Basepoint

Journal / Insights

Is AI replacing bookkeepers?

AI is shrinking bookkeeping jobs, not ending them. What BLS data shows, what AI does well, and where a person still checks the work.

BBranson·Oct 30, 2026·4 min read

Quick answer: Not outright, but it is shrinking the job. The US Bureau of Labor Statistics (BLS) projects employment of bookkeeping, accounting, and auditing clerks to fall 6 percent between 2025 and 2035, from 1,532,400 jobs to 1,446,800, and names software automation as the main cause. BLS still expects about 144,100 openings a year over that decade, all from people who retire or move to other work.

What the numbers say

The clearest data comes from the BLS Occupational Outlook Handbook, last updated in August 2026. It counted 1,532,400 bookkeeping, accounting, and auditing clerks in the US in 2025, with a median pay of $50,670 a year. By 2035 it expects 85,600 fewer of those jobs.

BLS is direct about why. "Technological change is expected to reduce demand for these workers," the handbook says, and it adds that software has automated many of the tasks these clerks used to do by hand.

The global picture points the same way. The World Economic Forum's Future of Jobs Report 2025, built on a survey of more than 1,000 companies across 55 economies, lists accounting, bookkeeping and payroll clerks among the fastest declining jobs to 2030. CPA Australia, reading the same report, notes the role ranks 7th on that list.

So the honest answer is that fewer people will do bookkeeping as a full time job. But a 6 percent drop over ten years is a slow squeeze, not a cliff. The work itself is not going away. Every business still needs clean books, and BLS still sees about 144,100 open roles each year as people leave the field.

What AI does well in bookkeeping

We build bookkeeping agents, so we see every day which parts of the job a machine handles well. The pattern is consistent. AI is good at work that is high volume, rule based, and repetitive:

  • Reading invoices and receipts from an inbox and turning them into bills in Xero or QuickBooks.
  • Matching bank feed lines to invoices and bills during bank reconciliation.
  • Suggesting the right account code for a transaction based on how similar ones were coded before.
  • Sending polite reminders on overdue invoices and logging the replies.

Research backs this up. A study by Jung Ho Choi of Stanford and Chloe Xie of MIT, published in the Journal of Accounting Research, looked at 277 accountants and over 200,000 transaction records from 79 small and mid-sized businesses. Stanford's summary of the work reports that accountants using generative AI finalized monthly statements 7.5 days faster, spent 8.5 percent less time on routine back-office processing, and produced books with a 12 percent rise in reporting granularity.

That last point matters. Faster books are nice. More detailed books are what let an owner actually see where the money goes.

What AI still gets wrong

The same study is also a useful warning. Accountants tended to step in when the AI was unsure, and following AI suggestions they disagreed with raised the risk of errors. The authors conclude that AI works best as a tool that augments professional judgment rather than one that replaces it.

That matches what we see in client books. The hard cases are predictable:

  • A supplier invoice that could be a capital purchase or an expense.
  • One payment that covers three invoices, minus a credit note nobody recorded.
  • A bank line with a vague description and no paperwork anywhere in the inbox.
  • Anything that touches tax treatment, payroll, or a judgment call about the business.

An agent can spot these. It should not guess at them. A wrong account code that nobody checks is still wrong at month-end, and it is harder to find three months later. This is why every setup we run has a person who approves anything the agent is unsure about.

What this means for bookkeepers and business owners

For bookkeepers, the job is moving up a level. BLS expects the people in these roles to do more analysis and advisory work, such as reviewing a client's books for ways to run more efficiently. The person who knows why a number looks wrong is worth more than the person who typed it in. Bookkeepers who learn to supervise automated tools, review exceptions, and explain the numbers to owners are in a better spot than those who compete with software on data entry.

For business owners, the takeaway is practical. You probably do not need to replace your bookkeeper. You can likely stop paying for hours of data entry, chasing, and matching, and spend that budget on review and advice instead.

That is the setup we build at Basepoint. Our AI accountant handles bookkeeping, accounts payable, invoice chasing, and bank reconciliation inside Xero, QuickBooks, and Gmail. It acts on what it is sure about and sets anything uncertain aside for a person to confirm.

Sources

B

Branson Tiong

Founder of Basepoint Labs

Get the next post by email

Monthly email. Unsubscribe any time.

Keep reading

All posts →