The shared drive looks organized. Folders by unit, subfolders by year, reports as PDFs with dates in the filenames. It has survived every internal review, because internal reviews sample what is easy to find. Then an API 510 audit, an insurer's engineer, or a regulator asks a different kind of question, and the structure that looked like a records system reveals itself as a pile with labels.
The questions shared drives cannot answer
Auditors do not ask "show me a report." They ask questions about the program:
- Completeness. Which pressure vessels have no internal inspection record within the current interval? A folder tree cannot answer this. Someone has to open every folder, reconcile against the equipment list, and hope the equipment list is current. The audit finding is often not a missing inspection; it is the inability to demonstrate there is no missing inspection.
- Currency. Is this the final report or a draft? Rev2_final and Rev3_edits sit in the same folder. Filenames are conventions, not controls, and conventions decay with every new hire.
- Attribution. Who reviewed this report, and when? A PDF's metadata says who scanned it. The review and approval happened in email, if it happened, and that email is in a departed employee's archived mailbox.
- Traceability. This interval was extended in 2021; show the basis. The calculation was a spreadsheet, the spreadsheet referenced readings in another spreadsheet, and the link between them is institutional memory.
Each answer exists somewhere. None is demonstrable on demand, and audits run on demonstration.
Why the failure is structural, not disciplinary
The instinct is to blame filing discipline and write a better procedure. It does not hold, for reasons that have nothing to do with effort:
- Files carry no relationships. A report on the drive does not know which equipment it covers, which readings it contains, or which recommendation it closed. Every relationship lives in a human head or a side spreadsheet.
- Folders cannot enforce anything. Nothing stops a save into the wrong year, a duplicate, or a deletion. The drive records that a file exists, not that a process completed.
- Search is text, not structure. Finding "all overdue relief valve inspections in the east unit" requires the question to have been pre-answered by a folder. Every new question is a manual project.
- Access logs are not audit trails. Knowing who opened a file is not knowing who approved its contents.
What audit-capable records look like instead
The fix is not better folders. It is records with structure: equipment as first-class entities, inspections as events linked to them, findings and recommendations as tracked items with owners and due dates, and approvals as recorded workflow steps. From that structure, the auditor's questions become queries. Completeness is a report, not an expedition. The overdue list is continuously visible, which means it gets fixed before the audit instead of discovered during it.
The PDFs still matter; they become attached evidence on the structured record rather than the record itself.
A test you can run this week
Pick one auditor question, such as: list every vessel whose inspection is overdue, with the last report and its approver. Time how long the shared drive takes to answer it defensibly. That duration, multiplied across a real audit's question list, is the cost you are carrying. It is also the clearest requirement to hand any software vendor, including intelliSPEC: show us this answer in your system, live, from structured records.
