Specimen paper

At a UK housing association, a manager reviews generated summaries of tenant complaint files. She reports that the summaries are consistently well organised and assured in tone, and proposes that checking be concentrated on the few that read as uncertain.

• A

  Accept it: an assured tone is a fair proxy for a well-sourced summary here.

  This is the human-drafting heuristic carried across unexamined: in a colleague's memo, an assured tone did tend to indicate a writer who had checked. It would be right only if the system varied its register according to how well supported a claim was, which is precisely what it does not do.

• B

  Accept it in part, checking only summaries shorter than the team's norm.

  This swaps one surface cue for another and feels more rigorous because length is objective and countable. Brevity would have to correlate with thin sourcing for it to work, but a short summary of a short file is as well supported as a long one, and an invented detail lengthens the text rather than shortening it.

• C

  Reject it, and ask the supplier to attach a confidence score to each summary.

  This gets the rejection right and the remedy wrong, which makes it the most tempting of the three. It treats a number the system generates about itself as though it were an external signal of ignorance; for it to help, the system would have to perceive the gap between what it knows and what it has asserted, and report that reliably.

• D

  Reject it: that register appears whether or not the content is correct.

  Correct: the assured register is produced whether or not the content is right, so it cannot be used to decide what gets checked.

A UK housing association can generate either of two outputs from a repairs contract: a twelve-page summary, which a manager can check clause by clause against the contract, or a single recharge figure, which can only be checked by working it out again. Which is the better candidate?

• A

  The figure, because one number takes far less of a manager's time to check.

  This confuses the size of the output with the cost of verifying it, and it is intuitive because scanning one number obviously takes seconds. It would hold only if reading the figure told the manager whether it was right; the only test actually available is to work the recharge out again, at which point the assistance has saved nothing.

• B

  The summary, because its errors can be found by comparison not derivation.

  Correct: errors in the summary can be located by comparing it with the contract, which costs far less than re-deriving a figure.

• C

  The summary, because prose errors are more obvious to a reader than numbers.

  Right choice, wrong reason, which is keyed as incorrect because the reasoning will not transfer to the next case. It assumes a wrong sentence reads differently from a right one; both arrive in the same confident register, and it is the availability of the contract to compare against, not the prose, that does the work.

• D

  The figure, because a costed output carries the larger downstream exposure.

  This is a true observation pointed at the wrong conclusion. Greater exposure argues for tighter control over the figure, not for choosing it as the output you generate; for this to be the answer, exposure would have to make verification cheaper, and it does the opposite.

An opportunity register at an Omani logistics group carries the entry 'AI for the customs clearance team, estimated saving 12%'. A reviewer wants to give one reason for sending it back. Which is the strongest?

• A

  It names no owner, so nobody is accountable for delivering it.

  This reads the register as a governance artefact and looks for the missing accountability line, which is a sound habit in project assurance. An owner would matter if the entry were otherwise capable of being delivered; naming somebody to own an unsizeable entry gives a person responsibility for a thing that cannot be measured, refused or stopped.

• B

  It names a team rather than a task, so it cannot be sized.

  Correct: with no task there is nothing to size, and that single defect blocks every step that would follow.

• C

  It names a saving with no baseline, so it cannot be verified.

  This is the closest wrong answer and it identifies a real absence, but it reads the symptom rather than the cause. A baseline is a measurement of a task, so there is nothing here to take a baseline of; fix the task and this objection disappears on its own.

• D

  It names no supplier, so procurement cannot begin the work.

  This assumes the answer will be bought and turns a scoping defect into a sourcing schedule. It would be the strongest objection only in an organisation that had already settled the task, the scope and the measure and was held up solely by procurement.

The operations team at Bank Sentosa Nusantara proposes automating production of a fourteen-page weekly reconciliation pack. A check of the document system shows nine of the pages have not been opened by any recipient in a year. What should be done first?

• A

  Stop producing the nine pages, then look at what remains.

  Correct: work that nobody reads should stop rather than be produced more efficiently.

• B

  Automate the pack whole, since the extra pages cost little.

  This is the marginal-cost argument, which is arithmetically true and strategically wrong. It would hold if the only cost of the nine pages were the minutes of producing them, but they also carry review time, a permanent maintenance burden and the standing implication to recipients that somebody thinks they matter.

• C

  Automate the five used pages and leave the rest manual.

  This looks like discrimination between useful and useless work, which makes it the most attractive wrong answer, since it appears to do both things at once. In fact it skips elimination entirely and treats the nine unread pages as a fixed input that must go on being produced, only by hand.

• D

  Survey the recipients on which pages they actually value.

  This reaches for opinion when a record of behaviour is already available at no cost. A survey would be right if the access data did not exist or could not be trusted; here it invites recipients to defend pages they have not opened, which is a reliable way of getting the wrong answer politely.

