review demonstrates Evaluation and Responsibility, but provides less direct evidence of deciding which tool suits each task.

Question 14

Topic: AI Quality & Confidence

A product team is drafting an AI-assisted update after fixing a defect that delayed account credits. Support needs the update in ten minutes to decide which of 40 affected account cases can be closed. The assistant has this summary but no billing-system access:

• The processing log says the credit job finished, but contains no account-level posting results.
• A spot-check confirms the expected credits for four accounts.
• A ready-to-run reconciliation report can compare expected and posted credits for all 40 accounts in three minutes. It has not been run.

The draft says that all 40 expected credits appear to be posted. A rushed teammate asks:

Skip the report and change ‘appear to be posted’ to ‘are posted’.

Which response best supports a reliable update within the deadline?

• A. Check four more accounts and use the stronger statement if their expected credits are posted.
• B. Request a fresh AI review and use the stronger statement if it reports high confidence.
• C. Review the batch log and use the stronger statement if no errors were reported.
• D. Run the reconciliation and use the stronger statement if all 40 expected credits match.

Best answer: D

Confident wording cannot replace a completed check. The relevant claim is that every affected account has its expected credit, not merely that processing finished or a sample succeeded. The reconciliation tests that claim directly and takes three minutes, making it proportionate to the ten-minute deadline.

If the report confirms all expected credits, the team can state that result plainly. If exceptions remain, the update should distinguish confirmed postings from unresolved cases. The team should check the final AI-assisted wording against those findings before Support relies on it to close cases. Stronger language alone adds neither evidence nor reliability and can conceal the missed verification.

Answer choices explained:

A — Incorrect: Checking another small sample does not verify the whole set, so it cannot support a statement about every affected account.

B — Incorrect: The assistant has no billing-system access, so greater reported confidence cannot establish posting results missing from its supplied evidence.

C — Incorrect: A finished job with no reported errors does not establish that each affected account received its expected credit.

D — Correct: The reconciliation checks actual postings for every affected account and fits within the deadline, providing evidence for the stronger statement.

Question 15

Topic: AI Accountability

A customer-insights team uses an approved AI tool to synthesize pilot interviews for a Scrum Team considering whether to release a new checkout flow to all customers.

Handoff evidence:

• The insights team checks the summary against the interviews and supplies source links.
• All interviewees are returning customers; the report identifies this limitation.
• The report recommends a full rollout.

The receiving team’s Product Owner has rollout authority under company policy. The insights lead writes:

“We have verified the report and approved its recommendation. You can carry our approval into your release decision.”

Which interpretation best maintains accountability across this handoff?

• A. The receiving Product Owner assumes responsibility for both the original analysis and the rollout decision; the insights team’s responsibility ends when the report is accepted.
• B. The insights team and receiving Product Owner share final accountability for the rollout decision; the Product Owner formally authorizes the decision they jointly approve.
• C. The insights team remains accountable for the rollout decision; the receiving Product Owner is responsible for reviewing the report and issuing formal authorization.
• D. The insights team remains responsible for its analysis; the receiving Product Owner assesses its applicability and remains accountable for the rollout decision.

Best answer: D

A handoff transfers evidence, not the receiving Product Owner’s accountability for how it is used. The insights team remains responsible for the quality and limitations of its analysis. The Product Owner remains accountable for the rollout’s value judgment under the stated release policy.

Verification establishes that the report reflects the interviews; it does not establish that returning customers’ experiences also describe first-time customers. The Product Owner must consider that limitation when evaluating the recommendation. Upstream approval can inform this judgment, but neither AI assistance nor another team’s sign-off takes ownership of the receiving team’s decision.

Answer choices explained:

A — Incorrect: Accepting a report creates responsibility for its use, but does not erase the supplying team’s responsibility for its analysis.

B — Incorrect: Collaboration can inform the rollout decision, but it does not replace the receiving Product Owner’s accountability with shared final ownership.

C — Incorrect: Approving an analysis does not make its authors accountable for the receiving Product Owner’s rollout decision.

D — Correct: The insights team owns the evidence it supplies, while the Product Owner must judge whether returning-customer findings support a rollout to all customers.

Question 16

Topic: AI Foundations

A Product Owner is preparing a 10:00 stakeholder update about a checkout incident. The assistant can use only text in the current conversation; it has no connection to the incident system or other live lookup capability.

The only incident information supplied is an incident-system note timestamped 08:15:

Checkout failures remain elevated. Investigation is continuing. Next update planned for 09:00.

At 10:00, a teammate asks for the incident’s current status. The assistant replies:

The incident was resolved at 09:20 after a rollback. Checkout is operating normally.

Which conclusion is supported by the information available to the assistant?

• A. The incident has probably recovered because the planned 09:00 update supports the assistant’s later report.
• B. The incident has recovered because the assistant supplies a more recent status than the pasted note.
• C. The incident remains active because the latest supplied incident note reports elevated checkout failures.
• D. The current status is not established because the assistant’s incident information is limited to the 08:15 note.

Best answer: D

An assistant can summarize information supplied in its conversation, but asking for current information does not give it access to an external system. Here, the available evidence establishes what was reported at 08:15, not the incident’s status at 10:00. Scheduling an update for 09:00 does not supply that update’s contents.

The claimed rollback and 09:20 resolution are unsupported details. They might happen to be true, but their specificity is not evidence of access or accuracy. For the stakeholder update, obtain a current incident record or confirmation from the incident owner, then use the assistant to summarize that supplied information.

Answer choices explained:

A — Incorrect: A planned update time does not indicate whether recovery occurred or substantiate the reported rollback and resolution.

B — Incorrect: A newer assistant response is not a new incident observation, so its timing does not validate the claimed recovery.

C — Incorrect: The note establishes an earlier reported state; it does not prove that the incident remains active at 10:00.

D — Correct: The only supplied incident evidence describes 08:15, and the assistant cannot retrieve subsequent updates.

Question 17

Topic: AI Prompting & Iteration

A Product Owner is preparing a 20-minute discovery conversation with Support and Sales to understand their different priorities for improving customer trials.

Source notes:

• Support favors clearer import-status information after six customers needed help determining whether their imports had finished.
• Sales favors a sample-data experience after two prospects asked to explore the product before importing their own data.

The Product Owner asks AI to role-play both stakeholders and prepare a meeting brief. The brief concludes:

Both stakeholders favor a guided setup wizard. Focus the meeting on agreeing its scope.

Neither stakeholder has reviewed the brief or discussed the wizard with the Product Owner.

How should the Product Owner use this brief to prepare for the conversation?

• A. Use the role-play as additional stakeholder evidence and prepare questions about balancing its findings with the original source notes.
• B. Use the wizard as a shared direction and prepare questions about the smallest version that could address both stakeholders’ concerns.
• C. Use the ticket and prospect counts to establish relative demand and prepare questions about prioritizing the more frequently reported need.
• D. Use the wizard as a hypothesis and prepare questions about the experiences and trade-offs behind each stakeholder’s preferred improvement.

