Definition
The rate at which AI-supported inspection deferrals extend actual inspection intervals beyond their engineering basis, measured as cumulative interval extension per asset class per year with the fraction lacking a revalidated engineering justification.
The Measurement Gap
There is no metric that tells you the cumulative effect of individually justified AI-supported inspection deferrals, which means inspection intervals can erode one defensible decision at a time until the actual interval bears no relationship to its engineering basis.
Formula
Plain language: For each asset class, compare the actual realized inspection intervals against the baseline intervals in the RBI or code-required basis. Sum the extension attributable to AI-supported deferral decisions. Divide by the baseline to get the erosion fraction, and separately report what share of the cumulative extension has been revalidated by an engineering assessment of the current interval (not just the individual deferral steps).
Notation:
IIER_c = (I_actual_c - I_basis_c) / I_basis_c × 100, with revalidated share RV_c = E_revalidated_c / E_total_c
Variables:
- I_actual_c = mean realized inspection interval for asset class c over the trailing period (days, NATIVE from inspection records)
- I_basis_c = engineering-basis interval for class c from RBI/code (days, MANUAL)
- E_revalidated_c = cumulative extension days covered by a revalidation of the current total interval (days, MANUAL from engineering records)
- E_total_c = total cumulative extension days attributable to AI-supported deferrals (days, DERIVED from deferral decision log)
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Each individual AI-justified deferral looked reasonable. Added together, how far have our real inspection intervals drifted from the RBI basis, and would the original engineering still sign off on where we've landed?