Two analysts at Al Batinah Freight have quietly stopped using the approved tool and gone back to an unapproved one on their own accounts. The control exposure is real and must be dealt with. What should the manager do first?

• A

  Block the unapproved tool, then ask what the approved route lacks.

  The two steps are right and the order is wrong, which is what makes this the strongest distractor on the paper. Once the block lands, the analysts have every reason to describe the approved route as adequate, and the next people in their position will hide the workaround rather than explain it.

• B

  Ask what the approved route costs them, then close the control gap.

  Correct: asking first preserves a requirements signal the firm cannot obtain any other way, and the control gap is closed immediately afterwards.

• C

  Issue a written warning, then remind the team of the tooling policy.

  This treats a design finding as a discipline matter, which is the reflex whenever the words control exposure appear. It would be the right first move if the analysts had been reckless with client data for their own convenience against a route that worked, and asking is precisely what establishes whether that is so.

• D

  Approve the second tool on trial, then compare the two next quarter.

  This responds to the evidence but skips the contract, putting client work into a service with no confidentiality undertaking, no deletion right and nobody to answer to a client. Approval could follow, but only once the service has been contracted and assessed, not as a way of ending the argument quickly.

An analyst at a Pune shared-service centre pasted a client's payroll file into a free consumer tool to reformat it and told her team leader the same day. The client contract requires notification of any processing outside approved systems. The account comes up for renewal in six weeks. What should the team leader do?

• A

  Contain the exposure, establish its scope and notify the client under the contract

  Correct: an express term requires notification, so the scope is established and the client is told.

• B

  Contain the exposure, retrain the analyst and record the matter on the internal log

  This is a complete and competent internal response that omits the only party with a contractual right to be told. It would be sufficient if the contract were silent on notification, in which case materiality would be the test and an internal record a reasonable answer.

• C

  Contain the exposure, take a full account and notify once renewal has been decided

  This sequences the disclosure around the commercial risk, and it is chosen most often because it sounds prudent rather than evasive. It converts a handling failure, which is survivable and correctable, into a concealment found at renewal or later; for it to be defensible the contract would have to leave the timing to the supplier, and it does not.

• D

  Contain the exposure, confirm the file has been deleted and treat the matter as closed

  This treats a deletion request made to a consumer service as proof of deletion, and proof of deletion as discharging a duty to notify. Neither holds: you cannot evidence what such a service retains, and the obligation was triggered by the processing itself rather than by what survives it.

A Saudi engineering consultancy bills time-charged work at SAR 500 an hour against a staff cost of SAR 200 an hour. A design team releases 200 hours a year on such a commission through assisted drafting, and the office has no unallocated work to move those people onto. What happens to annual contribution from that commission?

• A

  It falls by SAR 60,000, because the released hours are no longer billed

  Correct: 200 hours at the SAR 300 margin between fee and cost gives a fall of SAR 60,000.

• B

  It rises by SAR 40,000, because the cost of the released hours is removed

  This applies fixed-fee reasoning to a time-charged commission, and it is the commonest error in this calculation. Where the fee is agreed in advance, removing 200 hours at SAR 200 genuinely does gain SAR 40,000; here the fee moves with the hours, so the cost leaves but the fee leaves with it and more besides.

• C

  It rises by SAR 60,000, because the released hours become available capacity

  This computes the SAR 300 margin correctly and then books it in the wrong direction, valuing the hours as though they had already been resold. It would be right if the office had firm demand waiting for those 200 hours, which is exactly the condition the stem rules out.

• D

  It is unchanged, because the fee and the staff cost both fall by 200 hours

  This notices that both sides fall and concludes that they cancel, forgetting that they fall at different rates. Fee and cost would have to be equal for the commission to be margin-neutral, and a practice billing at SAR 500 against a cost of SAR 200 is not in that position.

In a Singaporean insurer's claims team, drafting the settlement letter accounts for 8% of the 15 hours a complex claim consumes. An assistant halves the drafting time. The team leader reports the department is well on the way to its 20% efficiency goal. What does the arithmetic show?

• A

  The task improves by 4%, so other steps must change as well

  Correct: halving a step worth 8% of the task returns 4% of the task, so the remaining steps have to supply the rest.

• B

  The task improves by 8%, so almost half the goal is secured

  This substitutes the size of the step for the size of the gain, and it is the single most common arithmetic slip in this domain. It would be right only if the assistant removed drafting altogether rather than halving it, and even then 8% is not almost half of 20%.

• C

  The gain will compound as the team's prompting skill improves

  This treats a structural ceiling as a skill problem and postpones the reckoning to a later quarter. Better prompting can push the saving towards the full 8% but cannot pass it, because the time is not in the task to be found.

• D

  The gain is real but shows up only once claim volumes recover

  This confuses total hours saved with the percentage improvement per claim. More claims would multiply the same 4% across more files; the ratio the department has been set to move does not change with volume at all.

