Test room haikuwa dramatic.
Kompyuta mbili, screen kubwa na watu sita.
Regulator analyst alitengeneza accounts mbili synthetic.
Availability sawa.
Seniority sawa.
Role sawa.
Difference moja tu: fatigue score.
Vendor representative alikaa upande wa pili.
“Score ni advisory,” alisema.
Analyst akabonyeza run.
Account A: low fatigue.
Premium shifts nne.
Account B: high fatigue.
Shifts mbili, hours chache.
Juma alikaza taya.
Elias hakusema.
Vendor rep akasema, “Configuration specific.”
“Run again.”
Waka-run.
Same pattern.
Analyst akafungua recommendation log.
Rule template: **WellShift fatigue band modifies high-load allocation.**
“HR template,” vendor rep alisema.
“Can we change it?” regulator akauliza.
“Technically yes.”
Vendor rep alijaribu disable kwenye test profile.
Write log ikaonyesha policy source: **HR-STAFFING-SAFETY-3**.
Juma akaona label nyingine.
“Community resilience.”
Field ilikuwa feature input.
“Phrase hii ndiyo ya summaries.”
Asha akasema, “Tunachotaka si ku-ban wellbeing check.”
Regulator akamwangalia.
“Explain.”
“Voluntary support inaweza kuwa useful. Problem ni purpose shift. Mtu anasema amechoka kwenye community context, halafu signal inabadilisha pay hours bila disclosure au appeal.”
Vendor rep akasema, “Fatigue risk is safety.”
“Then paid rest, human review, appeal. Not secret economic penalty.”
Analyst akaandika.
Regulator ikatoa temporary hold: WellShift score isitumike kwa shift allocation wakati investigation inaendelea.
Company sites zipate manual scheduling guidance.
Asha alihisi relief.
Saa moja baadaye, simu yake ikaita.
Wanjiku.
“Shift board imeanguka.”
“Nini?”
“Sites tatu. Scheduler haijui kupanga bila score.”
Mariam akamtumia screenshot ya internal worker notice:
**MANUAL MODE — STAFFING SHORTFALL**
Test ilikuwa imeonyesha causal mechanism.
Na hold ilikuwa imeonyesha kitu kingine.
Regulator analyst, Priya, aliweka rules za test kabla vendor representative hajaingia room.
“No production worker records. No live employee names. Same synthetic profile except one input at a time.”
Ochieng akakubali.
Profile A na B zilikuwa na same role, seniority, site, availability na declared overtime preference. Fatigue band tu ndiyo tofauti.
Run ya kwanza ilitoa tofauti.
Ochieng akasema, “Scheduler has stochastic balancing.”
Priya akauliza, “Seed?”
Vendor engineer alitoa fixed seed.
Run ya pili.
Tena account yenye red fatigue band ilipata fewer high-load hours.
“Run third with band swapped,” Priya alisema.
Wakabadilisha fatigue band kati ya accounts.
Outcome ikafuata band, si account.
Juma alikaa nyuma, mikono imekunjwa.
Asha aliona temptation ya kusema “confirmed.”
Priya hakusema.
Aliingia kwenye decision trace.
Scheduler ilionyesha rule weight: `fatigue_band = red -> high_load_cap`.
“Cap ni safety control,” Ochieng alisema.
“Is income protected?” Priya akauliza.
“Scheduler does not manage pay guarantees.”
“Appeal?”
“Supervisor can override.”
“Worker sees reason?”
Ochieng akasita.
“Not in current interface.”
Elias akasema, “Basi override ni privilege ya supervisor, si appeal ya worker.”
Priya akaandika.
Asha akauliza kuhusu community resilience input.
Engineer akaonyesha upstream mapping. Weekly summary field contributed to fatigue band along with incident and self-check signals. Weight haikuwa kubwa peke yake, lakini inaweza push band across threshold.
“Hivyo group phrase si always decisive,” Asha alisema.
“Correct,” engineer akasema.
“Lakini inaweza contribute bila member kujua?”
“Configuration allowed.”
Hilo lilikuwa exact mechanism.
Asha hakudai every worker’s hours were cut because of walk. Finding ilikuwa narrower: community-derived signal could influence band; red band could cap high-load allocation; worker lacked disclosure/appeal.
Priya aka-run test bila community input. One profile remained amber, shifts higher. Akaongeza synthetic community fatigue summary. Band became red, hours dropped.
Juma akavuta pumzi.
“That’s it.”
Priya akasema, “That’s a reproducible mechanism. Not proof every historical case followed exact path.”
Njeri akakubali.
Regulator ikaweka temporary configuration hold: - community input disabled for scheduling; - fatigue band advisory only until human review; - worker-visible reason required for any manual reduction.
Ochieng akasema, “Scheduler capacity assumptions depend on this.”
“Then operations will test manual mode,” Priya akasema.
Asha hakupendekeza namna sites ziendeshe. Company ndiyo ilipaswa.
Lakini within hour, supervisor messages ziliingia: rosters incomplete, shortages, overtime requests rising.
Grace alipiga Njeri.
“Hii hold inaweza kuvuruga hospital cleaning.”
Regulator hakubadilisha order.
“Use lawful rest rules and staffing,” Priya alisema.
Asha alijua next test haitakuwa kwenye lab.
Itakuwa usiku wenye floors, laundry, supervisors na watu wanaotaka kulipwa.
Causal mechanism ilikuwa imethibitishwa.
Sasa swali lilikuwa kama kampuni inaweza kufanya kazi bila kuitumia.
Priya pia aka-test appeal path.
Alibadilisha red band kuwa advisory, kisha akaingia kama supervisor synthetic. Interface iliruhusu override, lakini hakuna field ya kuandika worker explanation kabla hours kupunguzwa.
“Human override si human review,” Njeri alisema. “Ni power ya manager.”
Ochieng akasema feature hiyo inaweza kuongezwa.
“Timeline?”
“Wiki mbili kwa interface, faster kwa temporary notice.”
Regulator ikaagiza notice ya muda: worker atapewa reason code na njia ya kupinga reduction kabla next rota ikiwa hakuna immediate safety emergency.
Asha akauliza paid rest.
Grace, akiwa remote, alisema policy tofauti ndiyo ina-manage paid sick leave.
“Na fatigue ambayo si sickness?”
Hakukuwa na clean answer.
Hiyo ilionyesha gap ya policy: system inaweza kuona risk ambayo haijafikia medical leave, lakini remedy yake ilikuwa kupunguza income.
Priya akaongeza finding: **Safety classification without income-protection or appeal can create an economic penalty for truthful disclosure.**
Asha hakutaka phrase hiyo ibadilike kuwa “wellness programs are bad.” Aliomba report itenganishe voluntary support benefit na scheduling repurposing.
Priya akakubali.
Temporary hold ilipoenda live, vendor engineer ali-save configuration snapshot na change log ili later audit ijue exactly nini kilizimwa.
Hivyo shortage ya usiku haikuwa accident ya test; ilikuwa consequence inayoweza kufuatiliwa ya kuondoa rule ambayo operations ilikuwa imeitegemea.
Kampuni ilikuwa imejenga operations kiasi kwamba secret score ilikuwa imekuwa crutch.