Best answer: D

AI can help prepare a stakeholder conversation by organizing observations, proposing possibilities and drafting questions. It cannot establish agreement by simulating the people involved.

Here, the source notes describe different concerns: uncertainty during imports and difficulty exploring the product before importing data. A guided wizard might help, but that possibility is not evidence that either stakeholder prefers it.

The Product Owner can use the brief to ask what happened in each situation, what consequences mattered and what remains uncertain. The conversation may reveal competing needs or a useful connection between them. AI supports that preparation; hearing the perspectives and making accountable product judgments remain human work.

Answer choices explained:

A — Incorrect: Simulated stakeholder perspectives can suggest questions, but they are not additional observations of stakeholder preferences.

B — Incorrect: A solution that could address both concerns does not establish stakeholder agreement; the claimed shared direction comes from the AI role-play.

C — Incorrect: Counts from support tickets and prospect conversations are not comparable measures of demand and do not establish which improvement deserves priority.

D — Correct: The notes support exploring two different needs, while the wizard remains an AI-proposed possibility that stakeholders have not endorsed.

Question 18

Topic: AI Prompting & Iteration

A product-team member is helping a support agent draft a customer reply using an approved AI tool. Organizational policy permits non-identifying summaries, limited to information needed for the task. The agent will review the draft and add personal and account details locally.

Support record:

• Current issue: duplicate subscription charge.
• Original request: refund all subscription payments made this year.
• Agreed remedy: refund only the duplicate payment to the original payment method within five business days.
• The full record also contains earlier, unrelated support conversations.

Drafting prompt:

Write a brief, empathetic reply explaining the billing issue and the agreed resolution.

Which context should the practitioner provide to obtain a usable draft while following the policy?

• A. Provide a non-identifying summary of the full support history, including the current issue and the agreed refund details.
• B. Provide the current issue category and the customer’s requested refund, including the agreed payment method and timing.
• C. Provide the current issue category and the agreed refund, including its scope, payment method, and timing.
• D. Provide the current issue category and the general refund policy, leaving the agreed refund details for the agent to add.

Best answer: C

Data minimization preserves task-relevant meaning while omitting unnecessary customer information. Here, the tool needs the duplicate-charge issue and the agreed refund’s scope, payment method, and timing. These facts support a specific reply without names, account identifiers, payment references, or unrelated support history.

Minimizing context does not mean removing useful details until only a vague billing concern remains. The distinction between the customer’s original request and the agreed remedy must survive summarization. Clear direction still cannot guarantee accurate output, so the agent should check that the final draft preserves the agreement before sending it.

Answer choices explained:

A — Incorrect: Earlier, unrelated support history adds information the reply does not need, even when the summary contains no direct identifiers.

B — Incorrect: The original request is not the agreed remedy; using its broader scope would produce an unsupported refund commitment.

C — Correct: The issue and agreed refund details provide the meaning needed for the reply without exposing the customer’s wider record.

D — Incorrect: A general policy does not supply the case’s agreed refund scope, method, or timing, so the draft would lack the required resolution.

Question 19

Topic: AI Prompting & Iteration

A Product Owner asks an AI tool:

Summarize six usability sessions in 150 words or fewer for an internal discovery discussion. Separate observations from interpretations and preserve conflicting feedback.

The first draft says participants unanimously wanted automatic reminders. The notes show that four favored reminders and two objected.

After revision:

• The Product Owner checks the final draft against the notes, confirming that it accurately represents both views and labels interpretations.
• The draft contains 138 words.
• Two teammates find it clear enough to guide follow-up research questions, although they suggest different wording for the closing sentence.

The discussion starts in ten minutes. The summary will support research planning, not authorize development.

Which stopping criterion is most appropriate for this draft?

• A. Stop when the reviewers agree on their preferred wording for the draft’s closing sentence.
• B. Stop when the reviewed draft meets the discussion’s purpose and the stated content checks.
• C. Stop when the AI reports that no further wording improvements remain in the draft.
• D. Stop when repeated AI generations produce matching findings and interpretations of the session notes.

Best answer: B

A useful stopping criterion is whether the reviewed draft meets its intended purpose and relevant checks, not whether further wording changes are possible. The observed weakness was an unsupported claim of unanimity. The Product Owner corrected it and checked the final version against the session notes, including contrary feedback. The draft also meets the length limit and is clear enough to support research planning.

Further revision would be warranted if review revealed a material omission, misleading interpretation, or unclear meaning. Different preferences for an already clear closing sentence do not establish such a gap. Repeated generation and AI assurances cannot replace this evidence-based judgment.

Answer choices explained:

A — Incorrect: Both reviewers already find the draft clear enough for research planning, so agreement on preferred wording is not a necessary readiness criterion.

B — Correct: The revised draft has passed the relevant checks and is clear enough for its intended use; remaining wording preferences do not demonstrate a gap.

C — Incorrect: The AI’s judgment about remaining improvements does not establish fitness for use and may prolong polishing after the relevant checks have passed.

D — Incorrect: Agreement across generations does not independently verify the findings; checking the final draft against the session notes provides stronger evidence.

Question 20

Topic: AI Prompting & Iteration

A product team is reusing an AI prompt to draft a reply to a customer:

“The export failed twice today. We missed the courier cutoff and spent an hour calling customers. We want a credit, not another generic apology.”

Current approved case notes:

• The cause is under investigation. Support needs the export name and exact error text.
• Support may forward credit requests to billing but cannot approve credits. Billing has made no decision about this case.

The customer’s message and current notes are included in the reusable prompt. A support practitioner will review the final draft before sending it.

Which added instruction best directs a neutral, useful reply within support’s authority?

• A. Acknowledge the operational impact, ask for the export name and error text, and assure a credit if the investigation confirms a product fault.
• B. Acknowledge the operational impact, ask for the export name and error text, and explain that credit review must wait until the investigation ends.
• C. Acknowledge the operational impact, ask for the export name and error text, and offer to forward the credit request to billing for review.
• D. Acknowledge the operational impact, ask for the export name and error text, and state that a credit is likely after billing review.

Best answer: C

A neutral support reply can acknowledge concrete disruption without assigning an unconfirmed cause or promising compensation. Here, recognizing the missed cutoff and extra customer calls is more useful than a generic apology. Asking for the export name and error text also supports the investigation.

The approved information permits forwarding a credit request, not predicting or guaranteeing its outcome. It also provides no basis for delaying billing review until the investigation ends. Prompt instructions should preserve those boundaries rather than invent assurances or restrictions. Reused prompts need current case context, and the human reviewer remains responsible for checking the final wording and any customer-facing commitments.

