Will AI replace accountants? What the data says in 2026

No, AI is not going to replace accountants, but it is already replacing a large share of accounting work. The U.S. Bureau of Labor Statistics (BLS) puts accountants and auditors in its highest AI exposure category, yet it still projects their jobs to grow 5% by 2035 while bookkeeping clerk jobs shrink.

That split is the whole story. AI is very good at the recording, matching and first-draft parts of accounting, and the jobs made mostly of those tasks are the ones in decline. It is much weaker at the parts that need judgment, a client’s trust and someone willing to put their name to the numbers, and that is where accountants are heading. Below are the official numbers, the tasks AI already handles, how exposed different accounting roles are, and what to do if you work in the field.

The numbers: very high exposure, and still growing

In August 2026 the BLS published official AI exposure ratings for 831 occupations for the first time, alongside its 2025–35 job projections. It combines five research measures, including real-world AI usage data from Anthropic and Microsoft, and places every occupation in one of four groups (low, moderate, high or very high exposure) compared with other occupations.

Accountants and auditors land in the very high group, along with about a quarter of all occupations. So does nearly every job in finance, including all seven below. But look at how the same agency expects the number of these jobs to change between 2025 and 2035, next to their median pay in 2025:

Same exposure label, opposite directions. The two biggest groups are almost the same size: about 1.6 million accountants and auditors and 1.5 million bookkeeping, accounting and auditing clerks in 2025. The jobs that advise, examine and decide are growing faster than the 3% average for all occupations. The jobs that record and process transactions are flat or shrinking. The BLS says this directly in its outlook for accountants: AI and other automation will take over routine tasks such as data entry, but this “is not expected to reduce overall demand.” Instead, it expects the advisory and analytical side of the job to become more prominent.

Demand is not small, either. The BLS expects about 115,300 openings for accountants and auditors every year through 2035, many of them to replace people who retire or change careers.

Will AI take over accounting?

It is already taking over much of the routine work, and that is the part of the profession that is shrinking. The BLS’s outlook for bookkeeping, accounting and auditing clerks says software has automated many of their tasks, so the same work gets done by fewer people. It projects 85,600 fewer clerk jobs by 2035. Payroll and timekeeping clerks face the steepest drop in the field, at nearly 16%.

So if “accounting” means keying in invoices, matching bank lines and rebuilding the same month-end schedules, then yes: software and AI will take most of it over. That process started with accounting software decades ago, and AI models are speeding it up.

But the profession is much bigger than that. Someone still has to decide how an unusual transaction should be treated, explain the numbers to a board, spot the fraud that looks normal, plan a client’s taxes and sign the accounts. The BLS notes that publicly traded companies must have CPAs sign the reports they file with the SEC. That responsibility can’t be handed to software.

Which accounting tasks AI can already do

Two large studies measured how people actually use AI at work, instead of guessing.

Microsoft analyzed 200,000 anonymized conversations with its Copilot assistant and scored U.S. occupations on how much of their work AI was used for, and how well it did. Accountants and auditors scored 0.19, higher than about two in three of the 785 occupations in its published data. Bookkeeping clerks scored 0.24, in the top quarter, and tax preparers 0.31, in the top 10%.

The more useful finding is where AI did well. The researchers found Copilot stronger at the writing and research side of knowledge work than at analysis. In their data, AI completed the task in about two-thirds of conversations about calculating financial data, against more than 90% of conversations about advising on financial or business matters. For accountants, that means AI is already good at drafting memos, summarizing standards and explaining a tax rule, and still needs checking when it does the actual numbers.

Anthropic measured which tasks its Claude models are actually used for at work, giving full weight to fully automated use and half weight to use that only assists. Financial and investment analysts were among the ten most exposed occupations, with 57% of their tasks covered; the most automated was analyzing financial information to forecast business conditions. Accountants did not make that top ten. Across the whole economy, Anthropic found no systematic rise in unemployment among highly exposed workers since late 2022.

Put together with the task ratings our own test uses, accounting work splits into four layers:

• Most exposed: data entry, coding transactions, reconciliations and routine reports. These are rules-based and repetitive. Our model rates admin and data entry at 90% exposure.
• Heavily assisted: memos, client emails, policy summaries and first-pass analysis. AI writes a decent first version in seconds, and a person still has to check it.
• Partly exposed: calculations, forecasts and variance analysis. AI speeds them up but makes confident mistakes, so review matters more, not less. Our model rates numbers and analysis at 70%.
• Least exposed: judgment calls, client advice, audit sign-off, fraud investigation and leading a finance team. These depend on accountability, trust and context. Our model rates leading and deciding at 30% and face-to-face work at 25%.

