Skip to content
MirrorFirm

Why can't an agent just match payments by amount and reference?

Amount and reference matching handles routine cases, but combined remittances, stale references and deductions break it, and those are the cases that test an agent.

**Can an agent match on amount and reference alone?** Only for the easy cases. Exact-amount, exact-reference matches are routine and rule-based software already handles them. The cases that test an agent are the ones where those two signals fail.

**What goes wrong?** - One payment covers several invoices, so no single invoice equals the amount. - The reference names a previous owner, an old account number or a project rather than an invoice. - Two customers have similar names and the same amount is due from both. - The payer has deducted a credit, withholding or bank fee. - An invoice has been reissued, so the reference points to a cancelled document.

**What does the agent need beyond the payment line?** It needs open items, credit notes, customer history, written allocation policy and dispute status. It also needs a way to ask for help when the evidence is thin.

**Why does that matter for evaluation?** A benchmark that rewards only exact matches will not show whether an agent behaves safely when evidence is weak. Task packs for this workflow can include cases where the correct action is to hold the payment as unapplied and open an exception, and they can score that choice as correct.

Evaluate agents on the cases where simple matching fails, including the ones where the correct action is to hold the payment and escalate.

Audience:
Agent developers
Content type:
FAQ
Buyer question:
What it is and why it matters

Content release: r-a220eb6e099b243b

More on Payment allocation