Accounting software that uses AI mostly takes over routine work such as data entry, matching, and first drafts, while the accountant stays responsible for the result. The US Bureau of Labor Statistics (BLS) projects employment of accountants and auditors to grow 5% from 2025 to 2035, against a 6% decline for bookkeeping, accounting, and auditing clerks. The split suggests that exposure to automation depends on the task, not on the word "accounting."
Key takeaways
- BLS projects accountant and auditor jobs to grow 5% from 2025 to 2035 and expects clerk jobs in the same field to fall 6%.
- BLS expects automation of routine tasks to make advisory and analytical duties more prominent, not to cut overall demand for accountants.
- A July 22, 2024 PCAOB staff report found audit firms using generative AI mainly for administrative and research work.
- CPA Ontario states that a CPA remains solely responsible for the quality of work done with AI, as with any other software.
- This page is general information, not accounting, tax, legal, or investment advice.
What the labor data say about accountants and clerks
The BLS Occupational Outlook Handbook lists about 1,595,200 accountants and auditors in 2025, with a median wage of USD 83,680 in May 2025. It projects about 115,300 openings a year on average over the decade, many of them to replace people who leave the occupation. Its Job Outlook text says that some routine accounting tasks may be automated as cloud computing, AI, and blockchain spread, and that although these technologies will likely raise efficiency, the change is not expected to reduce overall demand.
The handbook entry for bookkeeping, accounting, and auditing clerks reads differently. It projects a 6% drop from 2025 to 2035, a loss of about 85,600 jobs from a base of about 1,532,400, and states that technological change is expected to reduce demand for these workers because software has automated many of their tasks. The median wage for clerks was USD 50,670 in May 2025. These are national projections for the United States. They do not predict what happens to any one employer, and Canadian figures were not used here.
Tasks that software handles and tasks that stay with people
The table is an illustrative example of how the division of work is commonly described, not a measurement. It follows the logic of the BLS text: repetitive, rule-based steps are the ones most open to software, and judgment steps move toward the accountant.
| Task | Typical role of software | What a person still decides |
|---|---|---|
| Entering invoices and receipts | Reads documents and proposes account codes | Corrects miscoded or unusual items |
| Bank reconciliation | Matches transactions by rules or learned patterns | Resolves breaks that do not match |
| Drafting memos or client emails | Produces a first draft from notes | Checks every statement before it is sent |
| Research on a standard or tax rule | Summarizes and points to sources | Confirms the answer in the authoritative text |
| Audit sampling and risk flags | Scans full populations for outliers | Investigates flagged items and concludes |
| Advice to a client | Prepares scenarios and charts | Applies judgment about the client's situation |
What audit regulators have observed
On July 22, 2024, staff of the Public Company Accounting Oversight Board (PCAOB), the US audit regulator, published a report on outreach to audit firms about generative AI. The staff spoke with global network firms and several other firms that audit more than 100 issuers. They found that use in audits centered on administrative and research activities, although most firms saw potential in parts of audit planning and performance. Firms also pointed to the need for strong supervision of the tools, including for data privacy and security. The document states that it reflects staff views, not a rule or policy of the Board.
No PCAOB standard written specifically for AI was found when this page was checked in October 2026, and the existing standards still apply to the audit work itself. Public company preparers were described in the same report as exploring generative AI in accounting and reporting, but as giving it lower priority than operational and customer-facing uses.
Professional responsibility in Canada
CPA Ontario, the provincial regulator, published guidance in 2024 stating that the obligations CPAs have when using any software also apply when using AI, and that the CPA continues to hold sole responsibility for the quality of the work performed. In practice that places the duty on the person who signs. A generated reconciliation, tax summary, or disclosure draft has the same status as one produced by a junior colleague: it needs review before it is relied on.
Confidentiality is the other recurring point. Client records entered into a public chatbot leave the firm's control, so firms typically set rules on which tools may receive client data. Those rules come from each firm and each provincial or state body, and readers should check their own regulator's current text.
Questions to ask about an AI feature in accounting software
- Which steps does the tool perform on its own, and which does it only suggest?
- Is there a log showing the input, the output, and who approved it?
- Where is client data stored, and is it used to train the vendor's models?
- How are errors found, for example by sampling a share of automated entries each month?
- Who inside the firm owns the review, and is that person trained to challenge the tool's output?
The list is a planning aid built from the regulator themes above. It does not certify any product.
Information, not advice
This article describes labor statistics and regulator statements as of October 2026. It does not recommend software, a career path, or a financial decision, and projections can be revised. Anyone choosing tools for a business or practice should consult a licensed accountant and the rules of the relevant regulator.
FAQ
Will AI replace accountants?
The BLS does not project that. It expects accountant and auditor employment to grow 5% from 2025 to 2035, while expecting routine clerk roles to shrink. Projections can change, and individual jobs may shift even when the total grows.
Why do clerks and accountants have different outlooks?
Clerk work consists largely of repetitive entry and matching, which software has already automated. Accountant work includes advisory and analytical duties that the BLS expects to become more prominent.
Are audit firms using generative AI?
According to the 2024 PCAOB staff report, the firms interviewed used it mainly for administrative and research work and saw potential for more. That report predates later developments, so current practice may differ.
Who is responsible for a mistake made by an AI tool?
Under CPA Ontario's 2024 guidance, the CPA keeps sole responsibility for the quality of the work, whatever tool was used. Contract terms with a software vendor are a separate legal matter.
Related reading
- How AI differs from plain automation
- BlackRock and AI in financial technology
- AI fintech companies in Canada
- AI in human resource management
More in AI in finance.
Last verified: October 2026
Sources
- Accountants and Auditors, Occupational Outlook Handbook, US Bureau of Labor Statistics, 2025 data, projections 2025 to 2035
- Bookkeeping, Accounting, and Auditing Clerks, Occupational Outlook Handbook, US Bureau of Labor Statistics, 2025 data, projections 2025 to 2035
- Staff Update on Outreach Activities Related to the Integration of Generative Artificial Intelligence in Audits and Financial Reporting, PCAOB, July 22, 2024
- Accountabilities for CPAs in the Age of Artificial Intelligence, CPA Ontario, 2024