LedgerPay's consumer app grew quickly after a partnership launch, taking monthly conversation volume from roughly 4,000 to 16,000 in a single quarter. The support team of nine could not be scaled at the same rate, and in regulated payments, hiring speed is limited by compliance training rather than recruitment.
The first move was outbound, not inbound
Analysis of inbound volume showed that a large share of contacts were predictable status questions triggered by known events: a payment held for review, a verification step not completed, a payout timing question after a bank holiday. The team built proactive notifications for the six highest-volume events before touching automation, which removed a substantial share of the new inbound volume outright.
Then automation, scoped narrowly
- Automated verification-status and payout-timing questions with strict read-only data access.
- Kept every dispute, chargeback and account-restriction conversation human-only, by policy enforced outside the model.
- Used AI drafting with mandatory human review for the middle tier, which is where most of the handle-time saving actually came from.
Operational changes that made it hold
The team introduced a written shift handover, a weekly risk-weighted quality review including AI-drafted replies, and a single dashboard showing the oldest unanswered conversation. That last number became their operating metric during the spike, in preference to averages.
Outcome
Satisfaction remained flat through a four-fold volume increase, and the team added two agents rather than the twelve the original model implied. Their compliance lead notes that the audit trail on automated actions was the deciding factor in getting approval at all.