Home · Business · Finance & Accounting · Invoicing & credit control
As of 13 August 2026, AI can only partly allocate credit notes to the right invoices.
This still needs a person who signs their name to it.
Can you do it?
15 minutesto a draft.
1 hourto something you’d act on.
Cost, all in£0
Skill neededpower-user
Who has to check ita colleague
What the alternative costsThe supplied tool information gives no price for an alternative bookkeeping service.
If this goes wrong, a credit note can be applied to the wrong invoice and leave customer balances, chasing activity or VAT records incorrect.
What to actually do
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.
Use a tool built for this
Second choiceDo it yourself
The distant thirdA chat interface, power-user skill, and roughly 1 hour until you can act on the result.
How to actually do it
- Open your accounting system and export the credit notes and open invoices for the same customer ledger and period, retaining invoice numbers, customer IDs, amounts, VAT amounts, currencies and statuses.
- Export or copy the written credit allocation policy, including rules for referenced invoices, partial credits, closed invoices, disputes and unapplied balances.
- Remove unrelated personal information, then paste the credit note table, invoice table and allocation rules into a chatbot with the prompt above.
- Check the proposed allocations against the source records by matching every customer ID, invoice number, credit note number, currency and amount, and recalculate the remaining balances.
- Give each exception to the finance colleague responsible for credit control and resolve it from the original customer account, correspondence or credit note rather than asking the model to guess.
- Enter only approved allocations into the accounting system, then compare the posted credit total and customer balances with the control totals produced from the source exports.
Prompt
Allocate the credit notes to the correct invoices using the records below. Credit notes: [Paste a table with credit note number, customer name or account ID, date, total amount, VAT amount, currency, reason and any stated invoice reference.] Open invoices: [Paste a table with invoice number, customer name or account ID, invoice date, due date, outstanding amount, VAT amount, currency and status.] Our allocation rules: [Paste your written rules, including whether a credit must be applied to the referenced invoice, how partial credits are handled, and what to do with disputed, closed or cross-customer items.] Return three tables: 1. Proposed allocations with credit note number, invoice number, customer, amount applied, remaining credit and remaining invoice balance. 2. Exceptions that you cannot allocate confidently, with the exact reason and the missing information needed. 3. Control totals comparing credit notes supplied, credit allocated and credit left unapplied. Match only on evidence in the supplied records. Do not invent invoice references, customer identities, amounts, VAT treatment or allocation rules. Do not allocate across customers or currencies unless the stated rules explicitly allow it. Flag duplicate invoice numbers, duplicate credit notes, amount mismatches, already-closed invoices and any item that could affect VAT. Do not post anything to an accounting system. Use exact source values and show your arithmetic.
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot establish the commercial reason for a credit when the note or customer correspondence is unclear.
- AI cannot decide whether a disputed or partially valid credit should be applied without your policy and an accountable finance decision.
- AI cannot guarantee that an invoice reference is genuine when records contain duplicates, cancellations or reused customer details.
- AI cannot take responsibility for incorrect customer balances, credit-control action or VAT records.
What caps this at PARTLY: judgement under ambiguity, legal accountability and verification cost.
How we scored this
Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.
| Axis | Score (0–2) |
|---|---|
| Output | 1 |
| Inputs | 2 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 2 |
| Total | 7 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can AI match credit notes to invoices?
- Yes, it can suggest matches from invoice numbers, customer IDs, dates and amounts. It should leave uncertain, partial and disputed items in an exceptions list rather than making a silent allocation.
- Can AI post credit note allocations into my accounting software?
- A general chatbot can produce the allocation file but should not be trusted to post it without a controlled workflow and approval. Check whether your accounting system has its own supported automation before sending approved entries.
- How do I stop AI applying a credit note to the wrong invoice?
- Give it complete records and explicit rules, require exact source values, and tell it to flag uncertainty instead of guessing. A finance colleague should compare every proposed allocation with the customer account before posting.
- Can AI handle partial credit notes and VAT?
- It can calculate proposed partial allocations when the source amounts and your rules are clear. It cannot decide the correct VAT treatment for an ambiguous case, so refer those items to your accountant or VAT adviser.
Nearby answers
Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.
The newsletter
AI news, new answers and product picks, straight to your inbox.