Stop counting an automation run as finished the moment the handoff message lands in the next queue. Most platforms treat queue acceptance as proof the receiving system will act, even when its worker pool is stalled or its schema rejects the payload.
RPA tools excel at single-system clicks because vendors built their metrics around local task completion. Once the script hands off via email, API stub or shared folder drop, the platform stops tracking. The dashboard still shows green.
Procurement teams buy these tools on demos that never include concurrent load from other departments. The handoff queue fills during month-end, yet the contract only required the bot to emit the message. Recovery logic sits in a different budget line.
The real cost appears in exception queues that grow without automated alerts. Staff who once owned the downstream process now spend hours reconciling orphaned entries that the automation platform declared complete weeks earlier.
Schema drift between the sending and receiving systems turns every routine update into silent failure. The automation vendor refuses scope changes because the handoff itself succeeded; the buyer must fund the mapper rewrite separately.
Measure instead by confirmed state change in the target record. Require the automation contract to include retry windows and explicit failure callbacks that surface in the same dashboard used for the originating step.
Without that linkage, end-to-end claims remain marketing copy. The organisation pays for two half-solutions while the gap between them widens every release cycle.