Accountancy Was Supposed to Be Automated by Now

A decade ago, accountancy featured on every list of professions about to be dismantled by software. The reasoning looked sound: the work involved structured data, defined rules, and repetitive processing, which is precisely what machines handle well. What happened instead is more interesting than either the prediction or the profession’s initial defensiveness about it. Enormous amounts of accountancy work genuinely have been automated, and the number of people working in the sector has not collapsed. Understanding why says something useful about how automation actually moves through knowledge work.

What the Software Actually Took

The tasks that disappeared were the ones nobody enjoyed. Manual data entry has largely gone, replaced by bank feeds and document capture that reads an invoice and codes it. Reconciliation, once a genuine time sink, is now mostly a matter of reviewing suggested matches. Compliance filings are prepared from structured data with far less human handling than a decade ago. What these have in common is that the input was already structured, the rules were stable, and the output was checkable. That combination is where automation delivers, and where it has delivered comprehensively. The regulatory environment accelerated it as well. Digital filing requirements pushed businesses towards software that produces machine-readable records as a by-product of ordinary bookkeeping, which handed the profession a far cleaner set of inputs than it had previously worked with. Automation tends to arrive where somebody has already done the unglamorous work of standardising the data, and in this case a compliance mandate did much of that work on the sector’s behalf.

The Work That Did Not Move

Meanwhile, a large category of work has proved stubbornly resistant, and it is not the technical end. Advising an owner on whether to sell now or in three years, structuring a transaction around a family’s actual circumstances, judging how aggressive a position can defensibly be, telling a client something they do not want to hear: none of this is a data-processing problem. It depends on context the software does not have, on judgement about ambiguous rules, and on a relationship that carries accountability. The pattern is consistent across professions, where automation absorbs the tasks with clean inputs and leaves the ones requiring interpretation and responsibility.

Firms Repositioned Rather Than Shrank

The commercial response has been a shift in what practices sell. When compliance work becomes faster to produce, its price falls, and a firm whose income depends on volume compliance faces a genuine problem. The ones that adapted moved up the value chain into advisory work, corporate finance, tax planning and sector specialisation, using the time released by automation rather than simply pocketing the efficiency. Price Bailey operates across that broader advisory range rather than positioning themselves solely around statutory filing, which is roughly the direction the whole profession has been travelling since the software arrived.

Data Quality Became the Bottleneck

An unglamorous consequence of automation is that the constraint moved. When a human was handling every transaction, errors got caught along the way by someone who noticed something looked odd. Automated pipelines process bad data as efficiently as good data, which means a misconfigured feed or an incorrectly mapped account can produce months of plausible-looking nonsense. Firms have consequently invested in review and exception handling rather than processing, and the skill that matters has shifted from doing the work to knowing what a wrong answer looks like. That is a genuinely different capability, and it is harder to teach. It has complicated training as well. Junior roles were historically built around processing work, and that repetitive exposure was how newcomers developed a feel for what normal looks like in a set of accounts. With much of that work now automated, firms have had to find other ways to build the same instincts, which is one of the less discussed consequences of the shift and a problem the profession has not entirely solved.

Clients Now Expect Different Things

Automation changed expectations as much as it changed delivery. A business owner who can see their position in real time on a dashboard is not satisfied by a set of accounts arriving nine months after a year end to tell them what already happened. The demand has moved towards forward-looking work: forecasting, scenario modelling, and advice on decisions that have not been made yet. That suits the firms who built the capability and squeezes those who did not, and it explains why the profession’s headcount held up even as the underlying tasks were automated away. The work changed rather than vanished.

Regulation Kept Its Own Pace

Technology has not removed the regulatory framework around this work, and in some respects has sharpened focus on it. The Financial Reporting Council sets and enforces standards for accountants and auditors in the UK, and questions about the use of technology in audit, the quality of evidence it produces, and where professional scepticism must still be applied have become a live part of that oversight. Automation does not transfer responsibility to a vendor. A signature on a set of accounts still belongs to a person, which places a firm limit on how far the process can be handed over regardless of how capable the tools become.

The Next Round Will Look Similar

Generative AI is now prompting a repeat of the earlier conversation, with similar confidence and similar imprecision. The likely outcome resembles the last one: substantial gains in drafting, research and first-pass analysis, minimal change to the parts that involve judgement, accountability and a client who needs to trust the answer. Professions rarely disappear when their tasks are automated. They reorganise around whatever is left that a machine cannot take responsibility for, and in this case that turns out to be a fairly durable core.