AI in finance and accounting

The obligations that reach an accountant using AI are not where most people look for them. Auditing standards bind auditors, and an accountant preparing a forecast, reconciling a ledger or drafting a disclosure note is not doing audit work. The binding material for that audience sits in the ethics code, which has required a professional accountant to exercise judgement about what steps to take when using the output of technology since December 2024 — and which almost nobody cites for AI, because the provision never uses the word. It says "technology", so it does not surface for the questions it answers.

The code goes further than a single provision. It defines objectivity to include freedom from undue reliance on technology, it names automation bias in terms as a bias to be aware of when exercising professional judgement, and it ends in a duty to refuse to remain associated with information the accountant still believes to be misleading. Those are not aspirations in a preface; they are numbered requirements, and the same provision on using technology output appears twice — once for the accountant in business and once for the accountant in public practice — differing on whose oversight of the system is in question.

Around that sit three further regimes with a common shape, and naming it is the point of this collection: the US issuer rules, the internal-control frameworks and the UK governance code all now require a person to attest, at a point in time, to something an AI system does continuously. That mismatch between a periodic certification and a continuous process is where the assurance work actually is. These pages cite each provision by number and official title and argue the substance in our own words, because most of this material is paywalled and none of it is ours to reproduce.