Skip to content
Bank and Credit Card Reconciliations

Triage Unmatched Bank Transactions and Miscoding with an AI Exception Review

View source

What the source describes

This Workplace AI Institute article argues that bookkeepers should let accounting platforms do bulk bank matching and then hand only the exception list to a general-purpose model such as ChatGPT or Claude. It describes exporting unreconciled items as a CSV and asking the model, using its analysis tool, to find likely duplicates, flag GL coding inconsistent with the payee, and list large items to confirm with the client. The article warns that models can be confidently wrong about specific dollar amounts and that the review is a triage tool, not a posting tool.

Result or outcome

According to the source, using the LLM for exceptions and explanations rather than raw matching turns a slow hunt for discrepancies into a structured review, but every suggested journal entry still needs human checking against source documents.

How to run it

  1. Let the bank feed and rules in Xero or QuickBooks Online do the bulk matching, so only unmatched items remain in the reconciliation.
  2. Export the unreconciled items as a CSV and remove bank account numbers, full names and tax identifiers before uploading anywhere.
  3. Attach the CSV to ChatGPT, Claude, Grok or Gemini on a firm-approved account, and run the prompt on this page.
  4. Work through the likely duplicates, miscodings and large items, checking each against bank statements, receipts and invoices before changing anything.
  5. Send client questions for unexplained items, then post corrections yourself in the accounting platform and complete the reconciliation.

Try it yourself

Copyable prompt

Act as a senior bookkeeper. Attached is a CSV of bank transactions for [client name] for [period] that [Xero / QuickBooks Online] could not auto-match. Columns: [Date], [Payee], [Description], [Amount], [Suggested Account]. Our chart of accounts is: [paste account list]. Use your data analysis tool to calculate, not estimate. Produce: 1) a table of likely duplicates (same amount, dates within [3] days, similar payee), 2) a table of transactions whose suggested account looks wrong for the payee, with a more likely account and a one-line reason, 3) any item over [amount] that should be confirmed with the client, written as a client-friendly question, 4) the variance total if every flagged item were corrected. Do not post anything; this is a review list only.

Replace the [bracketed] placeholders. Remove client names, account numbers, and personal data before pasting into any AI tool.

What to review

The source warns that models can be confidently wrong about specific dollar amounts, so tie each flagged amount and date back to the bank statement line. Confirm suggested accounts against the receipt or invoice, and check that a supposed duplicate is not two genuine payments. Treat the output as a triage list; every journal entry still needs your review before posting.

Related free download: Bank Reconciliation Cheat Sheet (Printable cheat sheet (PDF))

Source: Workplace AI Institute (blog)

Not professional advice. This is general information, not accounting, tax, legal or financial advice. Check AI output yourself and consult a licensed professional for your situation. See our Terms of Use.

This entry summarizes a public source. Features, plans, and availability change, so confirm details with the vendor. Tool names are trademarks of their respective owners.

Related workflows

  • Reconciliations

    Reconcile a bank statement export against your chart of accounts with Claude Cowork

    Claude Academy walks through giving Cowork a folder with a checking account statement export and the chart of accounts, then asking it to match transactions to accounts, flag anything that does not line up, and draft journal entries for discrepancies. Claude may ask clarifying questions and build a plan for review, then returns the reconciliation logic, specific discrepancies and journal entries ready for review. Suggested follow-ups include exporting entries as CSV and creating a reconciliation summary for the records.

    Claude, Claude Cowork +1

    Claude Academy (Anthropic tutorial)
  • Reconciliations

    Build reconciliation working papers from a bank statement and invoice folder with Claude

    A finance educator and former auditor, writing in the Tips & Insights on Finance & AI newsletter, describes dropping 100 invoices and a bank statement into Claude Cowork and running the reconciliation skill from Anthropic's Finance plugin. The post lays out six steps: install the plugin, create Bank_Statement and Invoices subfolders, create a Cowork project, call the skill, ask Claude to add a verification tab, and review only the exceptions. The author stresses that messy inputs still produce poor results.

    Claude Cowork, Anthropic Finance plugin +1

    Tips & Insights on Finance & AI (newsletter)
  • Reconciliations

    Build a Formatted Reconciliation Workbook from CSVs with Claude

    Anthropic's help article says Claude can execute code to create and work with files directly in a conversation, producing Excel spreadsheets, PowerPoint presentations, Word documents and PDFs that you can download or save to Google Drive. It says code execution and file creation are available on Free, Pro, Max, Team and Enterprise plans, on the web, Claude Desktop and Claude Mobile. Its examples include building financial models in Excel with working formulas and performing advanced analyses on uploaded data, which is what a reconciliation workbook built from bank and ledger exports needs.

    Claude, Microsoft Excel

    Claude Help Center
  • Receipts and Expenses

    Categorize Leftover Expenses with AI After Bank Rules

    This Brocent guide argues that QuickBooks and Xero bank rules should handle recurring transactions deterministically, and that ChatGPT is best used for the residue of unrecognised payees, split transactions and anomaly review. It recommends exporting uncategorised transactions to CSV with the real chart of accounts, asking for a suggested category with a confidence note and flags, and reviewing everything before posting. It stresses never auto-posting AI output, keeping an audit trail and leaving tax judgements to an accountant.

    Works in ChatGPT, Claude, Grok, Gemini

    QuickBooks Online, Xero

    Brocent Blog

Some links on this page are affiliate links. If you sign up through one, we may earn a commission at no extra cost to you. It does not change what we recommend. See our affiliate disclosure.