AI Accountancy
AI accountancy needs evidence, not confidence.
AI can now produce answers that look polished, fluent and persuasive. That is useful. It is also dangerous. In accountancy, the important question is not whether the answer sounds right. The important question is whether the answer is supported by evidence.
What AI changes
AI collapses the cost of production. A first draft that took a trainee an afternoon now takes a minute. Reconciliations can be compared, ledgers scanned for anomalies, disclosures checked against a checklist, and a hundred pages summarised before the kettle boils. That is not a marginal improvement. It rewrites the economics of every accountancy practice, because the part of the job that filled most of the timesheet is no longer the expensive part.
What AI does not change
The accounts still need to be right. The client still relies on them, the bank still lends against them, HMRC still assesses on them, and a buyer still prices a business off them. A model can produce something that looks like a set of accounts, but it cannot take responsibility for one. Responsibility is not a feature you can automate. It belongs to a named, qualified person, and every rule of the profession is built on that fact.
Why evidence matters
AI sounds exactly as confident when it is wrong as when it is right. There is no wobble in the voice. The mistake that gets through is never the obvious one; it is the plausible one, the figure that sits neatly in the table and sails past everybody because nothing about it looks wrong. The only defence is the discipline the profession has always claimed to practise: show me the evidence. A fluent answer is not a working paper, and an inconsistency is not automatically a finding.
Why professional judgement becomes more valuable
When production is cheap, review is the scarce skill. Someone has to decide which anomalies matter, which estimates are reasonable, which explanations survive scrutiny and which numbers deserve a second look. That is judgement, and it is learned by doing, questioning and being challenged. The accountants who can judge evidence will be worth more in an AI workflow, not less, because everything the machine produces flows towards them.
The future is not AI replacing the accountant. It is AI raising the standard of review.
How accountancy firms should think about AI adoption
There are two roads. Bolt AI on as a clever assistant, and it quietly gets promoted to author: a tired human gives a polished draft the lightest glance and out it goes. Or build it into a workflow, where nothing the machine produces reaches a client or a regulator until a named person has checked the part that matters against something real. The second road is more work. It means deciding, for every output, what checked actually means and who owns it. It is also the only road that survives contact with a bad week.
What SME owners should expect from AI-enabled accountants
Not just faster accounts and a lower fee. Better questions, caught errors, and an accountant who can explain what the numbers mean rather than just producing them. If your accountant is using AI, ask them a simple question: who checks the output, and against what? A good firm will have a precise answer. A firm that says the software handles it has told you everything you need to know.