You signed the automation contract once the vendor dashboard showed high completion rates inside their controlled test flows. Those rates stopped counting once the bot reached the first external system that required write access the procurement never requested.
The bot script executes every mapped field correctly yet halts at the ERP interface because the service account lacks the specific transaction code for batch updates. The vendor records this as a successful run and moves the exception to a shared queue that finance never monitors.
Procurement documents listed only read-only credentials for the target systems because security reviews treated the automation as another reporting layer. No one updated the change request process to include elevated permissions after go-live.
Each downstream rejection now requires manual intervention that the original business case subtracted from headcount projections. The gap appears only after the first month-end close when reconciliation files pile up outside the bot's scope.
Vendors measure success at script termination, not at the point where the receiving system accepts and processes the record. Your contract contains no clause that shifts payment to verified downstream acceptance.
Teams discover the missing permissions during incident reviews rather than design workshops because the original process mapping stopped at the handoff diagram. The diagram treated every interface as a simple data push.
Future automation requests now carry extra budget lines for credential audits and role testing that should have been part of the first deployment. Those lines remain optional until the next failure surfaces.