Question 360
A release runbook lists technical rollback commands but does not say who can order rollback or which symptoms warrant it. What addition is needed?
- Decision authority, trigger conditions, and communication responsibilities
- More screenshots of the deployment interface without changing the decision process
- A statement that rollback should be avoided whenever possible
- A requirement that every user approve rollback individually
Explanation
Correct Answer
A:Commands explain how to act but not when or by whose authority. The runbook needs an operational decision process so responders can use those commands promptly and appropriately.
Incorrect Answers
B:Screenshots may help execution but do not resolve authority or trigger conditions. The missing information concerns the decision to initiate rollback.
C:A general preference provides no actionable criteria. Responders need conditions and authority for deciding when rollback is warranted.
D:That process is impractical for incident response and does not assign accountable authority. The organization needs a defined operational owner.
Question 361
After repeated incidents, teams restore service but never update tests, architecture, or training. What lifecycle practice is missing?
- A longer incident status message with no assigned follow-up work
- Track restoration time as the only measure of incident improvement
- A blameless review that turns contributing factors into owned corrective actions
- File the incident transcript without assigning any remediation owner
Explanation
Correct Answer
C:Restoration ends the immediate outage but does not reduce recurrence by itself. A structured learning process connects evidence to tracked improvements in the system and operating practice.
Incorrect Answers
A:Communication records the event but does not implement learning. Improvements need owners and completion criteria to affect future behavior.
B:Fast recovery is valuable but does not show that recurring causes were addressed. Owned follow-up work is needed to reduce recurrence and improve the system.
D:Archiving preserves information but does not turn it into an action. Follow-up needs ownership and completion criteria to change future behavior.
Question 362
A vendor cites a successful customer deployment as evidence that the proposed assistant will work. That customer uses a different data type, language, and decision authority. What should the architecture review do?
- Treat any deployment by the same vendor as equivalent acceptance evidence
- Identify which conditions transfer and validate the material differences locally
- Discard the reference entirely because no two organizations are identical
- Adopt the reference customer's acceptance thresholds without consulting local owners
Explanation
Correct Answer
B:A reference can provide useful evidence without proving fit to every workflow. The differing conditions define what remains to be tested for this deployment.
Incorrect Answers
A:Vendor identity does not make tasks, data, or authority equivalent. The known differences can materially change performance and risk.
C:The reference may still inform feasibility or operating considerations. The useful response is to separate transferable evidence from unvalidated differences.
D:Thresholds reflect the other organization's needs and consequences. The local deployment requires criteria appropriate to its own workflow.
Question 363
A pilot has already spent $300,000 that cannot be recovered. Continuing would cost another $80,000 and is expected to produce only $25,000 in additional benefit, with no other strategic value identified. Which basis should guide continuation?
- Compare the additional future benefit with the additional future cost
- Continue until total reported benefits equal the amount already spent
- Continue because stopping would make the original spending a loss
- Allocate the past spending across more releases to justify the next release
Explanation
Correct Answer
A:The recoverable decision concerns what happens next. Under the stated assumptions, another $80,000 buys only $25,000 of benefit; prior spending does not make that increment attractive.
Incorrect Answers
B:Past spending cannot be recovered merely by continuing. Chasing a historical total can add further losses when the next increment has negative expected value.
C:The original spending is already unrecoverable in this scenario. Continuing changes future costs and benefits, not the recoverability of that expenditure.
D:An accounting allocation does not change incremental cash flows or benefits. Spreading the old cost across releases cannot create value in the proposed next investment.
Question 364
A late change removes a field from a case record. The team cannot identify which prompts, tools, tests, and reports depend on it. What maintained artifact would help assess such changes?
- A list of open defects without links to affected requirements
- A chronological list of project meetings without recorded dependencies
- A deployment inventory containing service names but no contract relationships
- A mapping from requirements and data contracts to their consumers and tests
Explanation
Correct Answer
D:Traceability connects a changed assumption or interface to the parts that rely on it. That makes impact assessment more systematic than searching from memory.
Incorrect Answers
A:Defects can reveal known problems while omitting healthy components that still depend on the field. Traceability should cover relevant consumers, not only current failures.
B:Meeting history may contain context but does not reliably identify consumers of a particular field. The team needs explicit relationships across the implementation.
C:Service names establish what is deployed, not which component consumes the changed field. The impact assessment needs data and requirement dependencies.
Question 365
A critical dependency has a likely delivery range of two to six weeks. The sponsor requests an exact release date. What should the project update communicate?
- No date-related information until the dependency is fully resolved
- The latest possible date without explaining why it was chosen
- The earliest possible date presented as a guaranteed release commitment
- A conditional forecast, the dependency range, and the actions that could narrow uncertainty
Explanation
Correct Answer
D:The schedule depends on unresolved external timing. A conditional forecast supports planning without disguising uncertainty as a firm commitment.
