Guide
ChatGPT prompts for finance, and where to save them
The finance prompts worth keeping are the ones for writing around the numbers: a variance note with your materiality threshold, a month-end close checklist with owners and days, a polite payment reminder, and a one-page summary for the board. Your spreadsheet stays the source of every number; the AI explains and drafts. Copy the four below, use them in a company account that doesn't train on your data, and save each where your AI reads it on its own: your standards in custom instructions, each month's files in a project, each repeated task as a skill.
Short answer
- Four prompts: variance note, close checklist, payment reminder, board summary.
- The spreadsheet calculates; the AI explains. Recompute any number it states.
- Company account only for financial data.
- Save them: standards in custom instructions, months in projects, tasks as skills.
The four prompts
1. Variance note
"Explain budget vs actual for [month] from this table. Only comment on lines over [5% and $2,000]. For each: favourable or unfavourable, the driver (timing, price, volume or one-off), and what we'll do. Don't calculate anything new. Table: [paste]."
2. Month-end close checklist
"Build our month-end close checklist: each task, an owner from [names], and the business day after month end it's due. Cover bank and card reconciliations, accruals, payroll, invoices, and the review. Our current list: [paste]."
For a full close with a reviewer on every reconciliation and a "done when" rule for each task, see ChatGPT for accountants.
3. Payment reminder
"Write a payment reminder for invoice [number], [amount], [days] overdue. Friendly first reminder, the invoice details, how to pay, a contact. No threats. Under 100 words."
4. Board summary
"Turn this month's management pack into a one-page summary: the three numbers that matter, what changed and why, cash runway, and the one decision we need. Use only the figures in the pack. Pack: [paste]."
Where to save each
| What | Where | Why |
|---|---|---|
| Thresholds, currency, house style | Custom instructions | Applied in every chat |
| One month's pack and notes | A project per month or per entity | Instructions and files apply inside the project |
| A task you repeat every month | A skill | Loaded when you ask for the task |
From OpenAI's and Anthropic's help pages, read 5 November 2026.
A worked example
Illustration: a made-up business. Omar is the finance manager at a 35-person design agency. Each month he wrote the variance note from scratch and re-explained the agency's threshold to ChatGPT. He moves it:
- Custom instructions: "Finance at a UK agency. GBP. Materiality: 5% and £2,000. Never compute; explain the figures given."
- A project for each month's close with the pack and last month's note.
- The variance note as a skill, with his format and two past notes the directors liked.
In October's project he types "variance note for October, table below". The draft flags freelancer costs 18% over budget as "timing: two projects pulled forward". He checks the figure against the ledger, keeps the explanation, and adds the one thing the AI couldn't know: the client has already paid for both.
Save your first prompt
- Pick the prompt you rewrite every month.
- Add your thresholds and two past examples.
- Move your standards to custom instructions.
- Save the task as a skill and test it on last month's figures.
- Delete the old copy from your notes.
Next step: finance rules for the whole team
Before anyone else in finance pastes client or company figures, agree the rules: what finance can put into AI. For skills built from these prompts, see Claude skills for finance.
Skills for this
- AI practiceTurn a prompt into a skillTurns a prompt that works into a skill the whole team can reuse, with a valid name, a when-to-use description, steps, an output template, an example, and checks.
- FinanceBudget variance noteExplains budget vs actual variances over a threshold: favorable or unfavorable, the driver (timing, volume, price, one-off), and the forecast impact.
- 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.
- 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
What are good ChatGPT prompts for finance teams?
Prompts for the writing around the numbers: a variance note with a materiality threshold, a month-end close checklist with owners and days, a polite payment reminder, and a one-page summary for the board. The AI writes; your spreadsheet stays the source of every number.
Can I paste financial data into ChatGPT?
Only into a company account whose plan doesn't train on your data by default, such as ChatGPT Business or Claude Team, and within your contracts. Never paste passwords into any AI tool, and never paste bank details or unreleased results into a personal account.
Can ChatGPT do my calculations?
Let your spreadsheet calculate and give the AI the results to explain. Recompute any number the AI states before it goes anywhere.
AI skill
Have your AI turn a prompt that works into a skill your team can reuse
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. When one of these finance prompts gives you a draft you'd sign off, Turn a prompt into a skill has your AI turn it into a skill with your thresholds and format, so month end starts from your standard.
14 days free · No card needed
Sources
- OpenAI Help: ChatGPT custom instructions — applied across chats; read through search-result text (read 2026-11-05)
- OpenAI Help: Projects in ChatGPT — read through search-result text (read 2026-11-05)
- Claude Help Center: Using skills in Claude — every plan with code execution on (read 2026-11-05)
- ChatGPT pricing — Business: no training on business data by default (read 2026-11-05)
- Anthropic Privacy Center: commercial products — not used for training by default (read 2026-11-05)