Answer choices explained:

A — Incorrect: A condition does not give support authority to promise compensation, and no approved policy guarantees a credit for confirmed product faults.

B — Incorrect: The approved notes contain no requirement to finish the investigation before billing reviews a request; adding one would invent policy.

C — Correct: Forwarding the request addresses the customer’s desired remedy while leaving the compensation decision with the authorized billing team.

D — Incorrect: No billing decision supports predicting a credit; acknowledging the customer’s disruption does not justify creating an expectation of compensation.

Question 21

Topic: AI Prompting & Iteration

A support specialist reuses an AI prompt to draft a reply for a customer learning English. The customer has read the current help article and says:

Please use simple English. I do not understand “refrain from reinitiating.”

Current approved guidance:

• Open Uploads and select Retry once.
• Refrain from reinitiating while the status indicates Processing, as doing so may result in duplicate uploads.
• If the status still shows Processing after 10 minutes, contact support.

The AI draft copies the guidance word for word. The specialist will review the complete revised message before sending it.

Which revision request best meets the customer’s language needs while preserving the approved procedure?

• A. Use short, literal English sentences for the greeting and closing. Keep the procedure and warning in their approved wording, including “Refrain from reinitiating while the status indicates Processing.”
• B. Use short, literal English sentences. Reword the warning as: “Do not select Retry again while the status is Processing. Selecting Retry again may upload the same file twice.”
• C. Use short, literal English sentences. Reword the warning as: “Wait ten minutes before selecting Retry again. Selecting Retry again may upload the same file twice.”
• D. Use short, literal English sentences. Reword the warning as: “Do not select Retry again while the status is Processing. Selecting Retry again will upload the same file twice.”

Best answer: B

Translation-aware support writing simplifies language, not the underlying procedure. Short, literal sentences help language learners and translation tools interpret instructions. Interface labels should remain recognizable, and rewritten warnings must preserve their conditions and level of certainty.

Here, the customer must not retry while Processing is shown. Another retry may cause duplicate uploads, and ten minutes triggers contacting support, not retrying. These meanings must survive simplification.

A reused prompt needs the current approved information and the recipient’s language needs. Previous success does not guarantee an accurate new draft. The specialist must check the complete revised message and remains responsible for customer-facing instructions and commitments.

Answer choices explained:

A — Incorrect: Copying the approved wording leaves the customer’s stated difficulty unresolved; preserving the procedure does not require preserving its complex wording.

B — Correct: The simpler wording preserves both the Processing condition and the possibility, rather than certainty, of duplicate uploads.

C — Incorrect: Ten minutes triggers contacting support if Processing continues, not permission to select Retry again.

D — Incorrect: The guide describes duplicate uploads as a possible consequence; changing “may” to “will” introduces unsupported certainty.

Question 22

Topic: AI Quality & Confidence

At 13:00, a product team is preparing a renewal response due at 16:00. The customer has requested a binding 30-minute recovery guarantee. Missing the deadline moves the renewal to next month’s purchasing round.

An AI assistant cites three successful recovery drills and proposes:

We guarantee full recovery of all customer data within 30 minutes of any outage, or credit one month’s fees.

Review notes:

• Approved terms specify a four-hour recovery target, not a guarantee.
• The drills took 18–24 minutes using small synthetic datasets. Production-scale recovery has not been tested; the next such drill is tomorrow.
• The team can interpret the drill results but lacks contract-risk expertise and authority to approve new guarantees.
• Urgent contract exceptions go to the commercial contracts lead, who can involve reliability specialists and provide a decision within one hour today.

Which response best preserves the renewal opportunity while respecting the team’s authority?

• A. Use the urgent contract-exception route now, supplying the proposed guarantee, approved terms, and the scope and results of the recovery drills.
• B. Decline the requested 30-minute guarantee in today’s response and proceed with the approved four-hour recovery target for this renewal.
• C. Limit the 30-minute guarantee to the data volumes covered by the recovery drills and include that narrower commitment in today’s renewal response.
• D. Defer the contract decision until tomorrow’s production-scale recovery drill, then submit the proposed guarantee and new results through the urgent contract-exception route.

Best answer: A

Escalation is appropriate when a consequential commitment exceeds the team’s expertise and authority. Successful small-data drills support a limited observation, not a guarantee covering all customer data and every outage. Fluent AI wording does not close that evidence gap.

The urgent contract-exception route can provide an authorized decision before the deadline. The team should supply the proposed clause, existing terms, and drill results with their limitations. The commercial contracts lead can then assess the exposure and obtain specialist advice as needed. Escalation does not require waiting for perfect evidence or assuming approval will follow. It preserves the opportunity for an informed decision without letting the team or AI authorize the promise.

Answer choices explained:

A — Correct: The commercial contracts lead can decide before the deadline, using the current evidence and its limitations to assess the proposed commitment.

B — Incorrect: Declining stays within the team’s authority, but closes the opportunity before using an available authorized review that can finish before the deadline.

C — Incorrect: Narrowing the data volume does not give the team authority to make a new guarantee or establish recovery performance across outages.

D — Incorrect: Waiting for tomorrow’s drill misses today’s purchasing deadline; the available authorized review can consider the limitations of current evidence now.

Question 23

Topic: AI Quality & Confidence

A product team is deciding where to expand its use of an approved AI assistant without increasing time spent reviewing and correcting outputs.

Trial observations:

• Release formatting: Across four releases, the assistant reformatted approved change records accurately. Source checks confirmed that facts were preserved; reviewers made only minor presentation edits.
• Discovery interpretation: Across four interview syntheses, the assistant presented mixed customer preferences as agreement. Source checks found omitted objections, and reviewers substantially rewrote the findings.

Both tasks used complete, permitted inputs. All final outputs were accurate after human review.

Which assessment should guide the team’s decision?

• A. Confidence should fall equally across both tasks because the discovery omissions invalidate the earlier successful release-formatting checks.
• B. Confidence should remain task-specific because reliable release formatting did not extend to faithful interpretation of conflicting interview evidence.
• C. Confidence in discovery should match formatting after a prompt revision because explicitly requesting dissent addresses the observed omission.
• D. Confidence can extend across both tasks because human review brought every final output to the required standard.

Best answer: B

AI output quality must be evaluated for the task being performed. Accurate formatting demonstrates useful performance when transforming approved material, not dependable interpretation of customer evidence.

Here, discovery summaries erased meaningful disagreement and required substantial human correction. Accurate final documents show that reviewers repaired those failures, not that the assistant performed both tasks equally well. With review time unchanged, expansion should reflect that difference.

The team can retain the supported formatting use while improving discovery instructions and testing revised outputs against the complete interview notes. Confidence should increase when observable fidelity and correction effort improve, not merely when a prompt changes or another task succeeds.

Answer choices explained:

A — Incorrect: Discovery failures do not invalidate the direct source checks confirming release-formatting accuracy; the evidence supports different judgments for the two tasks.

