Can a financial institution use AI to draft its internal audit report and examine records without an auditor rechecking each one? Our analysis supports both, where the responsible individuals can perform their roles effectively and the evidence justifies reliance on the automated work.

Suppose AI checks a quarter’s trades against a pre-trade approval requirement and flags ten breaches. An auditor confirms all ten. That supports reporting those breaches, but does not establish whether another twenty were missed or whether the control operated effectively across the quarter. The same limitation applies when people perform the initial checks.

Human oversight and manual examination are different tasks. A competent auditor must assess and adopt an AI-drafted report, but that does not necessarily require repeating every underlying check. The evidence must support the intended audit conclusion.

Written by Frederic Thieltgen and Chris Holland, our article works through the Singapore regulatory framework, including testing and monitoring, independent validation, outsourcing and AI governance. It explains how an FI can assess automation procedure by procedure and identify when further testing, human intervention or suspension is required.

For the full article, click here.