CFOs spend 26% of time checking AI outputs, survey finds

Datarails has published survey findings showing that finance teams spend an average of 26% of their workday verifying or correcting AI-generated outputs. The research covered 270 Chief Financial Officers and finance leaders at large US companies.

All respondents said they use AI in some part of their finance processes, indicating universal adoption among the group surveyed. Yet 96% said they spend at least 10% of their time checking finance-specific AI outputs, while 8% said those checks take up more than half of their working time.

The figures suggest that AI has created a new operational burden in finance even as adoption has become routine. The survey also found that only 5% of finance leaders trust AI to produce board-ready financial reports without human review, and just 4% trust it to handle month-end close without oversight.

Concerns about trust were strongest around auditability. Three quarters of respondents, or 75%, said lack of auditability was the main reason they hesitate to rely fully on AI for critical finance tasks. That ranked ahead of accuracy and hallucination concerns at 71% and regulatory or compliance concerns at 54%.

Problems with data quality and consistency also featured heavily. Nearly two thirds of respondents, or 65%, said their most common frustration with AI was getting confident answers based on the wrong data. Meanwhile, 56% said team members had received materially different outputs from a large language model despite using the same prompt and data.

"What this survey shows is that although AI adoption is now standard for CFOs, caution hasn't disappeared," said Didi Gurfinkel, Chief Executive Officer and Co-Founder of Datarails.

"Finance teams have stopped asking 'will we use AI?' and started asking 'how am I going to most effectively check its work?'" Gurfinkel said.

Budget pressure

The survey points to rising costs alongside growing use. Nearly one third of respondents, or 32%, said their organisations exceeded AI budgets by at least 10% over the past 12 months, while 53% said they plan to expand AI licences across their organisations over the next year.

Budget ownership was not always clearly defined. While 84% of Chief Financial Officers said they own their organisation's AI budget, 13% said there was no clear owner.

General-purpose AI tools remain common in finance departments. Respondents said their teams use an average of 2.5 large language models, with Microsoft Copilot used by 93%, ChatGPT by 65% and Claude by 64%.

Jobs picture

The findings also indicate that expected AI-driven job cuts in finance have largely not occurred. Sixty per cent of respondents said that as AI takes on more routine finance tasks, they are moving staff into higher-value work rather than reducing headcount.

Only 3% said they are actively cutting full-time roles in the Chief Financial Officer's office because of AI. That contrasts with broader concerns that automation would directly reduce staff in back-office functions.

"Fears of job losses have been largely allayed, but the challenge of AI output verification is critical. Tackling it will allow finance teams to spend less time on intensive checking and focus instead on the strategic work they were actually hired to do," Gurfinkel said.

Data challenge

The survey suggests many finance teams still lack the data foundations needed for reliable AI use. Only 4% of respondents said their organisations have achieved a single source of truth for finance and operational data, while 73% described their data as mostly centralised and 23% said they still depend on disconnected systems and manual reconciliation.

Those weaknesses appear linked to poorer AI results. Among teams whose main operational challenges are manual reporting and data consolidation, 86% said AI had generated a confident answer based on the wrong data, compared with 65% of all respondents.

That gap is also driving a new technology priority. Some 32% of finance leaders said they are prioritising a finance operating system, described in the survey as a governed data layer built for AI. That made it second only to planning and financial planning and analysis tools at 42%.

Pressure to roll out AI remains high despite those shortcomings. More than three-quarters of respondents, or 76%, said they are under high or very high pressure to implement AI fully, but only 7% said their finance function is fully ready to deploy it across all workflows.

Global Surveyz Research conducted the survey online on behalf of Datarails among Chief Financial Officers and finance leaders at US organisations with more than 1,000 employees and annual revenue of at least USD $100 million. Respondents completed the research through a global panel in July, and every participant reported using AI in some capacity in finance processes.