AI in finance and accounting
Section 200 names two categories of threat arising from technology use in a finance function. The self-interest limb is familiar — inadequate data, an unsuitable tool, nobody who can explain it. The self-review limb is the one that gets skipped: where the accountant or the employing organisation supplied the knowledge, expertise or judgement that built the system.
§ 1 — Who it binds
Professional accountants in business applying the conceptual framework wherever technology is used in the work of the function — tools bought in, tools configured in house, and models the finance team itself specified or trained. These threats are identified at the level of the function’s arrangements rather than of a single output. The Code takes effect through the IFAC member body that has adopted it.
§ 2 — In practice
Section 200 is where the conceptual framework is applied by professional accountants in business, and 200.6 A2 sits under a heading the Code wrote for this exact purpose. It names self-interest threats first, and its three examples are recognisable to anyone who has adopted a tool in a hurry. The data available might not be sufficient for the technology to be used effectively. The technology might not be appropriate for the purpose it is being put to. And the accountant might lack sufficient information and expertise, or access to an expert, to use the technology and to explain it and its appropriateness. Each is stated as a threat to compliance with the fundamental principles, which means it has to be evaluated and addressed rather than accepted as a feature of the environment.
The self-review limb is the half that gets skipped, and it is the more interesting one. A self-review threat arises where the technology was designed or developed using the knowledge, expertise or judgement of the accountant or of the employing organisation. Read plainly, a finance team that specified, tuned or trained a model is now reviewing the output of its own judgement — the definition of the threat, arriving somewhere nobody expects it. “We built our own forecasting model” is normally offered as evidence of control; under this paragraph it is the fact that creates the threat. The answers are ordinary ones: separate the people who set the logic from the people who review its output, or bring in something the team did not build to challenge it. The independent validation line supervised institutions draw is the same idea reached by another route.
200.7 A4 supplies three factors bearing on how significant the threat is, and all three are organisational rather than personal: the level of corporate oversight and internal control over the technology, third-party assessments of its quality and functionality, and regular training so that relevant employees can understand, use and explain it. That matters for how an assessment is written. An individual accountant cannot change the level of corporate oversight, but they can establish what it is, and the paragraph makes that fact part of their own evaluation. Where oversight is thin the threat is more significant and the safeguards have to be heavier at the individual level, which is the practical link between this paragraph and the judgement R220.8 requires each time an output is used.
§ 3 — When it applies
15 December 2024
Effective date of the Technology-related Revisions to Parts 1 to 3, which introduced the heading on identifying threats associated with the use of technology into Section 200. An effective date for a code provision rather than a statutory compliance deadline.
§ 4 — What discharges it
The artefacts an assessor asks to see, and what makes each one sufficient rather than merely present.
01
Self-interest and self-review recorded separately for each tool in use, because the safeguards differ. A single entry noting that AI carries risk has identified nothing the paragraph asks to be identified.
02
Names and dates for the logic, the thresholds and the training decisions. It is the only way to establish whether the people evaluating an output are the people whose judgement sits inside it.
03
What corporate oversight of the tool actually consists of, whether any third-party assessment exists, and what training relevant employees have had. Established rather than assumed, since all three move the threat.
04
Separation of reviewer from builder, an independent challenge, or an external assessment — with proof it happened during the period rather than a description of an intention.
§ 5 — Worked example
A group finance team builds its own machine-learning model to predict which customer accounts will go bad, using rules the credit controllers wrote and thresholds the financial controller set. The model feeds the expected credit loss provision. The same team reviews its output each quarter as part of the provisioning review, and the review is signed by the financial controller.
Which threat is present, and does the quarterly review address it?
The self-review threat, and the review as constituted makes it worse rather than better. The model was developed using the knowledge, expertise and judgement of the very accountants who now evaluate its output, which is the paragraph’s own description of the threat; that everyone involved is competent and well-intentioned is not to the point, because this failure is structural. The review is performed by the people whose judgement is embedded in the thresholds, so a threshold that is systematically generous will not look wrong to whoever chose it. Two of the significance factors are also unfavourable: there is no third-party assessment of the model, and corporate oversight of it is the same finance function. The addressing steps do not require dismantling anything — a reviewer from outside the team, a challenge against an independent expectation such as actual write-off experience by cohort, and a record of who set each threshold and when. What the Code will not accept is the model’s provenance being offered as the reason the provision is reliable.
§ 6 — What a weak answer looks like
The threat register that stops at the supplier. The function has assessed the vendor — financial standing, security certifications, a service level — and offers that as its technology threat assessment. It addresses none of the three self-interest examples the paragraph gives, because those concern the data available, the fit between the tool and this purpose, and whether anyone here can explain it. And it cannot reach the self-review limb at all, since that threat arises from what the organisation itself contributed to the system. A supplier assessment answers a procurement question.
§ 7 — Exposure
No Code fine; an unaddressed threat is pursued by the member body that adopted the Code
Section 200 applies the conceptual framework, and a failure to identify, evaluate or address a threat is a disciplinary matter for the adopting body — reprimand, fine, conditions on practice, or withdrawal of membership and the practising certificate. It reaches the accountant rather than the employing organisation, even where the threat arises out of arrangements the accountant did not choose.
§ 8 — Elsewhere
Where another instrument addresses the same obligation. These are correspondences, not comparisons — the Council does not rank one framework against another.
MAS expectations separate independent validation from peer review for models a supervised institution relies on. The self-review limb reaches the same structure through the accountant, in any organisation and whether or not it is supervised.
Finance and accounting standards
This paragraph identifies the threats that arise from having the technology in the function at all. R220.8 governs the judgement made on each occasion its output is actually used.
A correspondence indicates that two instruments address the same underlying obligation. It is not a mapping endorsed by either body, not a statement that one satisfies the other, and not a judgement about which is more demanding.
§ 9 — Where this is assessed
Examined in 2 credentials, in the domains named on each card.
Responsible use and accountability · 15% of the paper
For finance and accounting professionals who use AI in analysis, close and reporting work — and verify the output before the numbers are relied on.
Governing AI use across the finance function · 20% of the paper
For senior finance and accounting professionals who set the terms on which AI is used across a finance function — and answer for the estimates, controls and disclosures it touches.
§ 10 — provenance
This page sets out what the instrument requires and what discharges it. The official text is the authority — these go straight to it.
§ 11 — Also read
Certification
What this page sets out is examined, on work the candidate has not seen, in the credentials beside it. Every paper is machine marked against its key and every result is reviewed and released by a named person, and every credential is on the public register.
AIAC-FNAF
Certified AI in Finance & Accounting, Fundamentals2 hours · remote, proctored · valid 2 years
AIAC-FNAP
Certified AI in Finance & Accounting, Professional2 hours · remote, proctored · valid 2 years
Ten questions on using AI at work. About two minutes, free. You finish as one of six types, with a card to share.
SHARE BY 31 OCTOBER, US$199 FEE WAIVED
A free quiz from the AI Assurance Council. Not a certification.