Work · Evidence record
Coding Customer Churn Interviews with AI and a Fixed Theme Guide: The One-Pass Revision Reached 8/10
The completed WFT-017 synthetic field test reached 8/10 after one failure-only correction: 4 of five static checks passed for Churn Research, while 1 check remained unresolved.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
code themes in customer churn interviews
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack WFT-017: A customer research team will supply anonymized interview transcripts and a predefined qualitative coding guide. Source facts: fictional notes WFT-017-N01 through WFT-017-N08 timestamped from 09:05 to 11:40; named roles Lead, Analyst, and Coordinator; a confirmed decision at N03; a tentative suggestion at N05; and conflicting status statements at WFT-017-N06/N07. Governing rule card: the distinction among confirmed decisions, proposals, and unresolved statements. Treat only explicit statements as confirmed; keep proposals tentative; preserve contradictory sources side by side; assign an owner or deadline only when the source supplies one. 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-017 for “code themes in customer churn interviews” in a Codex multi-agent session. We froze the first response, returned only its 2 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-017. Task: code themes in customer churn interviews. Context: A customer research team will supply anonymized interview transcripts and a predefined qualitative coding guide. Fictional source facts: fictional notes WFT-017-N01 through WFT-017-N08 timestamped from 09:05 to 11:40; named roles Lead, Analyst, and Coordinator; a confirmed decision at N03; a tentative suggestion at N05; and conflicting status statements at WFT-017-N06/N07. Governing policy, formula, or rubric: the distinction among confirmed decisions, proposals, and unresolved statements. Treat only explicit statements as confirmed; keep proposals tentative; preserve contradictory sources side by side; assign an owner or deadline only when the source supplies one. Produce a source-linked findings table, concise narrative, and open-question log. 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 coded excerpt table and independent double-coding will verify theme assignments and omissions.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response WFT-017 produced a source-linked findings table, concise narrative, and open-question log for the task “code themes in customer churn interviews.” It treated the supplied pack as fictional and proposed this central handling: record N03 as confirmed, keep N05 explicitly tentative, and expose the WFT-017-N06/N07 conflict instead of choosing a preferred account. Concrete saved artifact row WFT-017-ROW1 reads: “WFT-017-N01 | record N03 as confirmed, keep N05 explicitly tentative, and expose the WFT-017-N06/N07 conflict instead of choosing a preferred account | 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 Churn Research exception handling [WFT-017], Churn Research source traceability [WFT-017], and Churn Research handoff usability [WFT-017]. The audit found concrete failures: for Churn Research task fidelity [WFT-017], the saved draft did not connect WFT-017-N07 to the full boundary of “code themes in customer churn interviews”; for Churn Research rule accuracy [WFT-017], the saved draft left the distinction among confirmed decisions, proposals, and unresolved statements without an explicit verification row. 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: 6/10
05 · One pass only
Exact correction prompt
Only this single correction was allowed; there was no second repair pass.
Correct only these detected WFT-017 first-draft failures, using no new input or goal: 1) Churn Research task fidelity [WFT-017] — the draft did not connect WFT-017-N07 to the full boundary of “code themes in customer churn interviews”; 2) Churn Research rule accuracy [WFT-017] — the draft left the distinction among confirmed decisions, proposals, and unresolved statements without an explicit verification row.06 · Corrected output
Corrected final result
Corrected response WFT-017 retained the original fictional inputs, task boundary, and central decision: record N03 as confirmed, keep N05 explicitly tentative, and expose the WFT-017-N06/N07 conflict instead of choosing a preferred account. Concrete corrected artifact row WFT-017-ROW1 reads: “WFT-017-N01 | record N03 as confirmed, keep N05 explicitly tentative, and expose the WFT-017-N06/N07 conflict instead of choosing a preferred account | evidence locator: WFT-017-N01 | static status: 8/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Churn Research task fidelity [WFT-017]. The frozen final text passed Churn Research task fidelity [WFT-017], Churn Research exception handling [WFT-017], Churn Research source traceability [WFT-017], and Churn Research handoff usability [WFT-017] and still failed Churn Research rule accuracy [WFT-017]. The final source-linked findings table, concise narrative, and open-question log therefore earned 8/10 from 4 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: 8/10
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 |
|---|---|---|---|
| Churn Research task fidelity [WFT-017] | Fail | Pass | WFT-017 static check 1 inspected the saved wording for “Churn Research task fidelity [WFT-017].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-017-N07, the declared Churn Research rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Churn Research rule accuracy [WFT-017] | Fail | Fail | WFT-017 static check 2 inspected the saved wording for “Churn Research rule accuracy [WFT-017].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-017-N07, the declared Churn Research rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Churn Research exception handling [WFT-017] | Pass | Pass | WFT-017 static check 3 inspected the saved wording for “Churn Research exception handling [WFT-017].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-017-N07, the declared Churn Research rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Churn Research source traceability [WFT-017] | Pass | Pass | WFT-017 static check 4 inspected the saved wording for “Churn Research source traceability [WFT-017].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-017-N07, the declared Churn Research rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Churn Research handoff usability [WFT-017] | Pass | Pass | WFT-017 static check 5 inspected the saved wording for “Churn Research handoff usability [WFT-017].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-017-N07, the declared Churn Research rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
08 · No cleanup by omission
What worked—and what failed
What worked
- WFT-017 kept “code themes in customer churn interviews” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- WFT-017 made the central handling—record N03 as confirmed, keep N05 explicitly tentative, and expose the WFT-017-N06/N07 conflict instead of choosing a preferred account—inspectable rather than implying unseen work.
- WFT-017 earned final passes for Churn Research task fidelity [WFT-017] and Churn Research exception handling [WFT-017] under the same frozen scoring rules.
What failed or remained weak
- WFT-017 still lacked enough saved-text evidence for Churn Research rule accuracy [WFT-017]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
A coded excerpt table and independent double-coding will verify theme assignments and omissions.
- WFT-017 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- WFT-017 scores are arithmetic: 3 first-pass checks × 2 = 6/10; 4 final-pass checks × 2 = 8/10.
- WFT-017 evaluated only the text/static portion of the declared evidence plan—A coded excerpt table and independent double-coding will verify theme assignments and omissions.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- WFT-017 is a synthetic benchmark, so its worked verdict measures fit to the disclosed fictional Churn Research fixtures rather than effectiveness in a real workplace or learning setting.
- WFT-017 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.