AI and IFRS: What's Actually Changing in Financial Reporting Judgement (2026)

IFRS compliance has always leaned on judgement: how quickly should a bank recognise expected credit losses under IFRS 9, when exactly does a performance obligation get satisfied under IFRS 15, what discount rate belongs in a lease liability under IFRS 16. AI tools are now sitting inside a lot of that work — drafting disclosure notes, flagging anomalies in a trial balance, summarising long contracts for revenue recognition review. What's actually changed, and what hasn't, is worth being precise about.

Where AI Is Genuinely Being Used in IFRS Reporting Today

The realistic use cases cluster around three areas. First, disclosure drafting: large language models are good at producing a first-pass draft of a note based on prior-year wording and this year's numbers, which a preparer then edits rather than writes from scratch. Second, anomaly and pattern detection: AI tools scan large transaction sets for outliers that might indicate a IFRS 15 revenue recognition issue — a contract modification that wasn't reassessed, a bill-and-hold arrangement that doesn't meet the criteria. Third, contract summarisation: reviewing hundreds of customer contracts for the five-step model used to be a slow manual exercise, and AI now does the first pass of extracting performance obligations and variable consideration terms for a human reviewer to check.

What AI Still Can't Do — the Judgement Core of IFRS

None of this touches the actual judgement calls the standards require. Determining whether there has been a significant increase in credit risk under IFRS 9's expected credit loss model is a forward-looking assessment involving macroeconomic scenarios and probability weightings — an AI tool can help build the scenarios, but the conclusion about which scenario is most likely, and how much weight to give it, is still a human call that has to be defensible to an auditor. The same is true of lease term reassessment under IFRS 16, variable consideration constraints, and materiality judgements generally. AI can narrow the analysis and surface the relevant facts faster; it doesn't remove the professional scepticism that IFRS preparation and audit both depend on.

The New Risk: Over-Reliance Without an Audit Trail

The practical risk finance teams are running into isn't that AI gets the judgement wrong — it's that nobody can show their working. If an AI tool suggested a probability weighting for an ECL scenario, or flagged (or failed to flag) a contract modification, and that becomes part of the basis for a reported number, the team needs to be able to explain how that conclusion was reached when the auditor asks. Firms that have adopted AI well in this area treat it as a drafting and research assistant that produces a documented first pass — not as the final word — and they keep a record of what the AI proposed versus what a qualified preparer decided and why. This matters as much for external audit evidence as it does for internal control sign-off: an auditor reviewing judgement-heavy balances will ask how a conclusion was reached, and "the AI suggested it" is not an answer that stands on its own.

Where This Sits Alongside the Rest of the Close Cycle

It's worth separating this from the broader wave of AI use in the financial reporting close cycle — reconciliations, variance analysis, month-end commentary. That work is largely mechanical and AI has made genuine speed gains there with comparatively low judgement risk. IFRS standards-application is a narrower, higher-stakes slice of the same broader reporting function, and it deserves a different level of scrutiny before a team leans on AI output.

Building IFRS and AI Skills Together

For anyone working toward ACCA Strategic Business Reporting, or already qualified and working with IFRS day to day, the practical skill isn't "using AI" in the abstract — it's knowing which parts of an IFRS judgement an AI tool can genuinely speed up versus which parts still need a human signature. That's a different skill set to general AI tools accountants are adopting across bookkeeping and month-end close, because IFRS work carries audit and regulatory exposure that most administrative AI use cases don't. Recent industry surveys put AI adoption across finance functions well above half, up sharply from a few years ago — but adoption of a tool for drafting speed is a different thing from delegating a recognition judgement to it, and the two shouldn't be confused when a team is setting its own policy on where AI is and isn't allowed to touch reported numbers.

FAQ

Will AI replace the judgement calls IFRS requires?
No. Standards like IFRS 9's expected credit loss model and IFRS 15's five-step model are built around estimates and professional judgement that have to be defensible to an auditor — AI can support that judgement with faster analysis, but it doesn't remove the need for it.

What's the biggest practical risk of using AI in IFRS reporting?
Losing the audit trail. If an AI-generated draft or flagged anomaly feeds into a reported number, the team needs a documented record of what was proposed and what a qualified preparer actually decided, and why.

Where should someone start if they want to combine IFRS knowledge with AI skills?
Start with the standard itself — IFRS 9 and IFRS 15 judgement areas specifically — then layer on the general AI tools already used for drafting and anomaly detection, rather than treating "AI" as a separate qualification.

IFRS reporting is a good example of where AI genuinely helps without changing what the job actually is: the standards still require someone to make and defend a judgement call, AI just changes how much of the surrounding work that person has to do by hand first.