Accountant / Bookkeeper

Accountants rank 4th by employment among the highest-exposure quintile in the paper's appendix, and accountants and auditors sit in its substitution rather than augmentation column. The paper says nothing about which accounting tasks moved; the link to the ledger-side tasks is this site's reading of its own task model. US private-sector payroll records from ADP covering millions of workers, November 2022 to June 2026. The 11% fall is for 22-25-year-olds in the two most AI-exposed quintiles; the same age group in the three least-exposed quintiles grew about 10%. Experienced workers show no comparable gap, and the authors state they find no evidence of widespread, economy-wide displacement. The findings are descriptive, not causal, and are measured at occupation level, not task level.

Verifiable change in hiring, headcount, hours or job scope. Highest weight — but causal attribution still has to be argued, not assumed.

The US tax authority's regulator of tax practitioners, reading existing Circular 230 duties for generative AI. It quotes the due-diligence duty in preparing, approving and filing tax returns and other papers, and says that when using generative AI practitioners must thoroughly review all AI-created documents and language before delivery to a client or submission to the IRS, verifying facts, citations and calculations, and cannot rely solely on AI. The guidance itself names no profession; Circular 230 (§ 10.3(b)) is what lets certified public accountants practise before the IRS. It binds work on US federal tax filings only, not statutory accounts or filings elsewhere, and it gives no figures on how many accountants use AI.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

One Big Four firm's own tax practice, global; a launch announcement whose scale figures (3 million tax deliverables, 30 million processes 'over the coming year') are forward-looking targets, not measured results. No headcount statement.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

Economists at Chicago Booth and Copenhagen reporting surveys run with Statistics Denmark and linked to administrative registers; the latest round, in late 2024, had 25,000 workers in 7,000 workplaces across eleven occupations, accountants among them. Asked what new tasks AI chatbots had created, accountants gave examples including reviewing AI-generated financial outputs for accuracy and completeness, and ensuring AI-driven accounting processes comply with financial regulations and standards; across occupations, 35% of new AI-related work was quality review and ethics and compliance. It is about chatbots, not ledger automation, in Denmark only; the share of accountants reporting new tasks is shown only in a chart; and it is a working paper, not peer-reviewed.

Measured, large-scale use of a tool for real work, where the decision to use it was the worker's rather than an employer's. It is more than a capability record — the work is real, not a demo — and less than a deployment record, because no employer put it into production, required it, or built a process around it. Weighted `cautious`: `automating` means the machine can do the task AND there are adoption signs, and this is an adoption sign — but usage can be experimental, and much of the measurement comes from a party with a stake, so one record is never enough and two independent ones are. Note who is counting. Vendor telemetry sees this directly and sells the tool, so such a record names that stake in its scope; a statistics agency asking firms whether their workers use AI in tasks sees the same channel with no stake at all, and that is the better source where it exists.

A ministry rule applying to government bodies, companies and all other organisations that keep accounts with information systems. Article 17 allows vouchers generated automatically with clear review rules to be reviewed by the system, requires vouchers not reviewed automatically to be reviewed by a person before further processing, and requires the automatic review rules to be queryable, verifiable and traceable, with their setting and changes going through a sign-off procedure and kept on file. Article 6 makes the head of the organisation first responsible for accounting informatisation; Article 22 requires organisations with the means to automate accounting work. The first two sentences of Article 17 already appeared in the 2013 rule for enterprises; the sign-off on rules is new. It governs how automated accounting is controlled, not how many accountants do it.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

An entry the US Department of Health and Human Services filed in the federal government's mandatory inventory of AI use cases, for the Centers for Disease Control and Prevention's financial office. It lists the use as deployed since May 2024 on a commercial automation platform, and says the invoice mailbox bot processes more than 15,000 emails a year and the purchase-card line-item review bot processes 3,000 to 5,000 transactions a month against more than 200 business rules, with output going for further processing by staff for payment. It is the agency's own description; the operational date is given as a month only, and it does not say how many staff hours changed.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

The global standard setter's revisions to the International Code of Ethics for Professional Accountants, adopted by member bodies that make the Code binding on their members (Singapore's institute, for one, adopted the same text effective 15 December 2024). R220.8 says an accountant in business who intends to use the output of technology, whether developed internally or by third parties, shall exercise professional judgment to determine the steps needed to meet the duty to prepare and present information properly; R320.11 requires an accountant in public practice to determine whether such use is appropriate; and the guidance lists automation bias — favouring automated output even when reasoning or contradictory information questions it — among the biases to guard against. It governs how accountants use technology; it does not measure how much of their work is now reviewing machine output.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

The traditional way in — doing entry and reconciliation until you understand the business — is the part being automated. You will need to reach judgement work faster than the previous generation did, which usually means seeking exposure to advisory and exception cases early rather than waiting to be given them.

Your exposure depends less on your title than on how your week is actually spent. If most of your hours go to production rather than judgement, that is the number to change. Supervising automation is a real and defensible role — but only if you can demonstrate you catch things it misses.

Stay and strengthen

Shift your hours from production to judgement

The tasks holding up this role are advisory and exception judgement. Hours spent there are the ones hardest to remove.

Reshape the role

Own the automation, don't just use it

Someone has to choose the tools, define the controls and be accountable when output is wrong. That role is being created faster than it is being filled.

Adjacent move

Financial analysis / FP&A

Reuses the same domain knowledge but sits on the forward-looking side, where the work is framing questions rather than recording answers.

Cross-industry

Audit, controls or compliance in another sector

Control thinking transfers across industries better than industry knowledge does. Regulated sectors reward it and are slower to automate the accountable layer.

Not as a whole. Specific tasks — entry, coding, reconciliation — are being automated fast because they are structured and checkable. Advisory, ambiguous treatment and accountable sign-off are not, because they need business context and someone who carries the consequence. The honest question is not whether the job disappears, but what fraction of your week is the automatable part.

The training is still valuable — understanding how money moves through an organisation transfers to finance, operations, controls and founding a company. What has changed is that the traditional entry path, learning by doing routine work, is narrowing. Plan to reach judgement work faster, and treat the qualification as a foundation rather than a destination.

The signal in this job is your firm's ratio of preparers to reviewers. When a practice that ran three preparers per reviewer runs one, the shift has happened — and that change is visible in who gets hired next, well before anyone announces anything.

Transaction entry and reconciliation — the two tasks marked automating here, and historically the hours that filled a junior's week. Advisory and ambiguous-case judgement are not, but they sit in the senior half of the profession, and a junior whose entry work disappeared does not automatically arrive there.

It protects the signature, not the hours behind it. Expect the same filings still to need a named person while taking far fewer hours each — that is a change to how the work is priced more than to whether the qualification matters. It remains worth having and it is not a shield.