After Programmers, AI Is Coming for Accountantseric

The first wave of AI anxiety arrived in software development. That makes sense: code is structured, digital and unusually easy for a machine to generate, test and revise. The next wave may be quieter, but just as important. It is arriving in accounting, especially in the part of accounting that looks like collecting documents, classifying transactions and completing familiar forms.

That does not mean accountants disappear. It does mean some accountants' jobs are going to be cut, and many others are going to change.

The return that used to need a professional

In Australia, individuals have long been allowed to prepare and lodge their own tax returns. The ATO's myTax service is available to people who want to lodge for themselves, with much of the information pre-filled from employers, banks, health funds and other institutions. The ATO describes myTax as available to individuals and sole traders who want to lodge their own return.

The permission existed before generative AI. The difficulty was confidence.

Many people could technically do their own return, but did not know what a label meant, which records mattered, or whether an apparently small detail changed the result. Paying an accountant was partly payment for expertise, and partly payment for reassurance.

AI changes that balance. A person can ask, in plain language, what a field means. They can provide a folder of statements and ask for a checklist. They can compare a draft against last year's return, find a missing category, or turn a pile of receipts into a structured record. The work becomes less mysterious and less intimidating.

Nobody knows your own situation better than you do. If you do not know that a particular payment or expense exists, your accountant cannot help you claim or report it. I did not know that income protection premiums could be relevant to my tax return until I looked more closely. An AI assistant that scans an organised list of expenses may be able to flag that kind of possibility, along with questions to take to an accountant. It is not a substitute for checking the rules, but it can help people notice what they would otherwise miss.

Privacy is a legitimate concern. Before sending records to an AI service, taxpayers can anonymise or redact names, addresses, account numbers, tax file numbers and other identifiers, and keep the original documents securely. They should also understand the provider's data-retention and training settings. Anonymisation reduces exposure, but it does not remove the need to use a service they trust or to review what is being uploaded.

That is enough to move a large amount of routine work away from professional firms.

Single returns are the obvious first target

For a single person with one employer, a bank account and a handful of ordinary deductions, the job is often relatively constrained. There may be judgment involved, but there are fewer relationships between the facts.

AI is particularly useful here because it can explain the process step by step without making the taxpayer search through a dense help page. It can ask the questions a careful preparer would ask:

• Was this expense paid personally?
• Was it reimbursed?
• What percentage was work-related?
• Do you have a receipt or another record?
• Is this income, a deduction, or a capital transaction?

The taxpayer still carries responsibility for the answer. But the barrier to getting to a sensible answer is much lower.

Families are where the complexity starts to compound

Couples and families are different. A return is no longer an isolated document. One person's taxable income can affect the other person's spouse questions, offsets and eligibility calculations. Children, childcare, family payments, private health cover, super contributions and income tests can all connect the returns.

This is where an AI assistant can be more valuable than a simple tax calculator. It can maintain a shared timeline, remember which person paid an amount, keep a record of carried-forward losses, and show how one figure flows into another return. It can turn “What do I put here?” into “This number comes from Eric's final return; this other number comes from Min's income test.”

That assistance does not make the tax law simple. It makes the complexity navigable.

It also creates a new risk: a fluent answer can still be wrong. AI can confuse taxable income with assessable income, treat a capital loss as an ordinary deduction, or confidently classify a transaction without enough facts. A good system needs source records, explicit assumptions and a final review, not just a polished paragraph.

What gets cut, and what remains valuable

The most exposed work is repetitive preparation: chasing standard documents, entering data, checking arithmetic, copying figures between related forms and answering the same basic questions. Firms that needed several people for this work may need fewer people, or may redirect them to more clients.

The valuable work moves upward. Accountants will still be needed for business structures, trusts, companies, international income, complicated capital transactions, audits, disputes and situations where the cost of an error is high. They will also be needed when a client does not merely need an answer, but needs someone to take professional responsibility for a judgment.

In other words, AI removes some preparation labour while increasing the premium on interpretation, review and accountability. That is the same pattern already visible in programming: fewer hours spent typing the obvious code, more time spent deciding what should be built and proving that it works.

The new middle ground

The interesting future is not “everyone fires their accountant” or “AI changes nothing”. It is a much larger middle ground.

Some people will still outsource everything. Some will do everything themselves. Many will prepare a first draft with AI, then pay an accountant for a targeted review. Others will use an accountant once a year and an AI organiser throughout the year to keep records tidy.

That is a meaningful change in the market. The professional relationship becomes less about being the only person who can operate the form, and more about being the person who can recognise the edge of the form.

Accountants' jobs are therefore genuinely being cut by AI, after programmers' jobs began to be reshaped by it. But the deeper story is about access. More people can now participate in their own financial administration, ask better questions and understand the consequences of their choices.

The safest version of that future pairs AI's patience and organisation with human judgment where the facts stop being routine.

The next post will look at what an AI-assisted tax workflow should record so that convenience does not come at the cost of an audit trail.