2 Credit Management AI Agents, Live in Production.

Credit memos are a small process with a long tail of confusion. The fast way to apply one is against whatever open invoice it will clear, which tidies the ledger and destroys the audit trail — a credit raised because of a delivery shortfall on one order ends up settling an unrelated invoice, and six months later nobody can explain either transaction. Reason codes exist to prevent this and are routinely ignored because respecting them requires matching credits to specific invoices rather than to balances. Credit limits fail differently: they are checked at onboarding and then reviewed when something goes wrong, so the drift in between — a customer whose payment behaviour has been deteriorating for two quarters while staying just inside their limit — goes unnoticed until it does not.

These agents keep both honest. Credit Memo Application decides which open credit memo settles which invoice, respecting age and reason codes, and lays out the application plan with the residual balances — the plan being reviewable before anything posts. Customer Credit Monitoring reviews every credit account against its limit and payment behaviour, ranks the ones drifting toward trouble, and prepares the credit watch list. Ranking by drift rather than by proximity to the limit is the useful part: an account at sixty percent of its limit whose payment timing has doubled is a better prediction of a problem than one at ninety-five percent that has always paid on the day.