ChatGPT and Claude prompts for accountants, bookkeepers and CFOs work best when they name the role, paste the real figures, ask for one output and demand the workings. That is the pattern behind every prompt in this guide. If you have been searching for "ai for accountants" or an "accounting gpt" and found only vague examples, this is the practical version: 25 tested prompts across bookkeeping and close, financial statements, cash flow forecasting, variance analysis and communication to leadership and clients, plus a cash flow forecast template and a month-end close checklist you can copy into your own files.
One line before we start: nothing here is financial, accounting or tax advice. These prompts produce drafts and analysis for a qualified accountant to review. Every figure must be checked before it is filed, sent to a lender or shown to a client.
AI bookkeeper skill
Selma - Bookkeeper AI Skill
Loads a standing bookkeeping process into Claude or ChatGPT: categorisation rules, reconciliation checks and a month-end close routine, so you stop re-typing the prompts below.
$29
Get the skill →
How to use these prompts
Each prompt below follows the same shape: a role ("act as a reviewing accountant"), the context ([paste trial balance]), one task, a fixed output format and an instruction to flag assumptions. Replace the [brackets] with your own data. The prompts work in Claude, ChatGPT, Gemini and Copilot. If you find yourself re-typing the same context every day (your chart of accounts, your reporting format, your materiality threshold), that is the point at which a reusable skill beats a prompt. We cover that in the task table near the end.
Two habits make a big difference. First, ask for the mapping or the workings every time, so a reviewer can check the output in minutes rather than redoing it. Second, tell the model to separate what it calculated from what it inferred. An AI bookkeeper that hides its guesses is more dangerous than no bookkeeper at all.
Bookkeeping and month-end close prompts
These are the prompts that take the typing out of the close. Paste real ledger data, ask for one output and always ask for the workings.
Prompt 1: Categorise a bank feed
Act as a bookkeeper for a [industry] business using [accounting software]. Here is our chart of accounts: [paste chart of accounts]. Categorise each of the following bank transactions to the most likely account, add a confidence level (high, medium, low) for each, and list separately every transaction where you need more information from me. Transactions: [paste bank lines with date, description, amount].
Prompt 2: Reconciliation breaks
You are reconciling our [bank account name] for [month, year]. Book balance per the ledger: [amount]. Statement closing balance: [amount]. Here are the ledger entries: [paste]. Here are the statement lines: [paste]. Match the items, list every unmatched item on either side with a likely explanation (timing, duplicate, missing entry, fee), and show the reconciling items that bring the two balances together.
Prompt 3: Accruals and prepayments
Review these invoices and contracts and tell me which need an accrual or a prepayment adjustment at [month-end date]. For each one, give the calculation, the journal (debit and credit, accounts from this list: [paste relevant accounts]) and the reversal date. Show your workings so a reviewer can check them. Documents: [paste invoice details with service periods and amounts].
Prompt 4: Fixed asset and depreciation check
Here is our fixed asset register: [paste asset, cost, acquisition date, useful life, method, accumulated depreciation]. Calculate the depreciation charge for [month] for each asset, flag any asset that appears fully depreciated but still carrying a charge, and flag anything acquired this period that has no depreciation started. Present as a table with a total that I can post as one journal.
Prompt 5: Month-end close review
Act as a reviewing accountant. Here is our month-end close checklist with statuses: [paste]. Here is the draft trial balance: [paste]. Identify anything on the checklist that is incomplete or inconsistent with the trial balance (for example a balance sheet account with an unusual balance, suspense accounts not cleared, intercompany not agreed). Rank the issues by risk of misstatement and tell me what evidence I should attach for each before sign-off.
Reporting and financial statement prompts
A financial statement generator prompt is only as good as the trial balance you paste in. Insist on a visible account mapping and on the model saying when a figure is missing rather than filling it in.
Prompt 6: Draft P&L from a trial balance
Using this trial balance for [period]: [paste accounts and balances], build a profit and loss statement in this structure: revenue by stream, cost of sales, gross profit and margin, operating expenses grouped as [your groupings], EBITDA, depreciation and amortisation, interest, tax, net profit. Map each trial balance account to a line and show me the mapping separately so I can check it. Do not invent any figure not present in the trial balance.
Prompt 7: Balance sheet with checks
Build a balance sheet as at [date] from this trial balance: [paste]. Group into current assets, non-current assets, current liabilities, non-current liabilities and equity. Confirm that total assets equal total liabilities plus equity, and if they do not, tell me by how much and which accounts are most likely unmapped. Add a short note on any account with a balance sign that looks wrong (for example a credit balance in receivables).
