A new survey of CEOs and risk professionals finds that the increased use of AI has introduced inaccuracies into reporting, including cases where errors have reached the board and public-facing publications.

The Verification Gap: AI Enthusiasm Runs Into Reality was released earlier this month by the software company Workiva and is based on a survey of more than 2,700 global finance, risk, sustainability, and legal professionals, including more than 800 C-suite leaders, as well as more than 350 institutional investors. A fourth of the respondents (26 percent) said internal audits uncovered errors that “reached external audiences or the board.” Moreover, only 11 percent of respondents said their data quality “is sufficient for AI use.”

The report suggests that organizational enthusiasm for AI hasn’t quite matched up with the actual efficiencies it promises. “Executives are more likely than practitioners to be ‘very confident’ in AI outputs appearing in annual reports without human review,” the report says.

A key challenge, as the 11 percent figure suggests, is that data governance is not rigorous enough to make AI outputs consistently reliable; 27 percent of respondents say poor data quality “has significantly blocked deployment in key workflows.”

Even so, substantial proportions of executives say they’re expecting to see measurable ROI on AI in the next year, be it through revenue growth (58 percent) or time savings (47 percent).

The findings follow other research surveys that have found that while AI adoption is growing, its benefits are elusive. The 2026 Enterprise AI Strategy Pulse Survey, released earlier this month by Plug and Play, found that while 74 percent of top-tier companies have “at least one AI solution live in production” and 93 percent are actively piloting one, half of the respondents can’t yet say whether it’s delivering ROI; nearly a third (31 percent) say their ROI has “met expectations.”

The Plug and Play report echoed Workiva’s finding that data governance presents a key challenge, with 71 percent saying that “data foundation” issues present a barrier to scaling up AI efforts. 

The Plug and Play report recommended that more organizations should establish clear guidelines around its AI goals to determine ROI. “Measurement is most useful when it actually informs decisions,” the report says. “That starts with setting clear baselines before deployment, using control groups or A/B testing where possible…and tracking total cost of ownership.”

The Workiva report similarly recommends that organizations establish systems to improve data quality and redesign job roles to clarify how and when employees will work with AI agents, with transparency about the AI risks. 

“Confidence in AI is not the same thing as having control,” the report concludes. “The leaders in the best position to close the gap between AI confidence and control are treating data quality, governance, and transparency as strategic advantages. They are building cross-functional partnerships across the C-suite to drive ROI from AI initiatives.”