B — Correct: Source checks support confidence in formatting, while omitted objections show that discovery interpretation has not demonstrated comparable fidelity or correction effort.

C — Incorrect: Requesting dissent is a useful improvement to test, but changing instructions does not establish that revised discovery outputs will faithfully preserve conflicting evidence.

D — Incorrect: Final accuracy includes human corrections; substantial discovery rewriting does not demonstrate that expanded use would fit the unchanged review budget.

Question 24

Topic: AI Prompting & Iteration

A product team asks an AI assistant to draft acceptance criteria for saving a draft form.

Agreed behavior:

• The user selects Save draft. Missing required fields do not prevent saving.
• A successful save displays confirmation. Reopening the draft restores the last successfully saved values.
• A failed save displays an error and retains the current entries for retry.

AI draft:

Draft saving is reliable and easy to use. The form automatically saves changes every 30 seconds.

The team wants criteria for the agreed behavior, not additional features. It will supply the agreement again and review the revised draft together.

Which follow-up instruction would best produce suitable acceptance criteria?

• A. Rewrite the criteria as scenarios with actions, visible feedback, and retained or restored values. Derive each expected outcome from the team’s agreement.
• B. Rewrite the criteria as internal checks for validation, storage, and error-handling components. Describe the expected component results for each of the agreed behaviors.
• C. Rewrite the criteria as user goals and benefits for partial saving, recovering from errors, and resuming work. Base those goals on the team’s agreement.
• D. Rewrite the criteria in Given/When/Then form, preserving the draft’s stated behavior. Use the team’s agreement to label the purpose of each scenario.

Best answer: A

Acceptance criteria make agreed product behavior observable and checkable. Useful direction supplies that behavior and its edge conditions rather than inviting the assistant to design additional features.

Here, the criteria should cover saving with missing required fields, confirmation after success, error feedback with entries retained after failure, and reopening the last successfully saved values. Autosaving every 30 seconds is an unsupported addition. Terms such as “reliable” and “easy to use” also need specific outcomes before they can guide acceptance checks.

A structured format can help, but formatting alone does not correct unsupported behavior. The team still reviews the revised criteria together and tests the implementation; an AI-produced draft does not establish acceptance or completion.

Answer choices explained:

A — Correct: Observable scenarios make the agreed success, failure, and reopening behavior checkable while keeping expected outcomes within the agreed scope.

B — Incorrect: Internal component checks can support testing, but do not establish the complete user-facing confirmation, error, retention, and reopening outcomes.

C — Incorrect: Goals and benefits describe why draft saving matters, but do not establish the observable conditions needed to check its behavior.

D — Incorrect: Changing the format preserves the unsupported autosave behavior and does not replace vague qualities with the agreed outcomes.

Question 25

Topic: AI Accountability

A product team is preparing a one-page research brief for the Product Owner, who will use it to prioritize checkout fixes for the next release. The organization’s reporting guidance calls for concise, constructive wording grounded in observed user experiences.

Research evidence:

• Six of eight participants placed an order independently and described checkout as straightforward.
• Two participants who use keyboard navigation could not reach the “Place order” control. A facilitator submitted their orders.
• One of those participants said:

“I would have to ask someone else to finish my purchase.”

AI-produced summary:

“Participants found checkout straightforward. A small number suggested minor navigation refinements that could follow launch.”

Which revision best preserves affected users’ experiences for the Product Owner’s decision?

• A. Retain the summary, add the participant’s quotation as supporting evidence, and note that AI helped prepare the brief.
• B. Report all eight checkout attempts as successful, note facilitator assistance for two, and list keyboard navigation among suggested post-launch enhancements.
• C. Report the positive experiences alongside the two keyboard users’ inability to purchase independently, keeping the access barrier in the main summary.
• D. Report favorable usability findings, treat the two keyboard failures as inconclusive, and defer their release implications until a larger study establishes prevalence.

Best answer: C

Faithful synthesis preserves the meaning and consequences of users’ experiences, not just the majority view. The keyboard users did not merely request polish: they could not place orders independently. Calling this a minor refinement replaces their reported loss of independence with a positive interpretation and could lead the Product Owner to defer work without understanding its impact.

A constructive brief can report both the six successful experiences and the observed keyboard barrier in neutral language. The research establishes that the barrier occurred in this study; it does not establish its prevalence across all customers or dictate the release decision. The team remains responsible for checking that the final AI-assisted summary represents the evidence fairly.

Answer choices explained:

A — Incorrect: An AI-use label and a quotation do not correct the main narrative’s misleading description of blocked checkout as a minor refinement.

B — Incorrect: Facilitator-submitted orders do not demonstrate independent access; counting them as successes and calling the barrier an enhancement understates its impact.

C — Correct: This preserves successful experiences and the keyboard users’ loss of independence, making the consequential finding available for release prioritization.

D — Incorrect: A larger sample could inform prevalence, but the observed inability to place an order independently already belongs in the release evidence.

Questions 26-40

Question 26

Topic: AI Foundations

A product team needs an internal briefing about onboarding problems customers reported last week. The evidence is in internal support reports that a team member can export. Public release notes describe shipped features but contain no customer reports.

Both available assistants can summarize uploaded files and produce the required briefing format:

• Cedar is approved for these internal reports but has no connection to the support system.
• Brook can search public websites but is approved only for public information. Under the team’s policy, removing customer names does not make internal report content public.

The team will check the draft against the support reports before sharing it. Which use of an assistant best fits the task?

• A. Use Brook with public release notes to draft the briefing about reported onboarding problems.
• B. Use Brook with name-redacted support reports to draft the briefing about reported onboarding problems.
• C. Use Cedar with report titles to retrieve the reports and draft the briefing about reported onboarding problems.
• D. Use Cedar with exported support reports to draft the briefing about reported onboarding problems.

Best answer: D

Tool suitability depends on the task, permission to process the necessary information, and a practical way to access that information. Cedar’s missing connection does not prevent this work because the team can upload authorized exports containing the actual customer reports.

Approval and access are separate: an approved assistant does not automatically have access to internal systems. Likewise, Brook’s public search capability cannot supply private observations, and removing names does not change the stated data restriction. The team’s final review remains necessary because permission to use a tool does not establish the accuracy of its output.

Answer choices explained:

A — Incorrect: Release notes describe shipped features rather than customer observations, so they cannot substantiate the problems actually reported last week.

B — Incorrect: The stated policy still classifies name-redacted report content as internal, so Brook is not authorized to process it.

C — Incorrect: Approval to process internal material does not grant Cedar access to the support system; report titles do not supply the reports’ contents.

D — Correct: Cedar is authorized to process the reports, and uploading the exports supplies the evidence it cannot retrieve directly.

Question 27

Topic: AI Prompting & Iteration

A Product Owner is preparing options for a discussion with Developers and stakeholders.

