Finance leaders trust AI’s numbers but not the board’s timetable

On 20 October AccountsIQ, which sells cloud accounting software to mid-sized businesses, announced research on how finance teams are handling AI. Four in five finance leaders (82%) told it they feel under pressure to adopt AI in their finance function faster than they are comfortable with, and 41% described that pressure as significant. Censuswide carried out the fieldwork in July and August 2026, surveying 1,036 senior finance decision-makers at companies with 20 to 500 employees: 836 UK and 200 Irish respondents across eight sectors.

The pressure is sharper in Ireland, where 91.5% of respondents feel it, than among the 80.3% of UK respondents who do.

Darren Cran, chief executive of AccountsIQ since October 2024, put the source of the pressure outside the finance team. “Finance leaders are being asked to move on AI at the speed of the conversation around it,” he said in the company’s release. “The discussion of AI starts in the boardroom and the press, and finance is expected to have an answer.”

The rest of the survey points away from doubt about the technology. Most respondents have accepted what AI calculates, and Cran puts the hesitation down to something narrower: a record of where each number came from and who approved it. AccountsIQ, for one, builds that record into its accounting software rather than leaving it to the finance team’s own procedures, so the pace of AI in finance depends on what vendors ship as much as on what boards demand.

Cran puts the pressure in the boardroom

Cran’s account puts boards and the press ahead of finance in the queue: finance is expected to have an answer because the discussion has already started elsewhere. He also separated trust from timing. “Most finance leaders already trust what AI produces, so the hesitation sits with how quickly they can bring it in with the controls they need,” he said.

Some 81% of respondents fully or mostly trust AI-driven outputs in finance processes, according to the research, and only 0.9% have no trust in them at all. Trust in their own data is higher still: 96% trust the accuracy of their financial data, and 91% say they forecast on current, governed data.

Some of that trust is untested

Not every respondent has AI to judge. The same research finds that 27.0% of UK finance teams in the sample have no AI features at all, and 18.6% of respondents say AI has simply not been used enough yet to trust fully. Some 92% say they could explain how an AI-generated output was calculated, a high figure for software that more than a quarter of the teams in one of the two markets do not yet run.

For some respondents, the 81% may reflect an expectation of how AI will perform rather than experience of AI-prepared journals and reconciliations. The concern that does come through clearly is dependence: 76% worry about their organization becoming too reliant on AI for decisions that used to be theirs.

The vendor sells the controls

Asked what would make them more comfortable implementing AI, respondents spread their answers across six conditions. Confidence that AI mistakes would be caught before reaching a report led, at 26.1%. Then came a person approving entries before they are posted (24.7%), clear rules for labeling AI-assisted numbers (24.6%), being told how often AI gets things wrong (23.9%), an explanation of each decision (22.9%) and an audit trail from any number back to its source (22.1%). Respondents could pick more than one, and no condition was chosen by more than about a quarter of them.

Cran described controls in terms of the time they take. “In finance, controls mean knowing where a number came from and who approved it,” he said. “Those take time to put in place, and a fast timetable can leave little room for them.”

The list reads closely against AccountsIQ’s own product page. Its AI, the company says, prepares work inside the ledger and holds every suggestion until a person approves it: receipts and payments are matched with the reasoning shown beside each pair, nothing reaches the ledger until a user accepts it and approved entries carry a full audit trail. A vendor’s survey finding that buyers want what the vendor sells is common enough. AccountsIQ positions itself between starter packages such as Sage, Xero and QuickBooks and full ERP suites such as NetSuite and SAP, and its AI page opens on “Drafted by AI. Decided by Finance.” before it mentions time saved.

Boards ask now, auditors ask later

AccountsIQ reports that 68% of finance leaders still decide without enough information, and it recommends that finance teams set their own AI timetable and put the six conditions in place first. Cran made the same case in terms of scrutiny after the fact: “Adopting at a pace that can be explained to a board, and its auditors makes the later questions easier to answer.”

Boards want an AI answer now, auditors will ask for the workings later and the finance director answers to both. The survey’s respondents have mostly settled whether AI output can be believed. At AccountsIQ the record of who approved each number ships inside the ledger, and in the mid-market the pace of adoption runs on vendors’ release schedules as much as on any board’s timetable.