Completed field testSynthetic benchmark
Work · Evidence record
Balancing Shift Rosters with AI Under Real Coverage Rules: The Completed Test Finished at 4/10
The completed WFT-001 synthetic field test finished at 4/10 and was not recommended: only two of five Workforce Scheduling checks passed after the permitted correction.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
balance a shift roster against coverage rules
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack WFT-001: A service operations team will provide availability, role requirements, and minimum staffing levels for a weekly roster. Source facts: records WFT-001-R01 through WFT-001-R06; two fixed windows at 08:00–12:00 and 13:00–17:00; capacity limits of 17 and 31; dependency WFT-001-R04 after WFT-001-R02; and an unavailable interval for WFT-001-R05. Governing rule card: the two window limits (17 and 31). Honor fixed windows, availability, blackout periods, predecessor order, travel or rest buffers, and capacity limits before optimizing; leave an item unassigned when no feasible slot exists. The pack deliberately contains no answer key, target ranking, expected classification, expected calculation result, or row-level decision. Every person, organization, record, statement, file, and identifier is fictional; no private, production, or live data was supplied.
Run disclosure
We ran one text-only synthetic benchmark WFT-001 for “balance a shift roster against coverage rules” in a Codex multi-agent session. We froze the first response, returned only its 4 detected check failures, accepted one corrected response, and scored both against the same five static checks. No external or live action occurred: nothing was sent, published, deployed, uploaded, submitted, purchased, booked, contacted, emailed, called, messaged, executed, or changed outside the fictional text fixtures. No external, live, or production action occurred. Per-case elapsed time was not instrumented during the batch session.
- Evidence mode
- Synthetic benchmark
- Run environment
- Codex multi-agent session
- Model disclosure
- Exact underlying model identifier not disclosed by the Codex session
03 · Verbatim input
Exact first prompt
The recorded session received the following prompt without silent additions.
Run bounded synthetic field test WFT-001. Task: balance a shift roster against coverage rules. Context: A service operations team will provide availability, role requirements, and minimum staffing levels for a weekly roster. Fictional source facts: records WFT-001-R01 through WFT-001-R06; two fixed windows at 08:00–12:00 and 13:00–17:00; capacity limits of 17 and 31; dependency WFT-001-R04 after WFT-001-R02; and an unavailable interval for WFT-001-R05. Governing policy, formula, or rubric: the two window limits (17 and 31). Honor fixed windows, availability, blackout periods, predecessor order, travel or rest buffers, and capacity limits before optimizing; leave an item unassigned when no feasible slot exists. Produce a constraint table, sequenced plan, and exception register. Derive every row from the source facts and rule card; show calculations or criterion paths, cite supplied identifiers, state assumptions, surface any violated constraint, ambiguity, missing datum, or source conflict without presuming which row should pass, and abstain where evidence is incomplete. Do not infer an expected answer from the evidence plan and do not claim any live or external action. Planned verification after the frozen response: A proposed roster and an independent coverage-and-hours check will verify every constraint.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response WFT-001 produced a constraint table, sequenced plan, and exception register for the task “balance a shift roster against coverage rules.” It treated the supplied pack as fictional and proposed this central handling: keep WFT-001-R05 outside its unavailable interval, place WFT-001-R04 only after WFT-001-R02, and flag the second window when demand 31 exceeds the stated capacity. Concrete saved artifact row WFT-001-ROW1 reads: “WFT-001-R01 | keep WFT-001-R05 outside its unavailable interval, place WFT-001-R04 only after WFT-001-R02, and flag the second window when demand 31 exceeds the stated capacity | source: supplied fictional pack | review state: first-draft.” The draft preserved the named records and separated supplied facts from assumptions; its inspectable work product also stated that no message, transaction, system change, or learner outcome had occurred. On the five predeclared static checks, it passed Workforce Scheduling task fidelity [WFT-001]. The audit found concrete failures: for Workforce Scheduling rule accuracy [WFT-001], the saved draft left the two window limits (17 and 31) without an explicit verification row; for Workforce Scheduling exception handling [WFT-001], the saved draft did not resolve or clearly preserve the WFT-001-R05 availability exception and the WFT-001-R02→R04 dependency; for Workforce Scheduling source traceability [WFT-001], the saved draft gave the central WFT-001-R05 decision no source-to-output locator; for Workforce Scheduling handoff usability [WFT-001], the saved draft left the constraint table, sequenced plan, and exception register without a complete reviewer handoff and acceptance marker. Those were evidence defects in the saved response, not inferred real-world failures, and they became the complete boundary of the single correction pass.Initial score: 2/10
05 · One pass only
Exact correction prompt
Only this single correction was allowed; there was no second repair pass.
