The future of accountants
A ledger can balance and still contain a duplicate invoice or an expense in the wrong period. For accountants exploring AI, a useful starting point is a small reconciliation where you can check both the arithmetic and the evidence behind each item.
Finance. Updated
This guide focuses on business accounting: transaction review, reconciliations and period-end reporting. It draws on US occupational information and UK professional guidance; tax, audit and licensing requirements vary. Task labels are editorial estimates of potential assistance, not measured automation rates or financial advice.
AI-assisted editorial guide by Future of My Job.
How AI is changing the work of accountants
Ways to examine your own workflow, with its context and limitations in view.
Accounting combines preparation with examination and explanation. The US Bureau of Labor Statistics describes all three as part of the occupation. ICAEW identifies possible uses for generative AI such as account categorization, report drafting and help writing reconciliation scripts. These are candidates for a test, not evidence that a tool can run a complete close accurately.
Separate a suggested match from the decision to accept it. Two amounts can match while referring to different invoices; a genuine invoice can belong to another period. Before trying an assistant, specify the account, period, matching key and evidence required. Compare its exceptions with your own review, and retain the explanation for each correction.
Start with one recurring task whose answer you can establish independently. The exercise below tests duplicates, cutoff and a missing entry using a deliberately simplified rule. Measure preparation, checking and correction time together. The result can inform your next experiment, but it cannot establish how staffing, demand or your own job will change.
Reconcile the duplicate and the wrong period
Budget 30 minutes, including review and correction. All suppliers and records are invented. This is a review-skills exercise with a toy expense rule, not accounting, tax or financial advice for a real business.
| Record / supplier invoice | Amount ($) | Service completed | Posted in March ($) |
|---|---|---|---|
| A / Cedar C31 | 480 | March 28 | 480 |
| B / Cedar C31 | 480 | March 28 | 480 |
| C / Harbor H19 | 720 | April 2 | 720 |
| D / Vale V22 | 350 | March 30 | 0 |
| E / Moss M41 | 250 | March 20 | 250 |
Establish your own answer (5 min)
Sum the posted column. Apply the fictional rule to identify the expected March expense and list the differences. Keep your calculation separate from the draft you will request.
Request an exception list (10 min)
Copy the fictional table, caption and prompt into a tool your workplace allows. If no tool is approved, work through the prompt manually or compare a colleague’s independent answer.
Check the reconciliation (10 min)
Use the answer checks to verify each inclusion and exclusion, then the totals. Correct the draft and record checking and correction time as part of the total task time.
Keep a useful review note (5 min)
Record any missed exception, your final explanation and the total preparation, drafting and review time. Choose one check to retain for your next approved experiment.
A prompt to try with the fictional table
Using only this fictional table and its stated rule, show the posted March expense, the expected March expense and the adjustment needed to reconcile them. Explain each exception by record letter. A and B are confirmed copies of one invoice. Do not infer fraud or invent missing documents. Do not propose real journal entries or apply tax or accounting rules beyond the toy rule supplied.
Show the answer checks
- Posted March expense is 480 + 480 + 720 + 0 + 250 = 1,930 dollars.
- Expected March expense is one Cedar invoice (480), Vale (350) and Moss (250): 1,080 dollars. Harbor’s service falls after the stated cutoff.
- The reconciliation removes the duplicate B (−480), removes C from March (−720) and adds the omitted D (+350). The net adjustment is −850 dollars; 1,930 − 850 = 1,080.
- A and B are duplicates because the exercise explicitly confirms that fact, not merely because their amounts match. No supplied fact establishes fraud or explains how the errors arose.
- These checks concern an expense total, not cash or the payable balance. Real recognition, supporting evidence and journal entries require the applicable framework and an authorized review.
Keep the checked reconciliation and a short exception log. A useful assistant should help you explain the difference without hiding a duplicate, dropping an omission or guessing why it happened.
Get my free work snapshotAccountant tasks most exposed to AI
Our relative estimates of where drafting or other assistance may be useful. They are editorial judgments, not measured automation rates. Actual usefulness depends on the tool, inputs and checks.
Suggesting transaction categories from a defined chart of accounts
HighWhy AI affects it
A bounded classification task is a useful candidate for assistance. Test unfamiliar vendors and mixed-purpose invoices; a confident category is not evidence that the underlying expense was understood.
Proposing matches and exceptions in a reconciliation
HighWhy AI affects it
Explicit identifiers and matching rules make a draft testable. Inspect duplicate records, unmatched items and timing differences before accepting a match or clearing an exception.
Drafting report commentary from checked balances
HighWhy AI affects it
A tool may help turn supplied figures into a first draft. Recalculate comparisons and remove explanations for changes that the source material does not establish.
Preparing a recurring schedule or formula
ModerateWhy AI affects it
A defined calculation can be tested against a small known-answer case. Review dates, ranges, rounding and changes to assumptions before reusing the output in another period.
Resolving unusual items and accounting treatment
LowWhy AI affects it
An unusual contract or estimate needs evidence and the applicable accounting framework. A draft can organize questions; it does not establish the right treatment for the business.
Explaining an unresolved balance to a reviewer or manager
LowWhy AI affects it
The useful work includes identifying what is missing, finding the person who can resolve it and explaining its effect on the report. That context must be established rather than inferred from a ledger total.
