From Event to Process Is Not Enough: Sampling Blindness, the Audit Orchestrator, and the Future of Systemic Visibility

Abstract

Prior research shows that AI, audit analytics, and continuous auditing can improve audit quality. Yet these approaches assume that more data and better tools are enough. This paper argues that the core problem is deeper. It lies in a structural limitation of auditing itself. I define this limitation as "sampling blindness," a condition in which periodic, sample-based auditing is inherently unable to detect risks that evolve within continuous and interconnected systems. Drawing on an analysis of audit reports, regulatory findings, and cross-sector cases, the paper shows that audit failure reflects limits in what auditors are able to see, rather than only limits in technology. To address this gap, I introduce the Audit Orchestrator, a framework that expands the auditor's field of view across three dimensions: transactional data, system logic, and organizational communication. Illustrative simulations show how this approach can identify risks earlier and more consistently. The paper also highlights an economic barrier, where time-based compensation discourages continuous monitoring.

Keywords: Algorithmic Accountability, Systemic Risk, Risk Management