1. Variance explanation drafter
When to use it: Month-end, when every account needs its paragraph.
Inputs: Budget vs actual rows, the context you know (one-offs, timing, hires).
Starting prompt:
Explain these budget variances: [paste rows]. For each line: state the variance, then the most likely explanation using only this context: [paste]. Mark any line where the context does not explain it as NEEDS SOURCE. Formal tone, 2 sentences max per line.Expected output: Draft explanations where unexplained lines are flagged instead of hand-waved.
Boundary: ChatGPT wordsmiths your context. If you do not give it the reason, it will invent a plausible one - watch for that.
Ready-made: the Budget & Variance Agent runs this as a standing workflow with gates, $19 one-time.
2. 13-week cash flow roll
When to use it: Weekly, when the forecast needs its update.
Inputs: Last week’s forecast, this week’s approved receivable/payable changes.
Starting prompt:
Roll this 13-week cash flow forecast forward one week: [paste forecast]. Apply these approved changes: [paste]. Keep every assumption visible per line and list what you changed and why.Expected output: An updated forecast with a visible change log.
Boundary: It applies your approved changes only. New assumptions come from you.
Ready-made: the Cash Flow Forecast Agent runs this as a standing workflow with gates, $19 one-time.
3. Management pack commentary
When to use it: When the pack is due and commentary is the last blank page.
Inputs: This month’s metric table with sources, last month’s commentary.
Starting prompt:
Write management pack commentary for these metrics: [paste]. Per metric: what moved, the direction, the source row. 2 sentences max. Mark any metric without a clear source as UNVERIFIED.Expected output: Consistent commentary with sources attached and unverified metrics flagged.
Boundary: It writes from your table; the story you tell the board is yours.
Ready-made: the Management Reporting Agent runs this as a standing workflow with gates, $19 one-time.
4. Spend request policy check
When to use it: When requests arrive faster than anyone can check them.
Inputs: The spend request details, your policy rules (thresholds, categories, required evidence).
Starting prompt:
Check this spend request against our policy: [paste request and policy]. Output: policy check per rule (pass/fail/unclear), missing evidence, and the approver level it routes to. No approve/reject recommendation.Expected output: A policy check sheet that routes the request to the right approver.
Boundary: It checks against written policy. Approval is always a person.
Ready-made: the Spend Request Agent runs this as a standing workflow with gates, $19 one-time.
5. Budget vs actual summary
When to use it: For the monthly owner meeting that starts in 20 minutes.
Inputs: The budget vs actual export, the 5 lines the meeting cares about.
Starting prompt:
Summarize this budget vs actual export: [paste]. Lead with these 5 lines: [paste]. Per line: budget, actual, variance, one-line reason if it is in the data. End with the 3 questions this data raises.Expected output: A one-screen summary that survives the first question in the meeting.
Boundary: Reasons only count if they are in your data; otherwise it should say so.
Ready-made: no agent for this one yet.
Quick comparison
| Workflow | Effort | Cadence | Risk | Ready-made agent |
|---|---|---|---|---|
| 1. Variance explanation drafter | One paste per run | Monthly | Draft-first, review before use | Budget & Variance Agent |
| 2. 13-week cash flow roll | One paste per run | Weekly | Draft-first, review before use | Cash Flow Forecast Agent |
| 3. Management pack commentary | One paste per run | Per event | Draft-first, review before use | Management Reporting Agent |
| 4. Spend request policy check | One paste per run | Per event | Draft-first, review before use | Spend Request Agent |
| 5. Budget vs actual summary | One paste per run | Monthly | Draft-first, review before use | none yet |
Free download
Get the Variance report example (CSV) - the companion file for these workflows, with sample rows filled in.
Platform notes
ChatGPT is a third-party product (https://chatgpt.com). We are not affiliated with or endorsed by them. Capabilities and names verified 2026-09-13; re-check before relying on specifics. The prompts above are platform-agnostic - they work in any capable assistant.
Want the job done instead of prompted? Browse the finance teams agent menu or the AI agents for finance teams guide.
Questions people ask
Are these ChatGPT workflows free?
Yes - every item includes a starting prompt you can copy free. Matching $19 standing agents are linked where they exist.
Are you affiliated with ChatGPT?
No. AI Agent Skills is an independent store; ChatGPT is named only because these workflows run on it. Facts checked 13 September 2026.
Last updated 13 September 2026 · Reviewed by the AI Agent Skills team.