How exposed is your accounting job?

Job titles hide big differences. We scored typical task mixes with exactly the same method our free test applies to a real person’s week. A typical bookkeeper, whose week is mostly data entry and routine numbers, scores 84. A typical accountant, with more analysis and some client contact, scores 72. Both sit in our “very high” band, next to data entry clerks and paralegals.

Now compare that with the advisory accountant in our careers index, someone who helps businesses make decisions with their numbers rather than producing them. The exposure drops to 60. A compliance and risk officer scores 42 and a financial planner 38. The further you move from recording numbers toward advising, deciding and signing, the lower the score.

The same logic applies inside a single job. Picture two accountants with the same title at the same firm. One spends most of the week on reconciliations and month-end schedules. The other spends half of it with clients, explaining results and planning the next year. The first is far more exposed than the second, and the gap will widen as AI improves.

Are accounting jobs safe from AI?

Some are much safer than others. Based on the BLS projections and the task research above, here is how the field splits.

Most at risk: bookkeeping, payroll processing, transaction entry and routine billing. These roles are made mostly of the tasks software and AI handle best, and the BLS already projects bookkeeping and payroll clerk jobs to shrink. Simple tax return preparation is also highly exposed, although the BLS still expects tax preparer jobs to grow 4.7%.

Safer: roles built on judgment and accountability.

• Audit and assurance, where a licensed person signs the opinion. Financial examiners, who check that banks and other institutions follow the rules, are projected to grow 9.3%.
• Tax planning and advisory work, as opposed to typing figures into returns.
• Forensic accounting, which means investigating fraud and often testifying as an expert witness.
• Controller and finance leadership roles. Financial managers are projected to grow 9.7%, with a median pay of $166,570.
• Compliance and risk, where more AI means more rules for someone to check and sign off.

These three careers in our index show where accountants can move without starting over. One honest note: the advisory accountant ranks last of our 50 careers on AI resilience, which is why it carries the No. 50 label. Even the advisory end of accounting involves a lot of screen-based analysis. Financial planning and compliance hold up better because more of the work is trust and sign-off.

So is accounting still a good career? For most people, yes. Accountants and auditors earned a median $83,680 in May 2025, against $50,980 for all occupations, and the field is growing faster than average. What is changing is the kind of accountant employers need.

The real risk: fewer entry-level jobs

The biggest change AI brings to accounting may be fewer junior roles rather than layoffs. Much of the work AI does well, such as reconciliations, sample testing and first drafts, is exactly the work that used to train new staff.

The research points the same way. Stanford’s Digital Economy Lab found that employment of workers aged 22 to 25 in AI-exposed occupations is now 19% below where it would be had it kept pace with their less-exposed peers, while experienced workers show no comparable gap. Anthropic found that young workers’ chances of being hired into exposed occupations fell about 14% compared with 2022, a result it calls only just statistically significant, with no similar drop for workers over 25.

If you are studying accounting or in your first few years:

• Learn to review AI work, not just produce work. Firms need juniors who can check an automated reconciliation and explain why a number looks wrong.
• Get close to clients and decisions early. Ask to join client meetings, write the commentary on the monthly pack and present it yourself.
• Aim for a license that carries responsibility. Credentials such as the CPA hold their value because accountability can’t be delegated to software.

How to AI-proof your accounting career

The same principle applies at any stage: let AI do the production and take ownership of the judgment. Here is what that means for each task area our test measures.

For accountants, five moves matter most:

1. Automate your own routine work first. Pick the reconciliation or report you rebuild every month and get AI or your accounting software to do the first pass. The person who set up the automation and checks it is in a far stronger position than the person it replaced.
2. Add a recommendation to every report. Don’t stop at what the numbers are. Say what they mean and what the business should do next. Turning this month’s report into three recommendations is a good start.
3. Own the sign-off. Volunteer for the reviews, approvals and filings that carry your name. Accountability is the one thing AI can’t take on.
4. Spend more time with people. Client meetings, budget negotiations and explaining results to non-finance managers are among the least exposed tasks in any job.
5. Specialize in the messy cases. Complex tax planning, acquisitions, fraud and new regulations. AI handles the standard case; people get paid for the exceptions.

For a full 90-day plan, see our guide to the skills AI can’t replace. If you are thinking about leaving accounting altogether, our ranking of the 25 most AI-proof jobs shows where low exposure meets good pay.