Guide
ChatGPT for accountants: start with the month-end close
ChatGPT helps accountants most with the work around the numbers, and the month-end close is the best place to start. Give it who does the books, your accounting system and its feeds, what the business has (prepayments, payroll, fixed assets) and the day you want to close by. Ask for the tasks in close order, one owner each, a reviewer who isn't the preparer, a business-day offset from month end and a short "done when" rule, ending with review and lock. It needs no client figures, and the figures themselves stay yours to check.
Short answer
- Start with the close: tasks, owners, reviewers, days, done rules.
- Also good: variance questions, client emails, a first read of a rule.
- Stays with you: every figure and every reconciliation.
- Client data only in an approved business account.
What to use it for
| Job | What you give it | What stays with you |
|---|---|---|
| Month-end close checklist | Who does the books, the system, what the business has, the target day | Doing and reviewing each task |
| Variance questions | Two months' P&L lines, in an approved business account | The reasons: it can only list what to ask |
| Client email | The facts and the one thing you need from the client | Every number in it |
| A first read of a rule | Your question | Reading the rule itself |
This page is about which jobs to give ChatGPT first, starting with the close. The prompts for a variance note, a payment reminder and a board summary, written to save so you stop pasting them, are in Finance prompts, saved.
The close brief to copy
Month-end close brief
"Build a month-end close checklist for [company, size, type].
Books: [who prepares, who reviews]. System: [accounting system, bank and card feeds, billing, payroll provider].
The business has: [prepayments or annual plans, payroll, contractors, fixed assets, inventory, more than one currency].
Close by business day [n] after month end.
Tasks in close order: transactions, cash, receivables and payables, accruals and prepaids, deferred revenue, payroll, fixed assets, reports, review and lock. One owner per task; a reviewer who isn't the preparer for reconciliations and the final review. A day offset and a short 'done when' rule on each. Mark anything you're unsure applies with [check]; don't drop it."
A worked example
Illustration: a made-up business. Harlow, a 15-person design agency, closed around day 12, and the numbers kept changing after the reports went out. It uses QuickBooks Online with bank and card feeds and Stripe for client invoices; some clients pay a quarter's retainer in advance; payroll runs through an outside provider. Priya, a part-time outsourced bookkeeper, prepares; Tom, the founder, reviews. The checklist ChatGPT returns for a day-5 close:
| Day | Task | Owner / reviewer | Done when |
|---|---|---|---|
| −2 | Chase missing receipts and card notes | Priya | Every card line has a receipt or note |
| 1 | Code bank and card transactions; reconcile bank | Priya / Tom | No uncoded lines; ledger matches statement, differences explained |
| 2 | Reconcile cards and Stripe payouts; receivables and payables to aging | Priya / Tom | Payouts match deposits, fees and refunds booked; aging totals match ledger |
| 3 | Accruals; recognize prepaid retainers; record payroll | Priya / Tom | Each accrual has support; schedule ties to deferred revenue; payroll matches the provider report |
| 4 | P&L and balance sheet; variance against last month | Priya | Reports drafted |
| 5 | Review the balance sheet and big changes; lock the period | Tom | Questions answered; period locked |
It notes what it left out (inventory and fixed assets: none) and one item to confirm: [check: does Harlow need a sales tax accrual?]. Nothing in the brief was a client figure. In the first close, the Stripe reconciliation turned up refunds that had been left in revenue, the kind of thing a bank-only reconciliation misses.
The client-data line
- No client data in a personal account; some personal plans train on chats unless you turn it off.
- Use a business plan that doesn't train on your data by default, within your engagement terms and professional rules.
- US tax return information has its own rules; the finance data guide covers section 7216 and the IRS's June 2026 AI guidelines.
- Check every figure the AI writes against the ledger.
Next step: keep your finance briefs
You'll run the close brief every month, and the variance and client-email briefs with it. Finance prompts, saved shows where to keep them so each month starts from your checklist, not a blank chat.
Skills for this
- FinanceMonth-end close checklistBuilds a month-end close checklist for a small company: reconciliations, accruals, deferred revenue, payroll, and review, with owners and day offsets.
- FinanceBudget variance noteExplains budget vs actual variances over a threshold: favorable or unfavorable, the driver (timing, volume, price, one-off), and the forecast impact.
- Operations and adminVendor invoice check before paymentChecks a vendor invoice against the PO or quote: details, arithmetic, tax, terms, duplicates, and changed bank details; returns approve or hold.
- FinanceOverdue invoice reminder emailsWrites overdue invoice reminders by stage, from before due to 30+ days late, firm but polite, with the invoice details and how to pay up front.
Questions
How can accountants use ChatGPT?
For the work around the numbers: a month-end close checklist, questions to ask about a variance, client emails, a first explanation of a rule you then read at the source. Don't let it produce figures you haven't checked, and keep client data in an approved business account.
Can ChatGPT make a month-end close checklist?
Yes, from your setup: who does the books, the accounting system and its feeds, what the business has (prepayments, payroll, fixed assets) and the target close day. Ask for tasks in close order, one owner each, a reviewer who isn't the preparer, a business-day offset and a done rule, ending in review and lock.
Is it safe to put client data into ChatGPT?
Not in a personal account. Use a business plan that doesn't train on your data by default, within your engagement terms and professional rules. A close checklist needs no client figures at all.
AI skill
Have your AI build your month-end close checklist
Put this skill in AI Skill Pod and everyone on your team gets it in their Claude, Cursor or ChatGPT (Pro, Business, Enterprise): one invite link, free to start. Month-end close checklist has your AI build your close from your setup: tasks in close order, one owner each, a reviewer who isn't the preparer, a business-day offset and a done rule, ending in review and lock.
14 days free · No card needed
Sources
- Claude Docs: Prompting best practices — give context, constraints and the output format (read 2026-11-05)
- OpenAI Help: Data controls — personal plans train on chats by default unless turned off; read through search-result text (read 2026-11-05)
- Anthropic Privacy Center: commercial products — not used for training by default (read 2026-11-05)