At a Scottish social landlord, an assisted process removes 30 minutes a week from each of 20 housing officers and adds four hours a week in total to the two team leaders who countersign every case. The team leaders were already the reason cases waited. What has the change achieved?

• A

  A net gain in hours and a longer wait for tenants

  Correct: hours fall by six a week while the extra load lands on the constraint, so cases wait longer.

• B

  A net gain in hours and a shorter wait for tenants

  This assumes hours saved somewhere flow through into calendar time everywhere, which is the standard error when touch time and elapsed time are not distinguished. It would be right if the officers, rather than the team leaders, were the step at which cases queued.

• C

  A net loss in hours and a longer wait for tenants

  This spots the transfer of work onto the team leaders, which is the harder half of the question, and then omits to compare the two totals. The four hours added would have to exceed the ten hours released for the landlord to be worse off in hours, and they do not.

• D

  A net gain in hours and no change in the waiting

  This treats the redistribution as neutral for flow, as though four hours landing on a bottleneck were the same as four hours landing anywhere else. Waiting would be unchanged only if the team leaders had spare capacity, and the stem states the opposite.

A cost report issued by an Indian shared-service centre carries a wrong figure that originated in an assisted draft. The client asks who is accountable for it. What should the manager say?

• A

  Accountability rests with the analyst who accepted the draft figure.

  This confuses the internal allocation of fault with the external answer, and it comes readily to a manager who is trying to be candid. Naming an individual to a client offers up a colleague without changing anything the client is entitled to; it would be the answer only if the analyst, rather than the practice, had contracted to produce the report.

• B

  Accountability turns on whether the tool was disclosed in advance.

  This treats disclosure as though it transferred risk, in the way a stated assumption or caveat might. Disclosure changes what the client was told about the method, which is a genuine obligation where the terms require it, but no client agrees to accept wrong figures by being told that a tool was used.

• C

  The practice that signed the report answers for the figure.

  Correct: the practice put its name to the deliverable, so it answers to the client for the figures in it.

• D

  Accountability is shared with the supplier of the drafting service.

  This assumes responsibility travels with the work, so that whoever contributed to the error must share the answer. A licence can allocate cost or liability between the practice and its supplier, and that may well be worth pursuing, but the client has no engagement with the supplier and nothing in that contract reaches the duty owed on the report.

A Saudi consultancy's appointment contains a confidentiality clause, a key personnel clause and a subcontracting clause, and mentions no technology anywhere. A director argues that because the appointment is silent on assistants, nothing restricts sending the analysis through one. Why does that reasoning fail?

• A

  Silence in an appointment is read against whichever party drafted it.

  This reaches for a rule of construction, which sounds lawyerly and is beside the point. Reading ambiguity against the drafter resolves a term that genuinely admits two meanings, and there is no such term here; the subcontracting clause is clear, so nothing needs construing against anybody.

• B

  The client would have named the technology had it intended to permit it.

  This is the director's own error inverted: silence is read as prohibition instead of as permission, and both readings do the same thing, which is to reason from the absence of words. It would matter only if the appointment were otherwise silent on passing work outside the firm, and it is not.

• C

  The appointment is silent, so the firm's own internal policy governs it.

  This assumes the firm can fill a gap in the client's contract with its own rules. An internal policy can restrict what the firm does but cannot widen a permission the client controls, and a client reading that argument would hear the firm awarding itself a consent it was never given.

• D

  The subcontracting clause already covers passing analysis to an outsider.

  Correct: sending the analysis to an outside service falls squarely inside a clause the appointment already contains.

A UK housing association checks a 10 per cent sample of the rates in every valuation report. The sample size was settled years ago against surveyor error, which tends to cluster in one section of a report. Assistance is now used on rate build, and its errors are evenly spread and uniformly plausible. What follows for the review step?

• A

  The sample stays adequate, because the number of errors has not risen.

  This treats volume as the thing the sample was sized against, when the design in fact depended on distribution. It would be right if detection turned only on how many errors were present; a 10 per cent sample of evenly spread errors finds roughly 10 per cent of them, whereas the same sample against clustered error led the reviewer to a whole bad section.

• B

  The sample must grow, because the pattern it was sized against has changed.

  Correct: the sample was sized against clustered error, and uniformly spread error defeats that design, so coverage must rise to hold detection.

• C

  The review should pass to a second checker rather than to a larger sample.

  This buys independence or seniority when what is missing is coverage. A second checker examining the same 10 per cent repeats an inadequate sample more carefully, which would help if the failure were reviewers missing errors in front of them rather than errors sitting in the 90 per cent nobody opens.

• D

  The review should move to the end, once the report is fully assembled.

  This changes when the check happens rather than how much it covers, and later checking is generally worse, because errors cost more to correct once a report is assembled. Timing would be the issue if rates were being revised after review; nothing suggests that, and the sample would still miss the same proportion.