Accounting & Auditing Conference - Virtual-5 - Heine - AI-Good-Governance-and-Real-Life-Examples

5 - Heine - AI-Good-Governance-and-Real-Life-Examples

Pdf Summary

The presentation argues that AI in financial reporting should be treated like an aircraft autopilot: useful for routine work, but dangerous when people stop monitoring it, misunderstand how it operates, or lose the skills needed to take over.<br /><br />Aviation accidents, including Air France 447 and Asiana 214, show how over-reliance, mode confusion and inadequate practice can turn automation into a hazard. The lesson for finance is not to prohibit AI, but to govern its use through clear procedures, trained reviewers, proficiency checks and documented evidence.<br /><br />AI is already used for drafting, research and reconciliations, and its role is expanding. Yet polished output can contain fabricated citations, unsupported explanations or unnoticed errors. Reviewers may be overly confident, fatigued, poorly trained or unable to verify work outside their expertise. Time saved can disappear into more tasks instead of being reinvested in review, while junior staff may miss the foundational work that develops professional judgment.<br /><br />The presentation recommends applying existing COSO and internal-control-over-financial-reporting principles to AI. It gives control practices for four use cases: reconciliations, variance analysis and disclosure drafting, technical research and contract abstraction, and AI-assisted control testing. Across these, reviewers should verify outputs against primary sources, independently reperform samples, confirm data completeness and tool versions, document review, and retain clear human accountability.<br /><br />Organizations should inventory AI use, assess risk, assign owners, define review standards, train people to evaluate outputs, preserve manual proficiency, monitor performance and prepare fallback procedures if tools fail or change. The central message: use AI, but deliberately reinvest some efficiency gains in verification and workforce skills. A reviewer who could not have prepared the work is not truly reviewing it.

Keywords

AI in financial reporting

automation risk

aircraft autopilot analogy

over-reliance on AI

human oversight

COSO internal controls

financial reporting controls

review and verification

AI governance

workflow proficiency