Incorrect Answers
A:The team can provide a useful range and conditional milestones now. Withholding all forecast information prevents informed planning around the uncertainty.
B:A conservative date may still be useful, but unexplained padding obscures the actual uncertainty. Stakeholders need the dependency and conditions behind the forecast.
C:The best case is not a guarantee. Presenting it without conditions hides the dependency risk and can mislead downstream planning.
Question 366
A product team wants users to attach real customer records to public feedback tickets. The records are unnecessary for most usability reports. What feedback process is preferable?
- Use structured, minimized reports and a restricted channel when sensitive examples are truly needed
- Require full records on every ticket to maximize debugging convenience
- Disable all feedback because some reports might contain sensitive data
- Allow public attachments if the user checks a generic consent box
Explanation
Correct Answer
A:Most usability evidence can be collected without exposing customer records. A separate controlled route can handle the exceptional cases that require sensitive detail.
Incorrect Answers
B:Convenience does not establish necessity for broad disclosure. The proposal would expose sensitive information even when a minimal report is sufficient.
C:Feedback remains valuable and can be collected through an appropriate process. The identified risk calls for better handling rather than abandoning the learning channel.
D:A generic checkbox does not establish authority to disclose customer records publicly. The process should minimize data and control any necessary sensitive examples.
Question 367
A vendor reports 95% autonomous answer success in a demonstration, but operators corrected some answers before scoring. What evidence is needed to assess the autonomy claim?
- Count an answer as autonomous whenever the model produced its first draft
- Separate unassisted outcomes from corrected outcomes and measure intervention effort
- Exclude every corrected case from the denominator and retain the reported label
- Report only final answer correctness because operator time is outside model latency
Explanation
Correct Answer
B:These measures identify how much of the reported success depended on people. They permit an autonomy claim to be assessed without discarding the potential value of assisted operation.
Incorrect Answers
A:Producing an initial draft does not establish completion without assistance. Human repairs that determine correctness must be distinguished from unassisted outcomes.
C:Removing intervention cases can hide the workload that required assistance. The denominator and treatment of those cases must remain explicit for an autonomy rate to be meaningful.
D:Final correctness can describe the combined service but not autonomous performance. Excluding operator work conceals the human contribution to both quality and effort.
Question 368
A production assistant's source policies change monthly, but its review process checks only model versions once a year. What lifecycle mechanism is missing?
- A release note listing each new document without assessing its effects
- Review triggers covering relevant policy, data, tool, and use-case changes
- A broader annual agenda without event-driven review triggers
- A rule freezing policy documents for the lifetime of the deployment
Explanation
Correct Answer
B:Application behavior depends on more than the model version. The operating process should revisit readiness when inputs, dependencies, or authorized use change materially.
Incorrect Answers
A:Documentation provides visibility but does not determine whether prompts, tools, tests, or user expectations remain compatible. Material changes need an impact review.
C:An annual meeting can still miss consequential monthly changes. Review timing should respond to the dependencies that actually change the workflow.
D:The business needs legitimate policy updates. The assistant should have a managed response to authorized change rather than prevent its source owners from maintaining policy.
Question 369
Two proposed assistants each claim all of the same ten hours of weekly manual reconciliation as savings. Both would replace that single activity. How should the combined business case treat the estimates?
- Model the shared workflow once and count only incremental combined benefit
- Add both estimates because each project produced its own calculation
- Assign the full combined saving to whichever project has the larger budget
- Reject both projects because overlapping benefits imply neither can help
Explanation
Correct Answer
A:Overlapping interventions cannot each eliminate the same hours independently. The program estimate should represent the joint workflow rather than sum incompatible standalone claims.
Incorrect Answers
B:Separate calculations can still claim the same underlying work. Summing them would overstate the benefit available from the shared activity.
C:Budget size does not determine the actual eliminated work. Attribution should follow the workflow and each intervention's incremental contribution.
D:The projects may still create value individually or together. The issue is estimating their combined effect without counting the same improvement twice.
Question 370
The team is unsure whether a scanned archive can yield usable text. It has one week for a technical spike. Which deliverable best supports the next investment decision?
- A polished production interface around a single clean scanned page
- A complete long-term operating model without testing extraction quality
- Measured extraction quality on representative scans, failure categories, and a recommendation
- A count of OCR packages installed during the week
Explanation
Correct Answer
C:A feasibility spike should resolve the targeted technical uncertainty with bounded effort. Representative results and failure analysis explain whether and how to proceed.
Incorrect Answers
A:The interface does not establish feasibility across the archive. A clean example can conceal the document conditions that determine the investment decision.
B:Operating plans may be needed later, but they do not answer the spike's central uncertainty. The immediate deliverable should establish technical feasibility.
D:Installed tools measure activity rather than usable extraction. The decision needs observed quality and remaining failure modes.