Prompt 8: Indirect cash flow statement
Prepare an indirect-method cash flow statement for [period] using the opening balance sheet [paste], closing balance sheet [paste] and the P&L for the period [paste]. Show operating, investing and financing sections, reconcile net profit to operating cash flow with each working-capital movement listed, and confirm that the closing cash agrees to the balance sheet. Flag any movement you had to assume.
Prompt 9: Management accounts pack
Turn these finance outputs into a four-page monthly management accounts pack for [company] for [month]: page 1 headline numbers and three sentences on the month, page 2 P&L versus budget and prior year, page 3 balance sheet and cash, page 4 KPIs [list KPIs]. Data: [paste]. Write the commentary in plain English for a non-finance owner, and mark every place where you need me to add context you cannot see in the numbers.
Prompt 10: Notes and disclosures draft
Draft the notes to the accounts for [entity] for the year ended [date] under [GAAP / IFRS / local framework] for these areas: accounting policies, revenue recognition, fixed assets, borrowings, related parties. Use this information: [paste]. Where the framework has specific disclosure requirements you are unsure about, say so explicitly rather than guessing, and list the questions I should confirm with a qualified accountant.
If you want the statement structure pre-built so you only paste the trial balance, the AI P&L Builder Skill ($29) and the AI Balance Sheet Generator Skill ($29) hold the mapping and the checks for you. There is also a financial statement generator overview page that explains how the three statements fit together.
Forecasting and cash flow prompts
Cash forecasting is where AI saves the most hours, because the arithmetic is tedious and the structure is standard. Give it dates, amounts and payment behaviour, and make it list assumptions.
Prompt 11: 13-week cash flow forecast
Build a 13-week direct cash flow forecast starting [date]. Opening cash: [amount]. Weekly receipts: [paste debtors ledger with due dates and expected payment behaviour]. Weekly payments: payroll [amount and dates], rent [amount and dates], suppliers [paste creditors with due dates], tax [amounts and dates], loan repayments [amounts and dates]. Show each week's opening cash, receipts, payments, closing cash and the lowest balance in the period. List every assumption you make in one block at the top.
Prompt 12: Scenario cash forecast
Take this base-case cash forecast: [paste]. Create two more versions: a downside where receipts slip by [x] weeks and revenue falls [y] percent, and an upside where [describe]. Show the three closing cash lines side by side by month, identify the first month each scenario breaches a minimum cash balance of [amount], and list the three actions with the largest cash impact that could be taken in the downside case.
Prompt 13: Rolling 12-month budget
Create a rolling 12-month budget from [start month] for [company]. Start from these actuals for the last six months: [paste]. Apply these assumptions: revenue growth [x] percent per month, headcount plan [paste], known cost changes [paste], seasonality [describe]. Output a monthly P&L budget and a summary of the assumptions with the three that most drive the result, so a CFO can challenge them.
Prompt 14: Revenue and collections forecast
Here is our aged receivables report: [paste customer, invoice date, due date, amount, days overdue]. Based on this payment history: [paste or describe typical days to pay by customer], estimate the cash we will collect in each of the next eight weeks, list the customers most likely to slip past 60 days, and draft a short collections priority list with the recommended action for each.
Prompt 15: Working capital and runway
Using these figures, calculate our working capital cycle (days sales outstanding, days payable outstanding, days inventory outstanding) and our cash runway at the current monthly burn: revenue [amount], cost of sales [amount], receivables [amount], payables [amount], inventory [amount], cash [amount], average monthly net cash outflow [amount]. Explain each metric in one sentence for a founder and suggest the two levers most likely to extend runway.
For a repeatable version of prompts 11 to 15, the AI Cash Flow Statement Template Skill ($4.99) fixes the layout, and Petra - FP&A Analyst AI Skill ($29) covers budgets, scenarios and drivers. We wrote a longer walkthrough in How to Use Claude for FP&A and How to Use Claude for Cash Flow Forecasting.
Analysis and variance prompts
Variance analysis is a pattern-recognition task, which is exactly what these models are good at. Your job is to supply the operational context that the numbers alone do not show.
Prompt 16: Budget vs actual variance
Here is our P&L for [month] with budget and actual by line: [paste]. Produce a variance analysis: calculate the variance in absolute and percentage terms, mark anything over [threshold] as material, and for each material line give the most likely driver based on the data and this context: [paste operational notes]. Separate what you calculated from what you inferred.
Prompt 17: Price, volume and mix
Break down the change in revenue from [period A] to [period B] into price, volume and mix effects using this data: [paste product or service, units, average price for both periods]. Show the bridge from period A revenue to period B revenue as a table, explain the method you used, and tell me which single product or customer group explains most of the change.
