Straight-through processing is the goal of every payment operation: a payment that moves from instruction to settlement without human intervention. In practice, a meaningful share of cross-border payments still passes through someone's hands. Each touch adds delay, cost and the possibility of error. Understanding why hands touch payments is the first step to removing them.

Why payments need repair

  • Incomplete or malformed data. Missing fields, inconsistent formatting or an unrecognised account structure force manual review.
  • Screening alerts. A match — genuine or not — routes the payment to an analyst.
  • Ambiguity in the instruction. Unclear beneficiary details or a mismatched reference.
  • Cut-off misses. A payment that arrives after a window may need re-entry on the next cycle.
  • Cross-system mismatch. Data that means one thing on one rail and something else on another.

Most exception handling is not caused by complexity but by disagreement between systems about the same payment.

What automation actually removes

Automation delivers most where the problem is data, not judgement. Cleaner reference data and consistent message standards remove the malformed-data category. Concentrating compliance in a single validation function removes the repeated-alert category, since one well-tuned check replaces several partial ones. Netting removes a category of exceptions entirely, because offsetting flows settle internally rather than travelling.

What automation removes less well is genuine ambiguity. Those cases still need a person — and a good design makes that person's decision cheap, by routing an over-limit operation to a human approver instead of rejecting it and starting over.

The structural point

Straight-through rates improve fastest when the payment path is shorter. Fewer hops mean fewer handovers, fewer opportunities for data to mutate and fewer screening queues. The goal is not to eliminate human oversight, which remains necessary for exceptions and approvals, but to reserve it for decisions that need it.

In short

  • Manual repair is usually a data problem, not a complexity problem.
  • Screening alerts are a major source of exceptions.
  • Standardised data and one validation function remove whole categories.
  • Netting removes exceptions by cutting the number of hops.
  • Humans should handle ambiguity and approvals, not plumbing.

This content is general information, not legal, tax or financial advice.

For the institutional view of automated processing, see /institutions/.