What stays valuable in accountant work
Capabilities to practice and examine in your own work. None guarantees job security.
Tracing a number back to evidence
Connect a balance to the supporting record and distinguish a source fact from an assumption. A clear reconciliation lets another person reproduce your conclusion.
Spotting exceptions that a total conceals
Practice checking invoice identity, period boundaries and omissions. An arithmetic check and a completeness check answer different questions.
Applying the right framework
ICAEW advises checking the jurisdiction and accuracy of accounting, audit and tax output. Establish which requirements apply before relying on a generated explanation.
Documenting a review someone can follow
Record the rule tested, the exception found, its resolution and the remaining question. This gives a reviewer something more useful than an unexplained acceptance of the tool’s answer.
AI tool categories for accountants
Starting points for an approved experiment. Check the specific tool’s capabilities and your workplace’s data policy.
Invoice capture and categorization assistants
What it affects
A possible starting point for extracting fields or suggesting categories. Compare source records with extracted amounts, identifiers and dates.
Reconciliation and matching tools
What it affects
A way to propose matches under explicit rules. Check what happens to duplicate, missing and out-of-period items; matching software is not necessarily generative AI.
Spreadsheet and code assistants
What it affects
A way to draft formulas or reconciliation logic. Test on invented data and review the output before using it in a live workbook or ledger.
Reporting and narrative assistants
What it affects
A way to produce commentary from supplied figures. Keep the calculation and supporting explanation visible so a reviewer can check both.
Where accountant skills transfer
Options to investigate against your interests, experience and local requirements. These are not personal recommendations or promises of demand.
Accounting systems or implementation work
Why it fits
Worth investigating if mapping accounts, testing workflows and resolving data issues appeal to you. Compare real job requirements with your systems experience and project skills.
Internal audit or controls work
Why it fits
An option if you enjoy investigating exceptions and documenting evidence. Learn what experience and qualifications employers require before assuming accounting experience transfers directly.
Financial planning and analysis
Why it fits
An option if you enjoy explaining results and discussing business drivers. Explore budgeting, forecasting and communication requirements; this guide does not establish that an adjacent role is safer.
Evidence and its limits
These sources inform the guide. Our task labels and exercises are editorial interpretations, not findings validated by these researchers.
Accountants and Auditors: Occupational Outlook Handbook
US Bureau of Labor Statistics. Accessed .
Describes US accounting and audit duties, including examining records, analyzing operations and explaining findings. It supplies occupational context, not an AI accuracy benchmark or a forecast for your employer.
Generative AI Guide: Opportunities and possibilities
Institute of Chartered Accountants in England and Wales (ICAEW). Accessed .
Identifies candidate uses including categorization, document drafting and assistance with reconciliation code. Professional guidance describes possibilities; it does not measure adoption, time savings or the reliability of a particular product.
Generative AI Guide: Using generative AI
Institute of Chartered Accountants in England and Wales (ICAEW). Accessed .
Supports checking provider terms, data handling, inputs and outputs, including jurisdiction-specific accounting and tax material. UK professional guidance is not a substitute for the requirements and approvals that apply to your work.
Generative AI and Jobs: A Refined Global Index of Occupational Exposure (2025)
International Labour Organization. Accessed .
Task-based occupational research that identifies job transformation as the most likely overall effect of generative AI. Exposure does not predict an individual job loss, and this study does not validate this site’s scores or task labels.
Common questions about AI and accountants
Will AI replace accountants?
The evidence cannot answer that for an individual accountant. Task exposure research describes possible changes to work, not an employer’s hiring decision. Look at your own mix of preparation, investigation and communication, then test where a tool helps after review.
Is accounting still worth studying?
Use your interests, training costs and local opportunities to make that decision. Learn accounting principles alongside the tools: you need a basis for recognizing a plausible but incorrect result. This guide cannot promise demand, earnings or a return on a qualification.
What AI skill should an accountant practice first?
Practice writing a precise task rule and checking the result. In this guide’s exercise, invoice identity and service completion matter as much as addition. Keep a record of errors and total review time so you can judge the whole workflow.
Can I upload client invoices or a company ledger to an AI tool?
Use only data and tool features your organization has authorized. ICAEW recommends checking data handling and provider terms. Keep this experiment fictional: do not include client records, bank details, payroll data or confidential financial information.
Illustrative AI exposure profile for accountants
Illustrative: a fictional work profile run through our scoring model. This is not a measured average for the occupation.
These are self-assessment indices, not job-loss probabilities or measured percentages of tasks. How the estimate works
- Low0–34
- Moderate35–54
- High55–74
- Very high75–100
Your own task mix may produce a different result. No score can tell you when your job will change.
Five dimensions on the same 0 to 100 scale. These model indices summarize the fictional answers: relative task exposure, potential for AI assistance, demands for human input and readiness to adapt. They do not measure AI capability.
- AI exposure
- 71
- Automation exposure
- 75
- AI augmentation potential
- 75
- Human resilience
- 45
- Transition readiness
- 72
See where your own work stands
Fifteen questions about the work you actually do, scored with the same rules as the profile above. An estimate, not a prediction.
Free · ~3 minutes · No account required