Anonymized field-feedback report

Publication status

• Report ID and version:
• Review status: draft / reviewed / withdrawn
• Organization type and sector:
• Country or regulatory context at a non-identifying level:
• Observation period:
• Independent reviewer:
• Publication date:

Evidence boundary

• Direct observation / internal measurement / interview / survey / vendor report:
• Who collected the evidence and for what purpose:
• System and workflow version:
• Baseline, denominator, sample, and missing cases:
• Raw evidence location and authorized reviewers:
• Personal, confidential, security-sensitive, or identifying details removed:

Preregistered transferred hypothesis

• Public source IDs or local hypothesis:
• Frozen snapshot version, UTC timestamp, and task-time registry version:
• Pre-pilot status: transferred measurement / adjusted hypothesis / local hypothesis:
• Transfer contract: task profile, work mode, architecture, exact A0–A4 action boundary, quality threshold, and experience level:
• Comparability and required adjustments:
• Planned manual minutes per case, cases per month, and eligible share:
• Exact net low, central, and high ranges per eligible case and across the whole workload:
• Retained human work: preparation, supervision, verification, correction, exception rate, and minutes per exception:
• Setup hours, amortization period, and calculated minutes per eligible case:

What happened

• Original process and problem:
• Work mode, architecture, exact A0–A4 action boundary, and permissions:
• Pilot protocol and preregistered thresholds:
• Observed business outcome:
• Total baseline human minutes for every observed case:
• Total human minutes with AI for every observed case, including rejected, failed, excluded, escalated, corrected, fallback, and missing cases:
• Observed whole-workload result and position below / within / above the range:
• Explanation for the gap between hypothesis and observation:
• Hypothesis to retain, narrow, widen, or remove:
• Recalibrated snapshot as a separate version, or reason none was created:
• Reliability by critical segment:
• Incidents, near misses, corrections, and withdrawals:
• Decision taken and by whom:

Transfer limits

• Population and workflows represented:
• Contexts not represented:
• Dependencies and implementation effort:
• Conditions required to reproduce the result:
• Claims this report does not support:
• Earliest reassessment date:

Publication review

• No personal data or re-identification key
• No client, supplier, or employee identity without explicit authority
• No secrets, exploitable security details, or privileged material
• Observation is distinguished from estimate and opinion
• Hypothesis and range were frozen before field results were read
• Extrapolation, observation, and recalibration retain distinct evidence statuses
• Failures and excluded cases are not removed from the denominator
• Supplier claims are labelled as supplier claims
• Transfer limits are prominent
• Publication authority and withdrawal contact are recorded

Short finding suitable for the public index: