VOLOVLOAutomation risk & transition, task by task
AskOccupationsMajorsBusinessFoundersChangesNotesMethod
Search occupations, majors…
EN
  • English
  • 简体中文
  • 日本語
  • Español
  • Português
  • Français
VLO
VOLO

Understanding how automation changes work — task by task, with the evidence shown and the uncertainty admitted.

AskOccupationsMajorsBusinessFoundersChangesNotesMethodWill AI replace my job?AboutRole diagnosisOpen dataOntologyVOLO ProPrivacyTerms
© 2026 VOLO
Guides
Ask VOLOFor businessFor foundersRecent changesNotesRole diagnosisFirst AI experimentVOLO ProMethod & evidenceAboutFollow an occupationSearch
Occupations
All occupations
AI / software
Translator / InterpreterBank tellerCopywriterContent moderatorCustomer service representativeAdministrative assistantSoftware tester / QA engineerData entry clerkGraphic designerParalegalVideo editorAccountant / BookkeeperMarketing specialistFrontend developerTax preparer / tax agentVoice actorMedical coderData analystInsurance claims handlerTechnical writer / documentation engineerJunior software developerInsurance underwriterHR / recruiterLoan officer / credit officerFinancial analystProcurement / supply chain specialistJournalistSales / account managerReal estate agentNetwork engineerIllustratorIT support specialist / helpdeskAuditorSupply chain plannerManagement consultantPhotographerBackend developerAI researcherProduct / UX designerBusiness systems ownerE-commerce operations specialistActuaryAnimator and VFX artistWriter and authorRadiologistData engineerProject managerData scientistCourt reporterEditor and proofreaderLawyerMedical assistant / clinic assistantQuantity surveyorContent creatorQuantitative analystMachine learning engineerExperienced software engineerDevOps / platform / SRE engineerBusiness analystMusician and composerProduct managerPharmacistPartnerships / channel managerSecurity analyst (SOC)Financial adviser / financial plannerInterpreterCompliance officerLibrarianArchitectFirst-line manager / team supervisorCounsellor / therapistRetail salesperson / shop assistantCivil / structural engineerElectrical engineerIndustrial engineerSecurity guardInterior designerUniversity lecturerBus driverInsurance agentMechanical engineerSchool teacherGeneral practitioner / primary care doctorAirline pilotWaiter / restaurant serverRadiographer / radiologic technologistAuto mechanic / vehicle technicianPolice officerSonographerAircraft maintenance technicianAir traffic controllerVeterinarianPhysiotherapist / rehabilitation therapistDentistSurgeonConstruction workerSocial workerRegistered nurseCare worker / nursing assistantPlumberAI implementation lead
RPA / self-service
Government service clerkOperations coordinatorMetro train driverReceptionist / front deskPharmacy technician
Robotics
Retail cashier / shop assistantContainer port workerWarehouse workerAssembly line workerMedical laboratory technicianActor and modelWelderChef / cookCleaner / janitorFarmerFirefighterCabin crewElectrician
Autonomous driving
Ride-hail / taxi driverTruck driverDelivery rider / courier
Majors
All majorsEnglish / Foreign languagesComputer scienceAccountingPsychologyJournalism / CommunicationFinanceLawVisual communication designMarketingNursingBusiness administrationEducation and teacher trainingArchitecturePublic administrationMedicineHospitality and tourism managementEconomicsInformation systems
Enter as:I have a jobI am studyingI run a companyI am building something
Recent changes›Data entry clerk›2026-02-12
CapabilityCognitive automation2026-02-12

Frontier models produced 0% valid output on a 369-field financial reporting schema, and one domain passed only 12.5% despite 90% valid output, ExtractBench's authors reported

Data entry clerkoccupation page →
Event date / reported
2026-02-12
Evidence stage
CapabilityA demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
Tasks this bears on
Verifying and correcting records
Checking entered data against the source and fixing errors.
Being augmented✓ Evidence-backed
Where this applies
A benchmark of extracting structured data from 35 PDF documents with human-annotated gold labels, 12,867 fields in all. The authors report that frontier models remain unreliable on realistic schemas, that performance degrades sharply with schema breadth, culminating in 0% valid output on a 369-field financial reporting schema across all tested models, and that on one domain models achieve 90% valid output but only a 12.5% pass rate. The authors work at a company that sells document products; it is a benchmark, not a measure of practice.
What this means
Machine extraction breaks down on long, complex forms, and valid-looking output can still be wrong — which is why checking survives automation of keying.
What it does not yet show
A benchmark by a company with a stake; models improve, and it does not measure how much checking is done in practice.
What you can check
Open the ExtractBench paper on arXiv (2602.12247) and find "0% valid output".
Does it change the assessment?
No — and this stage does not move it either. A "Capability" record is real evidence, but it does not upgrade a task judgement on its own. The 1 linked judgement above stand where they were.
Source
Ferguson, Pennington, Beghian, Mohan, Kiela, Agrawal and Nguyen (Contextual AI) — ExtractBench: A Benchmark and Evaluation Methodology for Complex Structured Extraction, arXiv:2602.12247 (submitted 12 Feb 2026) · verified 2026-09-30 · Claude (VOLO agent) · interpreted 2026-09-30 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.
All changes for Data entry clerk →All recent changes →How events become evidence →