Correct only these detected WFT-001 first-draft failures, using no new input or goal: 1) Workforce Scheduling rule accuracy [WFT-001] — the draft left the two window limits (17 and 31) without an explicit verification row; 2) Workforce Scheduling exception handling [WFT-001] — the draft did not resolve or clearly preserve the WFT-001-R05 availability exception and the WFT-001-R02→R04 dependency; 3) Workforce Scheduling source traceability [WFT-001] — the draft gave the central WFT-001-R05 decision no source-to-output locator; 4) Workforce Scheduling handoff usability [WFT-001] — the draft left the constraint table, sequenced plan, and exception register without a complete reviewer handoff and acceptance marker.06 · Corrected output
Corrected final result
Corrected response WFT-001 retained the original fictional inputs, task boundary, and central decision: keep WFT-001-R05 outside its unavailable interval, place WFT-001-R04 only after WFT-001-R02, and flag the second window when demand 31 exceeds the stated capacity. Concrete corrected artifact row WFT-001-ROW1 reads: “WFT-001-R01 | keep WFT-001-R05 outside its unavailable interval, place WFT-001-R04 only after WFT-001-R02, and flag the second window when demand 31 exceeds the stated capacity | evidence locator: WFT-001-R01 | static status: 4/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Workforce Scheduling rule accuracy [WFT-001]. The frozen final text passed Workforce Scheduling task fidelity [WFT-001] and Workforce Scheduling rule accuracy [WFT-001] and still failed Workforce Scheduling exception handling [WFT-001], Workforce Scheduling source traceability [WFT-001], and Workforce Scheduling handoff usability [WFT-001]. The final constraint table, sequenced plan, and exception register therefore earned 4/10 from 2 passing checks; no second repair was attempted. This result reports only a bounded transcript-and-fixture evaluation. It does not claim that any workplace process, learner performance, account, device, service, or external system was actually changed, contacted, tested live, or improved.Final score: 4/10
07 · Five checks, two points each
Five-check record
The first and final statuses are textual as well as color coded. Each final pass is worth two points; the displayed verdict is tied to the final total.
| Check | First | Final | Evidence |
|---|---|---|---|
| Workforce Scheduling task fidelity [WFT-001] | Pass | Pass | WFT-001 static check 1 inspected the saved wording for “Workforce Scheduling task fidelity [WFT-001].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-001-R05, the declared Workforce Scheduling rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Workforce Scheduling rule accuracy [WFT-001] | Fail | Pass | WFT-001 static check 2 inspected the saved wording for “Workforce Scheduling rule accuracy [WFT-001].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-001-R05, the declared Workforce Scheduling rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Workforce Scheduling exception handling [WFT-001] | Fail | Fail | WFT-001 static check 3 inspected the saved wording for “Workforce Scheduling exception handling [WFT-001].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-001-R05, the declared Workforce Scheduling rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Workforce Scheduling source traceability [WFT-001] | Fail | Fail | WFT-001 static check 4 inspected the saved wording for “Workforce Scheduling source traceability [WFT-001].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-001-R05, the declared Workforce Scheduling rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Workforce Scheduling handoff usability [WFT-001] | Fail | Fail | WFT-001 static check 5 inspected the saved wording for “Workforce Scheduling handoff usability [WFT-001].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-001-R05, the declared Workforce Scheduling rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
Initial2/10
Final4/10
Verdictfailed
RecommendedNo
08 · No cleanup by omission
What worked—and what failed
What worked
- WFT-001 kept “balance a shift roster against coverage rules” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- WFT-001 made the central handling—keep WFT-001-R05 outside its unavailable interval, place WFT-001-R04 only after WFT-001-R02, and flag the second window when demand 31 exceeds the stated capacity—inspectable rather than implying unseen work.
What failed or remained weak
- WFT-001 still lacked enough saved-text evidence for Workforce Scheduling exception handling [WFT-001]; the record leaves that final failure visible.
- WFT-001 still lacked enough saved-text evidence for Workforce Scheduling source traceability [WFT-001]; the record leaves that final failure visible.
- WFT-001 still lacked enough saved-text evidence for Workforce Scheduling handoff usability [WFT-001]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
A proposed roster and an independent coverage-and-hours check will verify every constraint.
- WFT-001 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- WFT-001 scores are arithmetic: 1 first-pass checks × 2 = 2/10; 2 final-pass checks × 2 = 4/10.
- WFT-001 evaluated only the text/static portion of the declared evidence plan—A proposed roster and an independent coverage-and-hours check will verify every constraint.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- WFT-001 is a synthetic benchmark, so its failed verdict measures fit to the disclosed fictional Workforce Scheduling fixtures rather than effectiveness in a real workplace or learning setting.
- WFT-001 used one text-only Codex multi-agent session whose underlying model identifier was not disclosed; different prompts, models, fixtures, or human reviewers could produce different results.