Prompt 18: Margin analysis by segment
Using this data by [product / customer / region]: [paste revenue, direct costs, allocated overheads], calculate gross margin and contribution margin for each segment, rank them, and identify segments where margin has moved more than [x] points against the prior period [paste prior period]. Suggest three questions the CFO should ask the operations team about the weakest segment.
Prompt 19: Trend and anomaly review
Here are 24 months of our general ledger balances by account: [paste]. Identify accounts with unusual movements compared with their own history (spikes, sign changes, flat lines where activity is expected). For each anomaly, give the month, the size of the deviation and two plausible explanations, one innocent and one that would need investigation. Present the top ten by size.
Prompt 20: KPI dashboard definitions
Define a finance KPI set for a [industry] business with [revenue range] revenue. For each KPI (aim for 10 to 12), give the formula using standard account names, the data source, the review frequency, a sensible target range with the reasoning, and the leading indicator that moves before it does. Format as a table I can hand to the reporting team.
Communication prompts for leadership and clients
The last group turns finished numbers into words that owners, boards, lenders and clients will read. Keep the word limits in the prompt; the models default to too long.
Prompt 21: Board finance summary
Write a one-page finance summary for the board of [company] for [month]. Data: [paste P&L, cash, KPIs versus plan]. Structure: what happened, why, what we are doing about it, what we need from the board. Plain English, no jargon, maximum 350 words, and end with three decisions or discussion points. Mark any statement that depends on an assumption I should confirm.
Prompt 22: Client month-end email
Draft a month-end email from a bookkeeper to a small-business client, [client name], for [month]. Include: the headline numbers [paste], three things they should know, two questions I need answered to finish the close (for example uncategorised transactions [paste]), and the deadline for their response. Friendly, short, no accounting jargon.
Prompt 23: Explain a variance to a non-finance manager
Explain to the [department] manager why their [cost line] was [amount] over budget in [month]. Here is the detail: [paste transactions and budget assumptions]. Use plain language, avoid blame, show the two or three items that caused most of the overspend, and end with one specific request that would prevent it next month.
Prompt 24: Lender or investor update
Draft a quarterly update to our [lender / investors] for [quarter]. Data: [paste revenue, EBITDA, cash, covenant metrics versus thresholds, headcount]. Cover performance against plan, covenant headroom, cash position and outlook, in under 400 words. Flag anything that could be read as a covenant concern and suggest how to present it honestly and calmly.
Prompt 25: Audit query response
Our auditors have asked: [paste query]. Here is the relevant supporting information: [paste]. Draft a clear, factual response that answers exactly what was asked, references the supporting documents by name, and does not speculate. List separately anything they asked for that we do not yet have so I can chase it.
Cash flow forecast template: the structure to paste into prompt 11
A cash flow forecast template does not need to be clever. It needs to be direct-method, weekly for the next 13 weeks, and honest about assumptions. This is the structure we use; copy it into a spreadsheet or straight into the prompt.
| Row |
What goes in it |
Source |
| Opening cash |
Closing cash from the previous week (week 1 uses the bank balance today) |
Bank |
| Receipts: customer collections |
Invoices due that week adjusted for each customer's typical days to pay |
Aged receivables |
| Receipts: other |
Tax refunds, grants, asset sales, financing drawdowns |
Known events |
| Payments: payroll and taxes on pay |
Net pay on pay dates, payroll taxes on their due dates |
Payroll calendar |
| Payments: suppliers |
Creditors by due date, plus recurring direct debits |
Aged payables |
| Payments: rent, loans, tax |
Fixed dates and amounts |
Contracts, tax calendar |
| Net cash movement |
Receipts minus payments |
Calculated |
| Closing cash |
Opening plus net movement; becomes next week's opening |
Calculated |
| Minimum cash and headroom |
Your floor (for example one payroll) and how far above it you are |
Policy |
| Assumptions block |
Every judgement call, dated, with the owner |
You |
Update it every week by replacing week 1 with actuals and adding a new week 13. The forecast becomes useful after about four weeks, once you can see how your assumptions perform against reality.
Month-end close checklist: the structure to paste into prompt 5
Prompt 5 asks the model to review your close against a checklist. Here is a checklist structure that works for most small and mid-sized businesses. Give each line an owner, a due day (day 1 to day 5 after month-end) and a status.
-
Day 1 to 2, sub-ledgers: all bank and credit card feeds imported and categorised; all sales invoices raised; all supplier bills entered; payroll journal posted; inventory count or valuation posted.
-
Day 2 to 3, reconciliations: every bank account reconciled to statement; credit cards reconciled; receivables and payables sub-ledgers agree to control accounts; intercompany balances agreed; payroll liabilities agree to payroll reports.