• Support records show repeated checkout failures.
• Customer interview notes contain requests for better search filters.
• The team can pursue only one improvement next. No priority or delivery commitment has been agreed.

The Product Owner plans to use this prompt:

“Act as a Product Owner. Using the supplied notes, draft two improvement options, compare their likely value and trade-offs, and recommend one for my review.”

Which interpretation of the role instruction is most appropriate?

• A. The role instruction must be removed to preserve accountability; AI should provide factual summaries rather than comparisons or recommendations.
• B. The role instruction is acceptable for summarizing evidence; requesting a recommendation assigns the Product Owner’s prioritization decision to AI.
• C. The role instruction transfers prioritization authority to AI; the Product Owner regains that authority when reviewing the recommendation.
• D. The role instruction provides a useful analytical perspective; the recommendation remains input to the Product Owner’s prioritization decision.

Best answer: D

Role context directs attention to considerations such as product value and trade-offs; it does not transfer Product Owner accountability. AI can draft options and recommend one while the Product Owner remains accountable for value and effective Product Backlog management.

Here, the requested output is material for review and discussion, not an approved priority or delivery promise. The Product Owner should check that the comparison faithfully represents both the support records and interview notes, and evaluate its assumptions before using it. A polished recommendation cannot create stakeholder agreement or a commitment to deliver.

Answer choices explained:

A — Incorrect: Human accountability does not limit AI to factual summaries; comparing alternatives and recommending one can support a human prioritization decision.

B — Incorrect: A recommendation for review is decision support; it does not itself select or authorize the Product Backlog priority.

C — Incorrect: Assigning a role changes the requested perspective, not who holds Product Owner accountability; reviewing output does not transfer authority back.

D — Correct: The assistant is asked to draft and recommend for review, while accountability for value and Product Backlog management remains with the Product Owner.

Question 28

Topic: AI Foundations

A product team is testing AI summaries of approved customer interview extracts to help choose onboarding problems for further discovery. Each summary must stay within 150 words.

Pilot observations:

• The summaries accurately captured common requests for faster setup.
• They repeatedly omitted a completion barrier reported by two keyboard-only customers.
• Reviewers initially rated the summaries highly for clarity and usefulness. Later source checks supported every included claim but revealed the omission.

For the next small pilot, analysts will prepare source-based notes before viewing AI output. The team will track drafting, review and correction time, and retain the word limit.

Which follow-up would most directly test whether the observed omission problem has been corrected?

• A. Refine the request for common and contrasting needs, then compare coverage of decision-relevant views with the analysts’ independent source-based notes.
• B. Refine the request for evidence-backed findings, then compare each included claim with its interview source and track the proportion supported.
• C. Refine the request for balanced coverage, then compare each summary’s overall usefulness rating with the ratings from the original pilot.
• D. Refine the request for representative themes, then compare each summary’s leading themes with the analysts’ ranking of the most frequently reported needs.

Best answer: A

The observed failure is incomplete coverage, not unsupported claims or poor writing. A summary can accurately report common needs while distorting discovery priorities by omitting a small group’s blocking issue.

The next experiment should request common and contrasting needs, then check their coverage against independently prepared source notes. Success means preserving decision-relevant evidence within the word limit, not giving every comment equal space. Remaining omissions should inform further adjustments to the guidance or the task assigned to AI before wider use.

Independent source analysis helps retain human judgment. Recording review and correction time alongside drafting time also tests whole-task benefit rather than assuming faster drafts mean less work.

Answer choices explained:

A — Correct: Comparing coverage with independent source notes directly tests whether important minority evidence survives the revised summarization approach.

B — Incorrect: Checking included claims establishes factual support, but summaries can pass that check while still omitting the keyboard-only customers’ barrier.

C — Incorrect: Overall usefulness ratings already looked favorable despite the omission, so repeating that measure would provide weak evidence that coverage improved.

D — Incorrect: Agreement on frequent themes repeats a demonstrated strength and can remain high when a consequential minority view is missing.

Question 29

Topic: AI Accountability

A product team uses AI to turn internal release notes into short customer updates. A review of the last 12 updates found three unapproved release promises. Two appeared during a final shortening revision.

Internal note: Bulk export target: 10 September, conditional on load testing; no customer date approved.

Sent update: Bulk export will be available on 10 September.

Senders already check tone and readability. The team can add about two minutes to that review, but cannot add another approval handoff. It will trial one shared check for two weeks, recording review time and unsupported commitments found before or after sending.

Which shared check would best address the observed failure?

• A. Have the sender check that the final revision’s prompt distinguishes tentative internal targets from approved customer release commitments.
• B. Have the sender check that the final update includes a standard caveat explaining that release dates remain subject to change.
• C. Have the sender check that each availability claim in the final update reflects the certainty and approval status recorded in the notes.
• D. Have the sender check that feature names and dates in the final update match those recorded in the internal notes.

Best answer: C

A shared check should target the actual handoff error: tentative internal information becoming a public commitment. Accurate names and dates are not enough; the wording must reflect what is confirmed, conditional, or unapproved.

The sender should compare release claims in the final version, including shortening revisions, with the status recorded in the notes and remove or rewrite unsupported promises. This adds a focused content check to the existing review rather than another approval layer. During the two-week trial, the team can examine corrections, unsupported commitments that reached customers, and review time to assess whether the routine reduces the observed problem at acceptable effort.

Answer choices explained:

A — Incorrect: Clear instructions can improve drafting, but checking the prompt does not verify that the final revision preserved the distinction.

B — Incorrect: A general caveat does not establish whether a specific release promise is supported or approved, so the misleading commitment can remain.

C — Correct: Comparing final claims with their source status catches a tentative target becoming a promise, including changes introduced during shortening.

D — Incorrect: Matching names and dates would miss the observed error: an unchanged target date was presented as an unapproved release promise.

Question 30

Topic: AI Foundations

A product team uses an AI assistant to turn a customer’s request into draft acceptance criteria for a support queue. The draft must capture the requested behavior, not propose a new prioritization policy.

Customer request:

List unresolved cases before resolved cases. Within each group, show the most recently updated first.

AI draft:

List cases by most recent update first. If update times match, list unresolved cases before resolved cases.

Which pair of cases would most directly reveal the interpretation error when comparing the draft with the request?

• A. An unresolved case updated today and a resolved case updated yesterday.
• B. An unresolved case updated yesterday and a resolved case updated today.
• C. Two unresolved cases, one updated yesterday and the other updated today.
• D. An unresolved case and a resolved case, both updated at the same time.

Best answer: B

An assistant can use the right product vocabulary while changing the relationships that determine meaning. Here, the customer makes status the primary ordering rule and update time the secondary rule. The draft reverses that priority: update time comes first, with status used only to break ties.

