As of 13 August 2026, AI can identify unpaid sales invoices.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Export the current sales invoice list from your accounting system, including invoice number, customer, invoice date, due date, total, credit notes and current status.
   2. Export the matching payment or bank transaction period, including transaction date, amount, reference, payer and any invoice number in the bank reference.
   3. Export separate records for credit notes, refunds, disputed invoices and known customer-account transfers.
   4. Remove unrelated personal bank data and paste the three datasets into the prompt in clearly labelled sections.
   5. Ask the model to produce the status table and exception list using the exact matching rules in the prompt.
   6. Compare every item marked apparently unpaid or partly paid with the original invoice, the accounting ledger and the bank transaction export, then correct any unallocated or wrongly matched payment.
   7. Send only the verified list to the person responsible for credit control, and keep disputed or cannot determine items out of automated chasing.

   Prompt

   Identify which sales invoices appear unpaid from the records below. Treat an invoice as paid only when the payment can be matched to the invoice number or to a clearly matching customer, amount and date. Account for partial payments, duplicate payments, credit notes, refunds, overpayments and payments that cannot be allocated. Do not assume that an unmatched bank receipt is unpaid revenue. Return a table with invoice number, customer, invoice date, due date, invoice total, matched payments, credit notes, balance apparently outstanding, status, and the evidence for the status. Use these statuses only: paid, partly paid, apparently unpaid, disputed or cannot determine. List every exception separately. Do not invent or infer missing figures. State which records you used and ask for any missing fields needed to improve the matching. These are UK business records, so keep the analysis to bookkeeping and do not give tax or legal conclusions.
   
   Invoice export:
   [PASTE INVOICE EXPORT]
   
   Payment or bank transaction export:
   [PASTE PAYMENT EXPORT]
   
   Credit notes, refunds and disputed invoices:
   [PASTE RELEVANT RECORDS]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.