-
Day 3 to 4, adjustments: accruals and prepayments posted with reversal dates; depreciation and amortisation posted; deferred revenue released; loan interest and foreign exchange revaluation posted; suspense and clearing accounts at zero.
-
Day 4, review: trial balance reviewed line by line against prior month; balance sheet accounts reconciled with support attached; P&L variance versus budget explained above your materiality threshold (prompt 16).
-
Day 5, reporting: management accounts pack produced (prompt 9); board summary or client email drafted (prompts 21 and 22); ledger locked for the period; open items carried forward with owners.
Paste that structure, with statuses, into prompt 5 and the model will tell you where the checklist and the trial balance disagree. It will not sign the close off; that remains your job.
Task table: which prompt, generator or skill to reach for
Prompts are free and quick. AI generators are small skills that fix one document format. Full skills load a whole role. This table maps the common accounting tasks to all three.
| Task |
Prompt |
Generator |
Skill |
| Categorise transactions, reconcile bank |
1, 2 |
- |
Selma - Bookkeeper ($29) |
| Accruals, depreciation, close review |
3, 4, 5 |
- |
Selma - Bookkeeper ($29) |
| Profit and loss statement |
6 |
AI P&L Builder ($29) |
Victor - CFO ($29) |
| Balance sheet |
7 |
AI Balance Sheet Generator ($29) |
Victor - CFO ($29) |
| Cash flow statement and forecast |
8, 11, 12 |
AI Cash Flow Statement Template ($4.99) |
Petra - FP&A Analyst ($29) |
| Budget, scenarios, variance |
13, 16, 17, 18 |
- |
Petra - FP&A Analyst ($29) |
| Expense reports and reimbursements |
1 |
AI Expense Report Generator ($4.99) |
- |
| Board, lender and client communication |
21 to 25 |
- |
Victor - CFO ($29) |
| Everything above in one library |
All |
Legal & Finance Prompt Library ($23) |
Finance Agent Bundle ($99) |
If you are choosing between skills, Best Claude Skills for CFOs and Finance Teams compares them side by side, and AI Prompts for Legal and Finance Work covers the contract side that sits next to accounting in many small firms. For more prompt collections across other roles, see the ChatGPT prompts page or the AI bookkeeper collection.
Frequently asked questions
Is ChatGPT or Claude better for accounting prompts?
Both handle the prompts in this article. Claude tends to be stronger with long pasted inputs such as a full trial balance or a 12-month general ledger export, and it is careful about flagging its own assumptions. ChatGPT is convenient if your team already lives in it and wants a custom GPT. The prompt matters far more than the model: give it the role, the real figures, one task and a fixed output format.
Can I paste client financial data into an AI model?
Only within your firm's confidentiality and data-protection rules and your engagement letter. Use a business or enterprise plan where inputs are not used for training, strip names and account numbers where you can, and never paste bank credentials or tax IDs. When in doubt, replace identifiers with placeholders and keep the real mapping in your own system.
Will AI replace my bookkeeper or accountant?
No. These prompts remove typing and first-draft work: categorisation suggestions, reconciliation breaks, variance commentary, a draft board note. Every figure still has to be reviewed and signed off by a qualified accountant, and the AI cannot take professional responsibility. Think of it as a fast junior who never gets tired but always needs review.
What is the difference between a prompt and a skill?
A prompt is a one-off instruction you type each time. A skill is a reusable file (a .md skill file plus README) that loads the role, the rules and the output formats once, so Claude, ChatGPT (as a custom GPT), Gemini or Copilot behaves like a trained specialist every session. Skills on KissMySkills are one-time purchases with a 30-day money-back guarantee.
Is this financial or tax advice?
No. This article and the prompts in it are productivity tools for drafting and analysis. They are not financial, accounting or tax advice. All outputs, figures and statements must be reviewed by a qualified accountant before they are used, filed or shared with clients or lenders.
The bottom line
AI does not close the books, but it removes most of the typing between the ledger and the sign-off. Use the 25 prompts above for the first draft of categorisation, reconciliations, statements, forecasts, variance commentary and the emails that follow. Keep the workings visible, keep the assumptions listed, and keep a qualified accountant on every output. When the same prompt is being typed every week, move it into a skill so the role, the rules and the formats load once.
The whole finance desk
Finance Agent Bundle - All 5 AI Finance Agents
Bookkeeper, FP&A analyst, CFO, cash flow and finance team skills in one pack. One-time purchase, .md skill files plus READMEs, 30-day money-back guarantee.
$99
Get it now →
Browse every AI bookkeeper and finance skill at KissMySkills.
