AI Errors Reach Boards: 26% Admit Internal Audit Failures

A Workiva survey highlights critical failures in AI oversight within organizations. Astonishingly, 26% of senior leaders reported that internal AI audits detected errors only after they had reached the board or external parties. This finding compounds the alarming statistic that 84% of executives express confidence in AI output for disclosures without human review, despite only 11% deeming their data quality adequate for AI applications.

Investor sentiment underscores the urgency for robust AI governance. Nearly all institutional investors (96%) stated that AI governance oversight policies influence their investment decisions. Currently, a standardized framework for AI governance disclosure is absent, forcing investors to meticulously analyze company reports and transcripts to gauge a firm's AI stance. This lack of transparency is concerning, especially as AI accuracy in financial disclosures is a major worry for 89% of investors, with 47% actively seeking signs of AI-generated errors.

The survey also exposed a disparity in AI confidence between executives and data professionals. Executives showed higher confidence (39%) in unreviewed AI output compared to the professionals handling the data (29%). This suggests a disconnect regarding the practical risks of AI implementation. Steve Soter of Workiva advises audit committees to demand concrete evidence of AI usage and validation processes from management, rather than accepting general assurances. External auditors are expected to pose similar questions, making proactive AI governance essential.