Concern about AI is rising as fast as adoption. Here are the five worries worth taking seriously, and why none is a reason to wait.

-----------------------------------------------------------------------------------------------------

The first time I pointed an AI tool at a reconciliation I'd normally spend a morning on, the speed struck me as much as it would anyone. By the second run, the question that mattered wasn't how it had done the work, but where the data had been and who was accountable for the result.

That second question is the one I hear most from finance teams, and it deserves a proper answer. I spend my time now creating and deploying AI tools for charity finance, and my current assignment spans a finance, facilities and digital remit. The closer I get to the work, the more convinced I am that the anxieties around it are not obstacles to adoption. They are the agenda for it.

Adoption is rising quickly. ACCA's Global Talent Trends 2026 found that 52% of finance professionals now regularly use AI, while 51% worry about its impact on jobs, up from 44% a year earlier (ACCA Global Talent Trends 2026). And our corner is the most cautious: charity and not-for-profit is the most AI-recruitment-sceptical sector and joint-lowest on adoption, at 43%.

I don't read that caution as a problem to talk people out of. I read it as finance teams doing their job. The task isn't to suppress the anxieties, but to work through them, so adoption happens on a footing people can defend to their auditors, trustees and themselves. So let me take the worries I hear most, in the order they tend to arrive.

Governance, and an actual policy

Most charities start using AI before they have written a line of policy about it. A short, plain AI policy — what tools are approved, what data may and may not go into them, who signs off new uses — does more to enable confident adoption than any technology. It turns a nervous "are we allowed to?" into a clear "yes, within these lines".

Cyber and data risk

Charities hold some of the most sensitive personal data there is, about donors, beneficiaries and staff. The discipline is simple to state and harder to live: sensitive or personal data should not go into tools that aren't approved to hold it. Much of my own AI work sits well away from personal data, on management accounts and board reporting. Knowing where the sensitive material is, and keeping it out of the wrong places, is most of the battle.

Shadow AI

Whatever your policy says, some of your people are already using tools you haven't approved, because they're stretched and these tools help. A ban doesn't stop that; it drives it underground, where you can't govern it. Better to point people towards approved, registered tools, make it easy to ask for a new one, and treat curiosity as something to channel rather than punish.

Training

It is easy to use these tools badly. The difference between sharper analysis and confident-sounding nonsense is largely training. ACCA's research is encouraging: 82% of finance professionals are confident they can learn AI skills, and 43% of employers now provide upskilling, up from 32%. A modest, deliberate programme will return more than the tools themselves.

Cost

This is the newest worry and the one I'd flag hardest. The more capable models are also more expensive, and vendors are changing how they bill as the market matures. Budget for AI deliberately, and watch the unit economics — what each use actually costs and saves — rather than assuming today's price is fixed.

None of these five is a reason to hold back. Each is a thing to get right so you can move forward without looking over your shoulder. The organisations that get the most from AI are not the boldest, but the ones that do this groundwork first.

Which is why I talk about governance before gadgets. The exciting part of this work is what it gives back to people: hours returned to the judgement and relationships only a person can offer. But you only enjoy that if the foundations are sound. As ACCA puts it, AI is a productivity layer, not a replacement, and accountability does not change. When the accounts are signed, the Finance Director still owns the numbers. Get the governance right, and you don't have to choose between caution and progress.

Alan de Sousa Caires FCCA is Director, Finance & AI Advisory, AI Finance Office

---