Exception Controls & Pilot Governance

[AI] Retain the approved control rationale. Release each expansion decision only when the population, executed permission and physical outcome can be reconciled.

Context

[HR] A distributed service network used bulk virtual status changes alongside normal physical processing. The register describes routine use, incorrect responsibility coding and competing risks: permissive access could mask unresolved work; restrictive access could delay legitimate exceptions. A technical dependency change created additional pressure on the operating model.

Objective

[PD] Establish whether tiered permissions can prevent unauthorized changes while preserving timely, legitimate recovery. The investment question concerns net mature service and resource consequences, rather than a reduction in submitted requests alone.

Approach

[RC] Reconcile the source register, permission versions, pilot membership, rate arithmetic, concentration measures and planning claims. [PD] The Permission-and-Exposure Reconciliation Contract (PERC) links a versioned permission to its executed population, fallback service and independently measured outcome. A separate finite operating model tests candidate evaluation designs.

Findings

[RC] The reported total lies outside the interval defined by the subgroup rates under the stated common-basis interpretation. [AI] This blocks the claimed aggregate explanation before causal attribution is considered. A limited permission pilot also cannot validate untested approval and central-service components. Recorded handover, approval and implementation stages are distinct.

Decision Value

[PD] The sponsor should require the metric owner to reconcile identities and rate definitions, the control owner to demonstrate rules on every route, and the service owner to accept overflow recovery. Release requires positive net resource value, improved physical outcomes and a prespecified legitimate-service safeguard. Unresolved evidence keeps the affected expansion or benefit claim on hold.

Implications

[AI] The case supports a controls-design and measurement-integrity mandate. It does not establish historical causal savings. Information value is modeled separately as improvement in a feasible decision, net of evidence cost; it is never added to the same operating benefit.

Reading the evidence

The report distinguishes historical record [HR], recalculated findings [RC], analytical inference [AI] and proposed design [PD]. These labels describe the evidence supporting each statement about this actual project.