A new survey of 1,036 UK and Irish finance leaders finds 82% feel pushed into AI faster than they are comfortable with. In Ireland it is 91.5%, and the pressure starts in the boardroom.
On 20 October a Dublin accounting software company published a survey of 1,036 senior finance decision-makers in the UK and Ireland. The headline was that 82% feel pressure to adopt AI faster than they are comfortable with. The Irish figure was 91.5%. These are CFOs, finance directors and financial controllers at companies with 20 to 500 staff, and in most of those companies the person applying the pressure sits on the board.
Treat the publisher with the usual scepticism. AccountsIQ sells finance software with AI features, the Irish sample was 200 people, and the fieldwork ran in July and August. None of that changes the direction of the finding. We hear the same thing in almost every finance conversation we have had this year. The board has read about AI. The CFO is being asked where the firm's AI is. Nobody has been asked what it should be for.
The pressure is coming from inside the building
The survey does not say where the pressure originates. AccountsIQ's chief executive, Darren Cran, put it down to "the speed of the conversation around it", meaning boardrooms and media coverage. That matches what we see. A non-executive director attends a conference, a competitor announces an AI programme, a peer firm mentions a pilot over dinner, and the question lands on the finance director's desk the following Monday: what are we doing on this?
That question has no good answer, because it asks about activity rather than outcome. A finance director who answers "we have three pilots running" has satisfied the board and told it nothing. A finance director who answers "nothing yet" looks like a laggard. So the pilots get started, the licences get bought, and the operational readiness that would make any of it safe is left for later. 41% of respondents described the pressure as significant. In Ireland, on the smaller sample, the pressure reading was eleven points higher than in the UK.
The board is the source of the problem and the only place it can be fixed.
The trust numbers are the alarming part
The same survey reports that 81% of finance leaders fully or mostly trust AI-driven outputs, and 92% say they could explain how an AI-generated output was calculated. Fewer than 1% said they had no trust at all.
Read those alongside the control questions, where respondents could tick as many options as they liked. Only 26.1% wanted assurance that AI mistakes are caught before they reach a report. 24.7% wanted a person to approve entries before posting. 22.1% wanted an audit trail from any number back to its source. 23.9% wanted to know how often the tool gets things wrong.
Those are the controls a finance function would demand of a new junior accountant, let alone a probabilistic system. Fewer than one in four finance leaders asked for any single one of them. High trust combined with a low appetite for verification is the exact condition in which an unreviewed number ends up in a board pack or a lender covenant calculation. The High Court has already put a price on unverified AI output in litigation. Finance will get its equivalent.
Then there is the 76% who worry about their organisation becoming too dependent on AI for decisions that used to be theirs. Three in four see the risk. Fewer than one in four are asking for the controls that would manage it. That gap is where the governance work sits.
What a board should ask instead
The fix is a change in the question. Stop asking management where the AI is. Ask what evidence would make a specific use of it safe. Four things in practice:
Name the decision before the tool. Which finance decisions are candidates for AI support: cash forecasting, accruals, intercompany reconciliation, supplier matching? For each one, write down who owns the number today and who will own it afterwards. If the answer to the second question is "the system", stop.
Set the implementation pace in the finance function, and have the board ratify it. A timetable the CFO can defend to the audit committee and the external auditor is worth more than one that matches a competitor's press release. Cran's own line is the right one: adopting at a pace that can be explained to a board and its auditors makes the later questions easier to answer.
Put the controls in before go-live. Error-rate measurement. A named human approving entries before posting. A labelling rule for AI-assisted figures in any pack the board sees. A traceable line from every AI-touched number back to its source. These are the controls fewer than a quarter of respondents asked for, and they are the minimum.
Fund the data work the survey is hiding. 68% of respondents say they still make decisions without enough information. AI applied to a ledger with inconsistent cost centres or three versions of the customer master produces faster wrong answers. Information architecture comes before AI, and it is usually the larger budget line.
The uncomfortable implication for directors
If 91.5% of Irish finance leaders feel rushed, most Irish boards are doing the rushing, and doing it from a position of limited AI literacy. Article 4 of the AI Act calls on deployers to support AI literacy among the people using AI on their behalf. A board that sets the pace on AI adoption is directing its deployment. The literacy duty reaches the boardroom.
A finance director who pushes back on the timetable, with the control gaps written down, is doing the board's risk job for it. Good boards will recognise that. The rest will find out what the trust numbers mean the first time an auditor asks where a figure came from.
The Acuity AI position
Pressure to adopt is a signal, and most of the time it signals that the board has not yet done its own diagnostic. We work with Irish boards and finance leadership on precisely this: which decisions AI should touch, what controls have to exist first, and what the organisation's data can honestly support. The output is a pace the CFO can defend and the board can ratify. Vendor-neutral, evidence first, and it starts with the question nobody has asked yet: what is this actually for? If that question has not been put to your board, a diagnostic conversation is the place to start.