Fisheries gives me a concrete case. The Florida Fish and Wildlife Conservation Commission Public Fisheries Data Module supplies 3,335 trip-ticket rows for 2000–2024. The Federal Reserve Bank of St. Louis FRED series CPIAUCSL supplies 300 monthly, seasonally adjusted CPI-U observations. The workflow preserves those sources, the calculation path, exceptions, and an inspectable FTI workbook. The practitioner can still reject it when the code runs. A reviewer did exactly that when the time control was unclear, years appeared on the wrong axis, and county results needed a map. Those objections became requirements.

A change reopens the boundary instead of inheriting the last approval. New FWC PFDM data or a revised CPIAUCSL vintage enters as a candidate, never as a replacement for the approved snapshot. Schema, period, duplicate, and reconciliation checks run before promotion; a failure quarantines the candidate and names every failed gate. A reviewer’s CPI catch became a fail-closed contract rather than a one-time correction. The public demonstration deliberately disables promotion. A firm pilot would require an authenticated practitioner to approve the candidate, preserve the prior snapshot, and retain a rollback path.

The evidence is mixed, which is useful. Over ten weeks, Project Clifford used LDR Hub to coordinate three interns across A-01 market concentration, A-03 fleet age, A-04 backlog, A-05 price index, and A-99 overcharge regression. They produced 28 report exhibits, and each intern defended an individual method without me answering for them. A separate exit review found that other tools still depended on the original builder. Fisheries has completed a reviewer correction loop, but not a full independent handoff. I would not call it organizational capability until a second analyst can reproduce the output, defend the method, adapt when inputs change, and identify the decisions that remain human.