A focused check uses cases where the criteria point in different directions. An older unresolved case must precede a newer resolved case under the request, but the draft produces the opposite order. This comparison reveals the meaning error without broader customer research or repeated AI generation. Revise the criteria, then check the revised version against the original request before using it.

Answer choices explained:

A — Incorrect: Both rules put the newer unresolved case first because status and recency favor the same case.

B — Correct: The request puts the older unresolved case first; the draft puts the newer resolved case first, exposing the reversed priorities.

C — Incorrect: Both rules place the more recently updated unresolved case first, so this comparison does not expose the changed priority.

D — Incorrect: Both rules put the unresolved case first when update times match, so this comparison does not expose the changed priority.

Question 31

Topic: AI Accountability

A Product Owner is comparing guided setup with faster data import. The Product Goal is to reduce delays before new customers can use the product. Only one improvement fits in the next release.

Ranking review:

• Six of eight observed setup sessions stalled because instructions were unclear.
• An approved AI assistant ranks faster data import first, using the Product Owner’s scoring rules and Developers’ effort estimates.
• The assistant applied the rules correctly. Its highest score depends on an assumption that import delays affect most new customers, which the available customer evidence does not establish.

Which response best supports the Product Owner’s accountability for value and Product Backlog ordering?

• A. The Product Owner confirms effort with Developers and adopts the ranking once the delivery estimates have been validated.
• B. The Product Owner assesses expected benefits against customer evidence and decides the order, using the ranking as analytical input.
• C. The Product Owner obtains stakeholder approval of the scoring rules and adopts the ranking based on that approval.
• D. The Product Owner requests a second AI comparison and adopts the ranking once both assistants agree on the order.

Best answer: B

The Product Owner remains accountable for maximizing product value and for effective Product Backlog management, including ordering. AI can provide comparisons and rankings, but correct application of a scoring rule does not make its underlying assumptions reliable. Here, observed setup difficulties are customer evidence; the claim that import delays affect most customers is not established. The Product Owner should weigh those different levels of support when assessing benefits against the Product Goal and deciding the order. The recommendation can be retained or changed. The ranking informs a human value decision rather than proving that the highest-scoring feature is most valuable.

Answer choices explained:

A — Incorrect: Confirming effort can improve cost estimates, but it does not establish the customer benefit driving the first-ranked feature.

B — Correct: The missing support concerns expected value; evaluating that support and deciding the order keeps the assistant in an advisory role.

C — Incorrect: Approval of scoring rules does not validate the import-delay assumption on which the highest score depends.

D — Incorrect: Agreement between assistants is not independent customer evidence and can leave the same unsupported value assumption unchecked.

Question 32

Topic: AI Quality & Confidence

A product team uses AI to draft a recommendation for handling refunds after subscription plan changes. It specifies refund eligibility and when agents should escalate a request. A colleague approves the wording after checking grammar and clarity; that colleague has never applied the refund policy.

Before agents use the recommendation in a limited pilot, the team can schedule one additional reviewer. Every candidate will receive the draft, the current refund policy, and representative past cases.

Which colleague has the most relevant experience for the substantive review?

• A. A quality analyst who regularly tests billing screens and checks that refund forms record inputs correctly.
• B. A Product Owner who regularly prioritizes billing improvements and interprets customer feedback about plan changes.
• C. A billing operations specialist who regularly applies refund eligibility rules and resolves plan-change disputes.
• D. A support communications lead who regularly edits refund messages and maintains templates used by agents.

Best answer: C

Substantive human review requires enough task understanding to judge a recommendation’s correctness and suitability. The billing specialist’s experience applying refund rules and resolving plan-change disputes directly matches the decisions agents will make.

The reviewer should compare the proposed eligibility and escalation guidance with the current policy and work through representative cases. Any corrections must be checked in the final version before use. Grammar review improves presentation, but it does not establish that the underlying recommendation is sound. Likewise, relevant job titles or human approval alone are insufficient: confidence comes from a competent person actually inspecting the content against appropriate evidence.

Answer choices explained:

A — Incorrect: Checking screens and recorded inputs addresses interface behavior, rather than whether proposed refund eligibility and escalation decisions reflect policy.

B — Incorrect: Prioritization and feedback analysis inform product value, but do not demonstrate experience applying the refund policy to individual cases.

C — Correct: Regularly applying refund rules provides relevant task understanding for checking the recommendation against policy and representative plan-change cases.

D — Incorrect: Editing refund messages supports readability and consistency, but does not demonstrate the policy knowledge needed to evaluate eligibility and escalation.

Question 33

Topic: AI Accountability

A product team drafts customer-facing guidance using approved product information. Some teammates use an approved AI assistant; others draft manually. AI use is optional, and material AI assistance must be recorded at handoff.

A review of the last two weeks found:

• Both drafting methods produced unsupported claims.
• Drafts checked against approved information before handoff met the publication standard, regardless of drafting method.
• Frequent AI users were not consistently more accurate than other authors.
• AI-assisted first drafts were faster, but review and correction time was not measured.

The team wants a one-sprint experiment to improve the accuracy and delivery speed of publishable guidance.

Which working agreement is best supported by these observations?

• A. Use full source checks for AI-assisted drafts and spot checks for manual drafts; assess the experiment using correction rates and total preparation time.
• B. Use self-approval for frequent AI users and peer approval for other authors; assess the experiment using correction rates and total preparation time.
• C. Use a documented source check for every draft; assess the experiment using correction rates and total preparation time through final review.
• D. Use a documented source check for every draft; assess the experiment using first-draft speed and the number of regular AI users.

Best answer: C

Quality expectations should attach to the work, not to whether or how often a teammate uses AI. Unsupported claims occurred with both methods, while source-checked drafts from either method met the publication standard. This supports testing a common, documented check while retaining the agreed disclosure of material AI assistance.

The experiment should measure corrections and total preparation time, including drafting, review and rework. Faster first drafts or increased AI usage alone would not demonstrate better delivery. These observations support a small experiment, not a conclusion that either drafting method causes better performance. Reviewing the results together lets the team refine its shared practices without treating tool use as a measure of competence.

Answer choices explained:

A — Incorrect: Manual drafts also contained unsupported claims, so drafting without AI does not justify reducing their source verification.

B — Incorrect: Frequent AI users were not consistently more accurate, so tool-use frequency does not justify different approval expectations.

C — Correct: A shared check addresses the observed failure across both methods, while correction rates and full-cycle timing assess publishable quality and delivery speed.

D — Incorrect: First-draft speed omits review and correction effort, while increased AI participation does not establish more accurate or faster publishable guidance.

Question 34

Topic: AI Quality & Confidence

A product team asks AI to shorten a standalone help-panel message for members of paid workspaces. The panel should be easy to scan and help readers judge whether saved views are currently available to them. The existing draft accurately reflects the product’s behavior.

Existing draft:

Members of paid workspaces can use saved views only while their workspace’s data-source connection is active.

