The decision this example supports
For a fund administrator or investment operations leader, the question is whether draft preparation can reduce useful work without increasing checking or obscuring exceptions. This is a hand-authored expected output, not an AI execution or client result. All inputs and quantities are invented solely to test the workflow.
The inputs
SRC-R1, previous-month approved operating report: 100 open service requests at month end. SRC-R2, current-month approved register: 110 open service requests. SRC-R3, current-month commentary file: not received. SRC-R4, earlier working register: 107 requests, marked superseded.
These are service-request counts, not investment performance figures. The permitted task is to prepare a draft operating update using approved sources and make unresolved information visible.
The expected draft
Open service requests increased from 100 to 110, a rise of 10 requests or 10%, using the approved month-end sources [SRC-R1, SRC-R2]. The current-month commentary has not been received [SRC-R3]. A reason for the movement cannot yet be stated.
Reviewer checklist: confirm matching cutoff dates and definitions, reproduce the arithmetic, check that SRC-R4 was excluded, obtain or waive the missing commentary through the approved process, and approve wording before circulation.
A claim such as increased client demand caused the movement would fail this test: none of the supplied sources establishes that explanation.
Where a plausible draft can fail
If the previous value is zero, percentage change should be marked undefined rather than fabricated. If reporting periods differ, flag the mismatch before comparison. If an approved source cannot be accessed, say which part cannot be completed.
A valid-looking number from the wrong version remains an error. Test version selection separately from arithmetic and separately from narrative claims.
How to judge the result
Measure the entire task through approval, including source retrieval, drafting, review and corrections. Record unexplained claims and version errors even when someone catches them before circulation.
A faster first draft only matters if the completed report meets the agreed standard with a worthwhile overall change in effort. No saving is established by this illustrative example.
Save this illustrative pack
The text download contains the sample inputs, expected output and exception checks described above. It contains no client information.