A recent executive survey from Workiva has uncovered a significant gap between leadership confidence in artificial intelligence and the reality of AI errors reaching the highest levels of organizations. According to the 2026 Midyear Executive Benchmark Survey, 26% of senior leaders reported that an internal AI audit caught an error after it had already reached the board or external audiences. The figure rises to 21% among Australian executives, where one in five say internal audits have detected AI errors that reached external audiences or board members.

Despite these incidents, confidence in AI remains high. Globally, 84% of executives said they would be at least somewhat confident in AI output appearing in an annual report without human review. In Australia, 79% of surveyed executives expressed at least some confidence in the accuracy of AI output without human review. However, the survey reveals a striking disconnect: only 11% of executives globally believe their organization's data quality is sufficient for AI use, and in Australia that number is just 14%.

Data Quality and AI Accuracy

The survey highlights that poor data quality is a major obstacle to effective AI deployment. Among Australian executives, 21% said poor data quality has significantly blocked deployment in key workflows, while 65% said it has at least moderately impacted the use of AI in financial and sustainability reporting. Kristen Pimpini, VP and General Manager APJ at Workiva, underscored the problem: "Bad data dressed up by good AI is still bad data; it’s just faster and harder to catch. Before leaders confidently hand more decisions to AI, they need to know where the numbers came from and be able to prove it when someone asks."

The consequences of inadequate data governance extend beyond operational challenges. The survey found that 96% of institutional investors globally indicated that AI governance oversight policies factor into investment decisions, with 62% calling it "very important." Yet there is currently no standardized framework for AI governance disclosure, forcing investors to infer a company's approach from risk factors, footnotes, and earnings call transcripts.

Investor concerns are well-founded. A separate analysis by the Conference Board and ESGAUGE found that while 72% of S&P 500 firms flagged at least one material AI risk last year, only 2% cited inaccurate outputs as a risk. This disconnect is reflected in investor sentiment: 89% of survey investors expressed concern about AI accuracy in disclosures, with 47% saying they watch closely for signs of AI-generated errors.

Confidence vs. Reality

The survey reveals a notable difference in confidence between executives and the professionals who handle the data directly. Globally, 39% of executives expressed high confidence in unreviewed AI output, compared to only 29% of the professionals who work with the underlying data. This suggests that those closest to the data are more cautious about its reliability.

Steve Soter, Workiva vice president and industry principal, treats practitioner caution as competence rather than hesitation. He noted that generic large language models "can produce polished, inaccurate outputs that give an illusion of quality." Soter recommends that organizations "Stop accepting general assurances about AI and start requiring evidence." He outlined three steps: trace then certify, convert assurance into evidence, and close investor signal gaps.

The Australian Perspective

The Australian findings paint a complex picture. While 74% of Australian executives say AI has improved the quality of at least some of the work produced, the survey also identifies significant risks. Among Australian executives:

  • 43% are concerned about the risk of intellectual property leakage from sharing proprietary data through unauthorized AI tools.
  • 39% are concerned about the verification gap, where employees fail to check AI-generated data before it is used externally or presented to the board.
  • 39% say their organization's most recent AI audit uncovered data lineage gaps or traceability issues.
  • 29% say their most recent AI audit uncovered gaps in governance policies or controls.
  • 21% say their most recent AI audit uncovered hallucinations or errors that reached external audiences or the board.
  • Only 38% say their most recent AI audit revealed validation of existing controls.

The survey also found that Australian executives are increasingly focused on AI training and governance. 88% say AI training is a top priority for the second half of the year. When asked about future AI governance needs, 59% say they will need platforms to manage agents and automated workflows, 51% say they will continue to need systems of record such as general ledgers, and 50% say they will need software that enables traceability and audit.

Measuring the return on AI investment remains a challenge. Australian executives report using a variety of metrics, including revenue growth (61%), time savings (53%), automation rate (48%), internal rate of return (36%), and revenue per employee (23%).

David Conley, Head of Reporting at Challenger, emphasized the broader goal: "Finance teams should be spending more time supporting their businesses and its initiatives, and less time on the process of reporting and closing out the month and year end."

The Workiva 2026 Midyear Executive Benchmark Survey polled 2,272 finance, risk, and sustainability professionals, including 847 C-level executives. The findings suggest that while AI adoption is proceeding rapidly, the foundations of data quality and governance have not kept pace, leaving organizations vulnerable to errors that can reach the highest levels of scrutiny.