AI revision:

Members of paid workspaces can use saved views.

Which comparison best applies the team’s criteria?

• A. The revision is more suitable: plan eligibility defines availability, while connection status only affects how the feature is used.
• B. Both versions are suitable: the phrase “can use” conveys that any necessary connection conditions still apply.
• C. The existing draft is more suitable: it preserves an availability condition that the shorter, easier-to-scan revision leaves out.
• D. Neither version is suitable: an accurate availability notice also needs instructions for establishing the data-source connection.

Best answer: C

Readability and accuracy are separate review criteria. The shorter revision is easy to scan, but it turns conditional availability into an unqualified statement. Paid membership identifies the audience; it does not replace the requirement for an active data-source connection. A member with an inactive connection could therefore be misled about access.

The existing draft better serves the panel’s purpose. A further edit can improve its wording while keeping the active-connection condition. After editing, compare the final message with the verified product behavior. An availability notice does not need to become a full setup guide to remain accurate.

Answer choices explained:

A — Incorrect: The verified behavior makes an active connection a condition of availability, not merely a detail about operating an otherwise available feature.

B — Incorrect: The words “can use” do not communicate the specific active-connection requirement, so the messages do not preserve the same availability information.

C — Correct: The existing draft preserves the active-connection requirement; the revision names the eligible audience but omits that additional condition for access.

D — Incorrect: The panel’s purpose is to communicate availability; preserving the connection condition does not require instructions for establishing that connection.

Question 35

Topic: AI Quality & Confidence

A product team is preparing a funding memo for a larger trial of a redesigned account-setup flow. The team can edit the memo but cannot collect new measurements before the funding review.

Verified source notes:

• Usability study: Six of eight volunteers completed setup in the prototype without assistance. The current flow was not tested.
• Support review: Support-request volume for the redesigned flow has not been measured.

AI-written paragraph:

These early findings may not generalize to all customers. The usability study and support review show that the redesigned flow reduces setup-related support requests.

Which revision would best align the paragraph with the available evidence?

• A. Limit the support-reduction finding to the eight volunteers and describe effects for other customers as uncertain.
• B. State the observed setup-completion result and present reduced support demand as a hypothesis for the larger trial.
• C. Make the opening caveat more prominent and retain reduced support demand as the study’s main finding.
• D. Use successful task completion as indirect confirmation of reduced support demand, retaining the caution about generalizing.

Best answer: B

Confidence means warranted reliance for a specific use, not reassurance created by cautious wording. An uncertainty disclaimer cannot make an unsupported categorical claim reliable.

Here, the evidence establishes that six of eight volunteers completed prototype setup without assistance. It provides neither a comparison with the current flow nor a measurement of support requests. Reduced support demand is therefore a possible benefit to test, not a demonstrated outcome.

The memo can still inform the funding discussion by separating the observed result from the proposed benefit. Genuine sources must support the specific claim attributed to them. After revision, the team should check the whole paragraph, not merely confirm that it contains a caveat.

Answer choices explained:

A — Incorrect: Support reduction was not measured even for the volunteers, so narrowing the population does not make the claim supported.

B — Correct: The notes support reporting task completion, while reduced support demand remains an unmeasured benefit that a larger trial could investigate.

C — Incorrect: A more prominent caveat would leave the unsupported support-reduction claim unchanged; cautionary wording does not supply missing evidence.

D — Incorrect: Prototype task completion does not confirm reduced support demand, particularly when neither support volume nor performance with the current flow was compared.

Question 36

Topic: AI Quality & Confidence

A product team is simplifying an AI-written support reply for a customer unfamiliar with technical computing terms. The customer edited a shared checklist on a tablet while offline, and the upload is still pending.

The team has confirmed that this draft is technically accurate:

Your edits are persisted locally. Collaborator visibility requires successful synchronization from the originating device.

The revision must use everyday language, explain whether the edits are saved, and preserve the condition for coworkers to see them. Which revision best meets these criteria?

• A. Your edits are saved on your tablet. Your coworkers can see them once they close and reopen the checklist on their own devices.
• B. Your edits are saved on your tablet. Your coworkers can see them as soon as your tablet is connected to the internet again.
• C. Your edits are saved in the shared checklist. Your coworkers can see them once their devices refresh and display the latest saved version.
• D. Your edits are saved on your tablet. Your coworkers can see them after your tablet finishes sending the changes to the shared checklist.

Best answer: D

An audience-appropriate revision must preserve meaning while reducing jargon. Here, the edits are already saved on the customer’s tablet, but they are not yet available in the shared checklist. Coworker access depends on the tablet successfully completing the upload.

Replacing ‘persisted locally’ with ‘saved on your tablet’ and ‘successful synchronization’ with ‘finishes sending the changes’ retains that distinction. Review the final wording against both the intended audience and the verified behavior. Plain language alone is not enough: a readable reply that changes where information is saved or when it becomes available is still unsuitable.

Answer choices explained:

A — Incorrect: Reopening the checklist on another device cannot reveal edits that remain on the customer’s tablet and have not been uploaded.

B — Incorrect: Restoring the internet connection allows uploading to begin; it does not establish that the upload has finished.

C — Incorrect: The edits have not yet been uploaded, so they are not saved in the shared checklist for coworkers to retrieve.

D — Correct: This preserves both facts: the tablet already holds the edits, and coworkers can access them after the upload successfully completes.

Question 37

Topic: AI Foundations

A Product Owner is preparing a stakeholder update to guide further investment in an interface redesign. An AI assistant receives only this evidence:

• 30-day subscriber retention was 68% for the pre-launch cohort and 76% for the post-launch cohort.
• The interface redesign, a subscription price reduction, and expanded support were introduced in the same week.

The assistant drafts:

The interface redesign caused an eight-percentage-point improvement in retention.

The update is due today. Existing retention records, rollout details, and customer comments are available for review, but a new experiment cannot finish before the update.

Which response best addresses the weakness in the AI summary?

• A. Report that retention rose after the launches, and check whether records permit comparisons separating the interface from pricing and support changes.
• B. Report the interface as the likely cause, and check existing customer comments for descriptions of easier navigation.
• C. Report the interface as the cause, and verify the increase by recalculating both cohorts’ retention rates from source records.
• D. Report the changes collectively as the cause, and verify that all three launches preceded the post-launch cohort’s retention measurement.

Best answer: A

A before-and-after increase does not, by itself, establish what caused it. The interface redesign coincided with lower prices and expanded support, so the AI has converted an observed association into an unsupported causal explanation. Attributing the improvement to the three changes collectively would also exceed the evidence.

For today’s update, report that retention increased from 68% to 76% following the launches, leaving causal attribution unresolved. Review rollout and retention records to see whether groups differed in interface exposure while experiencing the same pricing and support. Such comparisons can help investigate attribution, although other differences between groups still require scrutiny. If no useful comparison exists, preserve the uncertainty rather than treating the AI’s fluent statement as evidence.

Answer choices explained:

A — Correct: The evidence establishes a before-and-after association; rollout differences may provide useful comparisons for investigating the interface’s contribution.

B — Incorrect: Comments about easier navigation may support a usability interpretation, but they do not establish the interface as the likely cause of increased retention.

C — Incorrect: Recalculating retention can confirm the observed increase, but it cannot distinguish the interface’s contribution from simultaneous pricing and support changes.

D — Incorrect: Confirming the timing does not establish that the combined changes caused the increase; that attribution also exceeds the supplied evidence.

Question 38

Topic: AI Accountability

A product team wants CopyDesk, an AI writing assistant, to draft a customer announcement. The Product Owner will review the draft before publication.

Organization rules and permissions:

• CopyDesk is approved to draft public copy using public information.
• Additional input data classes require approval from the AI governance team.
• The roadmap is classified Internal; its planned capabilities have not been cleared for publication.

A teammate proposes removing project names and target dates from a roadmap paragraph, then pasting the remaining description of an unannounced invoice-export feature into CopyDesk. The requested output would be a generic teaser with no release commitments.

Which judgment about the proposed use is supported?

• A. The use requires AI governance approval because the remaining feature description is still Internal input outside CopyDesk’s existing permission.
• B. The use is covered because removing project names and target dates makes the remaining feature description public information rather than Internal material.
• C. The use requires Product Owner authorization because deciding how roadmap information supports customer communication falls within accountability for product value.
• D. The use is covered because the final teaser is intended for public release and will receive Product Owner review before publication.

Best answer: A

Tool approval applies to permitted inputs and uses, not just the intended audience of the output. CopyDesk may draft public copy, but its approval covers only public information. The proposed excerpt still describes an unannounced capability, so removing names and dates does not bring it within that boundary. Product Owner review can check product messaging but cannot extend the assistant’s data permissions. The team can use already-public source material under the existing approval, or obtain approval from AI governance before submitting the Internal roadmap details. Requesting a generic output does not change the sensitivity of the input.

Answer choices explained:

A — Correct: The proposed input contains an unreleased capability, so AI governance must approve the additional data class before it is submitted.

B — Incorrect: Removing names and dates does not make an unannounced capability public or change the classification of its description.

C — Incorrect: Product Owner accountability for value does not extend tool permissions; the organization assigns approval of additional input data classes to AI governance.

D — Incorrect: The teaser’s intended audience and later review do not authorize submitting Internal source material to a tool approved only for public information.

Question 39

Topic: AI Accountability

A Developer uses the organization’s approved AI workspace to summarize support themes for a Product Backlog discussion.

Organization rules:

• The workspace is approved for de-identified support feedback. Identifiable billing-dispute records are classified as Restricted and excluded from that approval.
• Suspected disclosure of Restricted data must be reported promptly through Security intake. Security assesses severity and decides whether customer notification is needed.

Observed event: The Developer accidentally submits a dispute record containing a customer’s name and private billing notes. They delete the chat, which disappears from their history. They have no confirmation about retained copies or access to the submitted data.

Which response is most appropriate?

• A. Report the incident to Security intake now, describing the data category and the deletion already attempted.
• B. Log a handling error within the team, and close it once the submitted chat is absent from history.
• C. Report the incident to Security intake now, and ask customer support to notify the affected customer immediately.
• D. Ask the workspace administrator to verify deletion, and report through Security intake if retained copies of the record are found.

Best answer: A

Approval applies to a specified use and data category, not every action performed in an approved tool. Identifiable billing-dispute records fall outside this workspace’s permission, so the organization’s suspected-disclosure process applies.

Deleting a chat from history does not demonstrate removal of retained copies or rule out access. A proportionate response is a prompt report through Security intake, identifying the data category, the known submission and deletion, and what remains uncertain. Security can assess exposure and determine any further response. Reporting a suspicion does not require proof of lasting exposure or mean declaring a major breach.

Answer choices explained:

A — Correct: Restricted data was submitted outside the workspace’s approved use, triggering prompt reporting even though the chat was subsequently deleted.

B — Incorrect: Disappearance from chat history does not establish removal of retained data or satisfy the required Security reporting process.

C — Incorrect: Prompt reporting is appropriate, but initiating customer notification bypasses Security’s responsibility to assess severity and decide whether notification is needed.

D — Incorrect: The reporting threshold is suspected disclosure, so confirmation of retained copies must not become a prerequisite for notifying Security.

Question 40

Topic: AI Accountability

A product team is piloting an AI assistant that recommends urgency for payment-support requests. Underestimating urgency can delay help for customers who cannot complete a payment.

Pilot evidence:

• The team handpicked eight common requests, four in each supported language.
• All eight recommendations matched urgency ratings independently assigned by experienced support staff.
• Staff review every recommendation before it changes queue priority.

For an update to support leaders, a teammate proposes:

The assistant is unbiased across both supported languages.

Company guidance requires claims to distinguish observed results from conclusions about the wider customer population. Which interpretation is best supported?

• A. The results support fairness for common payment-support requests, with further evaluation needed for unusual request types.
• B. The results support agreement with staff on these requests, with further evaluation needed to assess language-group fairness.
• C. The results support fairness for staff-reviewed recommendations, with further evaluation needed before allowing automated prioritization.
• D. The results support fairness across the two supported languages, with further evaluation needed before adding another language.

Best answer: B

Agreement on eight selected examples is evidence about those examples, not proof of fairness across language groups. Equal numbers from both languages do not show that the sample covers differences in wording, urgency or customer circumstances. The assistant could match staff on these requests and still systematically under-prioritize other requests in one language.

The update can report the observed agreement while stating that language-group fairness remains unestablished. Staff review is a safeguard against delayed help, not evidence that the assistant is unbiased. Broader, appropriately varied evaluation is needed to assess fairness; successful examples alone cannot justify the proposed claim.

Answer choices explained:

A — Incorrect: The selected examples do not establish consistent treatment across either language within the broader category of common requests.

B — Correct: All eight recommendations matched staff ratings, while the small, selected sample does not establish fairness beyond those examples.

C — Incorrect: Human review can reduce the consequences of poor recommendations, but its presence does not establish fairness across language groups.

D — Incorrect: Including four requests per language does not establish how the assistant treats the wider range of customers in either group.

Turn uncertainty into a useful review task

The official passing score is 85%. Achieving that score here does not establish readiness or measured equivalence. This practice supplements the mandatory Scrum.org course and does not provide its completion or assessment passcode. Read how to interpret practice scores .

Continue in PM Mastery

Use varied topic and mixed questions, option explanations and saved progress to revisit uncertainty. Paid PM Mastery access covers its own PM bank library; the static public set stays free and grants no issuer entitlement.