CFO survey: finance teams spend 26% of the workday checking AI output

Global reference

• Country: United States
• Source: Datarails, 2026 CFO Sentiments Survey, carried out by Global Surveyz Research
• Sample: 270 CFOs and finance leaders at companies with 1,000 or more employees and revenue of USD 100 million or more, all using AI in finance
• Fieldwork: July 2026
• Note: Datarails sells finance AI software, so this is vendor research

The numbers

• 26%: average share of the workday spent verifying AI output.
• 96% spend at least 10% of the workday on it; 8% spend more than half.
• 32% exceeded their AI budget by at least 10% in the past year. 53% plan to expand AI licences in the next 12 months.
• 76% feel high or very high pressure to implement AI fully. Only 7% say their finance function is fully ready.
• 4% have a single source of truth for finance and operational data. 73% describe their data as mostly centralised.

Why trust is low

• 75% name lack of auditability as a reason they do not fully trust AI with critical tasks; 71% name accuracy and hallucinations.
• 56% have seen team members get materially different outputs from the same prompt and data.
• Only 5% would trust AI with board-ready reports, and 4% with month-end close tasks, without human review.

Tools and people

• General-purpose tools in use: Microsoft Copilot 93%, ChatGPT 65%, Claude 64%.
• 84% say the CFO or finance leader owns the AI budget; 13% report no clear owner.
• 60% are moving staff to higher-value work; 3% are cutting headcount where AI has replaced work.

How to cut the checking time

• Ask the AI to show the source of every number it quotes, in a tie-out table: the figure, the account code, the column and the row.
• Tell it to write NOT FOUND, and not to estimate, when it cannot trace a figure.
• The reviewer then checks the table against the file, not the whole draft.
• Save the prompt as a standard, so that everyone runs the same one.

What to do now

• Measure checking time for each AI-assisted process for one month, and put it in the business case.
• Fix data sources before adding tools.
• Standardise and store prompts for recurring tasks.
• Name one owner for the AI budget.

Our take

Checking is not waste; it is the control. The aim is to make it faster, not to remove it. Finance teams that ask AI for evidence with every number will spend less time reviewing and will have a better audit trail.

Important. This summary is for general information and is not accounting, legal or financial advice. It is based on public sources, together with vendor research and press reports, and may be incomplete or out of date. Check the original sources, and take professional advice, before acting.