Two numbers from the same survey should not be able to coexist. In a poll of 2,272 finance, risk, sustainability, and legal professionals published this week, 84 percent said they are at least somewhat confident in AI-generated content appearing in annual reports without human review (39 percent very confident, 45 percent somewhat). In the same survey, 26 percent said their internal audit teams have already caught AI errors that reached the board or external audiences. Same respondents, same organizations, same survey.

The survey is Workiva’s 2026 MidYear Executive Benchmark Survey, an independent study Workiva commissioned from research firm Ascend2, fielded in May across sixteen countries at organizations with at least 250 employees or $250 million in revenue, and covered by Accounting Today this week. Standard caution applies: Workiva sells controlled financial reporting software, so it benefits commercially from discovering a control problem, and it got to name the phenomenon (the “verification gap”) after itself found it. The sample is large and multinational all the same, and the finding that matters is not flattering to anyone’s product. It is the distance between those two numbers.

Because think about what the annual report is. It is the most reviewed, most signed, most litigated document an organization produces, the one with certifications attached and securities law hovering over it. It is the last place a rational organization would remove human review. Yet 84 percent of the people responsible express at least some confidence in exactly that, while one in four of their own audit shops has already fished an AI error out of material that escaped to the board or beyond. A third number from the same survey completes the picture: just 11 percent say their data is actually in shape for AI use. Confident in the output, burned by the output, unready underneath it, all at once.

If the top of the organization sounds like this, recall what the bottom sounds like. A month ago we covered survey data showing nearly half of finance teams spending 15 or more hours a week checking AI output. Put the two datasets side by side and you get the real org chart of AI adoption in 2026: executives extending trust at the top, staff quietly absorbing a verification workload at the bottom, and the gap between them papered over by nothing more than the staff not having failed yet. Workiva’s chief accounting officer, Junko Swain, told Accounting Today the statistic “sets off alarms but it’s not surprising,” because it shows AI errors “aren’t hypothetical risks, and business leaders still need human oversight and data lineage.”

Our read: confidence is descending the org chart faster than verification capacity is ascending it. That is the exact recipe for the 26 percent to grow, because trust without review does not eliminate errors, it just delays where they surface, and the annual report surfaces them in public.

What to do with this if it describes your shop. If you are the executive: treat the 26 percent as your early warning, and before extending confidence, ask who reviewed the AI-touched sections of your last external document and where that review is documented. If nobody can answer in one email, you have your answer. If you are the controller or the staff accountant doing the checking: document what you catch. Every error you intercept is currently invisible value; a log of caught errors is both your budget case and, the day something slips through, your defense.

The bottom line: 84 percent of executives are, in effect, ready to let the machine sign. But an annual report is not content, it is a signature, with names, certifications, and liability attached. AI can do the work. It cannot sign the work, and the quarter of organizations that already watched an error reach their board have learned which half of that sentence the machine cannot deliver. The others are scheduled to learn it later, at a worse moment, in a more public document.

— Footnote

Footnote is an independent publication. It is not professional accounting, tax, or legal advice. Survey figures are from Workiva’s published survey and Accounting Today’s reporting, linked above; Workiva is a financial reporting software vendor and the survey’s framing is its own. Details are current as of August 14, 2026.