← All use cases

Use case #8 · CPA

Month-end close run as six Claude prompts on Xero/QuickBooks exports

Nicholas Bird, CPA (LinkedIn) ·

The short version

What they did

TB fluctuation sweep, close checklist from the chart of accounts, bank-to-ledger rec, balance sheet rec workpapers, management commentary, and one consolidated client question list.

Result (as claimed)

“Nothing posts without an accountant’s name on it.” Suggests: “If it finds things your reviewer missed, the ledger was never the bottleneck.”

Prompt

Their prompt:

Export last month’s trial balance and the three before it. Ask Claude for every account that moved more than 20% with nothing offsetting it.

A fuller version to copy:

I’m an accountant reviewing [month] before closing the books. I’ll attach [trial balances for that month and the three before it], with [notes on debit and credit signs]. Help me identify changes to investigate.

1. Compare the latest month with the previous month. Show all four balances for accounts that moved more than 20%, plus the dollar and percentage change.
2. Look for possible offsetting movements in the other accounts. Show the amounts, but don’t claim they are linked without supporting detail.
3. Separate accounts with no visible offset from those needing more evidence, and draft one short list of client questions.

Return a table and a plain-English question list. Show the arithmetic, flag zero or negative comparison balances, and don’t invent explanations. Say what transaction detail is needed before confirming an offset. Don’t post anything.

Do it yourself

These steps are my suggested way in, not Nicholas Bird’s setup.

What you need

  • Trial balance exports for the last four months from Xero or QuickBooks.
  • A chat assistant that reads spreadsheets, such as Claude or ChatGPT, on a business plan.
  • A closed month you already reviewed by hand, to test against.

Start simple

  1. Pick a month you already closed. Export its trial balance and the three before it.
  2. Paste the case’s Prompt and attach the files. Ask for account, prior balances, change and percent change in a table.
  3. Compare its list with the fluctuations you actually investigated that month.
  4. Ask it to turn the flagged items into one list of client questions, in plain language.
  5. Note what it caught, what it missed and what it got wrong.

Take it further

  • Add the other steps one at a time: a close checklist built from the chart of accounts, a bank-to-ledger rec, balance sheet rec workpapers, then draft management commentary.
  • Save each prompt in a project so the whole close runs the same way every month.
  • Keep the rule from the source: nothing posts without an accountant’s name on it.

Check your work

  • Recompute two or three flagged percentages yourself. This library’s test found false statements about depreciation and offsets alongside correct flags.
  • Ask it to cite the exact balances behind each claim, and tie them to the export.
  • Every entry or client-facing number gets an accountant’s review before it posts or goes out.

Testing

2026-10-18: All three planted movements surfaced in 31.792 seconds. The answer also made false statements about depreciation and the offset, so it did not pass the full accuracy criterion. The full run record includes the exact inputs, output, timing and scoring. The source claim remains unverified.

2026-10-21 repeat: The repeat found all three planted movements in 41.321 seconds, but added a misleading receivables flag and false exact-ratio claims. It did not meet the full accuracy criterion. The linked report shows the repeat; the first run remains in the reproduction kit. Both used identical inputs and prompts, with different default model identifiers. The source claim remains unverified.

Before you rely on it

  • This is general information about how accountants use AI, not professional guidance for your engagements.
  • You stay responsible for the work under your professional standards, including review and sign-off. An AI tool’s output is a draft, not evidence.
  • Before you send client data to an AI tool, read the tool’s data and training terms, and get any client consent you need. For tax return information, that includes IRC section 7216.

Source checked. The link, author, date and quote were confirmed against the source. The result itself has not been reproduced. Their prompt is quoted from the source; any fuller version is written for this library. Business size: Not stated. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.

More in Close/bookkeeping

New cases go out every Saturday in Inside AI.