Workiva released its annual AI governance survey on August 24, and the numbers reveal a dangerous contradiction. 26% of senior leaders say they found errors in artificial intelligence systems after they were already in production and available to the public. Even so, 84% trust AI-generated results without any human review.
The survey shows only 11% of companies believe they have adequate data quality to support AI-driven decisions. This combination — fragile data, known errors, and blind trust — worries investors: 96% rank AI governance as very important, and 89% are concerned about the accuracy of information companies disclose using AI tools.
The study surveyed executives and investors at publicly traded companies. The most serious finding is that errors were detected but did not trigger any process change. The lack of systematic review puts not only reputation at risk but also regulatory compliance, as bodies like the SEC advance rules for AI use in corporate communications.
For those running systems in production, the lesson is concrete: no AI pipeline should feed external reports or financial decisions without mandatory human validation. The blind trust documented by the survey is the shortest path to an incident that reaches the boardroom and the news.
The publication finds the data alarming. Errors in production are inevitable; absence of review after an error is found is negligence. Those operating AI systems must — before any new feature — define mandatory checkpoints, with a designated person responsible and an approval record.