Two different jobs
Mechanical means a careful reader could point at text (or its absence) and agree something is inconsistent, incomplete, or circular — without deciding whether the product is safe or the investigation is scientifically sound.
Judgment means interpreting evidence, scope, risk, and adequacy in context: was this the right method, the right batch impact conclusion, the right severity? Two competent reviewers can disagree, and both can be acting in good faith.
Usually mechanical
- Approval dated before the event it signs
- Missing version / revision / effective date
- Root cause that only restates the symptom
- Undefined abbreviations used as if defined
- Internal contradictions (limit A vs limit B in the same doc)
- Empty signature block or blank required field the template expects
- Cross-reference to a revision that does not match the attachment
Needs judgment
- Whether the root cause is true, not merely non-circular
- Whether impact assessment covered the right batches
- Whether the CAPA is proportionate to risk
- Whether data support a shelf-life or release conclusion
- Whether a deviation should have been escalated
- Whether a signature is authentic (not merely present)
Why the split matters under deadline
Inspection-readiness work is scarce attention. If a human spends the first pass hunting date order and missing identifiers, they have less left for the judgment calls an inspector will actually test in the room.
Mechanical triage is not a shortcut around science. It is a way to stop the avoidable stuff from consuming the budget of care.
What a triage tool should do
- Flag clear, specific problems with evidence from the document.
- Say REVIEW when the document does not contain enough to judge — not invent a pass.
- Never issue an overall score or “ready to submit” label. That is a clearance claim dressed as metrics.
- Never generate regulatory citations on the fly. Sources belong on fixed criteria written by a human.
PASS on a criterion is not “document approved.” It only means no issue was found against that one rule on this pass. A full PASS list is still a work list of zeros — not a certificate.
Where people get burned
- Treating mechanical silence as scientific adequacy
- Softening REVIEW into PASS so the report looks tidy
- Asking software to “verify” a wet-ink signature from a scan
- Collapsing twenty criteria into one green readiness percentage for management
If someone skips their own review because a tool looked clean, and the tool was wrong, the tool was worse than useless.
Browse the criteria → · Example: circular root causes →
See per-criterion findings on sample data — labeled simulated, no login.