
Standing Settlement Instructions Go Stale
The account details were correct when somebody typed them in. Everything since then has been an assumption, and the assumption is checked only when a payment fails.
A standing settlement instruction is the stored record of where and how to pay a given counterparty: the beneficiary, the account, the correspondent bank and the settlement conventions for a particular currency and instrument. It is static data that governs live money movement, and it degrades in four ways: the counterparty changes bank, a correspondent relationship changes, an entity is restructured, or the instruction was entered for one currency and is used for another.
- SSI records are set up once under time pressure and reviewed almost never, yet every payment relies on them.
- Drift is silent. A stale instruction usually produces a delayed or returned payment rather than an error, so nobody investigates.
- Currency and instrument matter. One instruction reused across currencies routes through the wrong correspondent and generates unexplained charges.
- The verification source must be independent of the counterparty contact who asked for the change, which is the same principle as callback verification.
- ChatFin reads confirmations and counterparty documentation, compares them against stored instructions, and flags every record that has drifted or was never confirmed.
Standing settlement instructions are the least examined data in treasury and among the most consequential.
They are created when a relationship starts, usually by someone under pressure to get a payment out, from a document that arrived by email. From that point they are treated as fact. Payments route through them, confirmations reference them, and the record is reviewed only when something breaks. When something does break, the symptom is usually a delay or a return rather than an error message, and the resolution is a manual repair to that one payment rather than a correction to the record.
This page covers how SSI records degrade, why currency and instrument specificity matters more than most teams assume, what an independent verification source looks like, and how to review a whole population rather than the records that happen to fail.

The four ways an instruction goes stale
Degradation is not random. Four events cause almost all of it, and each produces a distinguishable failure pattern once you know to look for it.
Counterparty changes bank
The beneficiary moves banking relationship and tells whoever they deal with commercially. That message rarely reaches whoever maintains the settlement record.
Correspondent changes
The intermediary bank in the chain changes or exits a corridor. The beneficiary account is still correct and the routing is not, which produces delays and unexplained deductions.
Entity restructuring
A merger, renaming or reorganisation on either side. The name on the instruction and the name on the account no longer match closely enough for the receiving bank.
Currency reuse
An instruction created for one currency applied to another. The account may not accept that currency, or it routes through a conversion nobody intended.
Cause four is self inflicted and the easiest to prevent. An SSI is specific to a currency and often to an instrument, and reusing one across both is a shortcut taken during setup that produces charges nobody attributes back to it.
Why the failures stay silent
The reason these records survive uncorrected is that failure is soft. A payment on a stale instruction rarely bounces cleanly with a message saying the account has changed.
That last point is the practical fix worth building first. A simple count of payments requiring repair, by counterparty and by reason, converts an invisible problem into a ranked list of records to verify.

What counts as verification
The same principle that governs vendor bank changes applies here, and for the same reason. A change communicated through a channel controlled by the requester proves nothing about the requester.
| Independent source | A contact or channel established before the change request, not one supplied within it |
| Documented confirmation | A bank issued confirmation or an authenticated message rather than a formatted email attachment |
| Dual control | A second person approving the static data change, separately from whoever entered it |
| Test payment | A small value payment confirmed as received before full value flows, where the relationship warrants it |
| Record of proof | What was verified, against which source, by whom, and when, retained with the instruction |
The last row is what converts the control into evidence. An instruction with a verification date and source attached can be relied on. One without is a record whose provenance nobody can state, which is a different thing entirely.
Where the treasury stack sits
Treasury management systems hold SSI records properly. Kyriba and GTreasury maintain counterparty static data with currency and instrument specificity, and Bottomline manages payment origination and formatting against those records. Fides and similar connectivity providers manage the bank channel layer.
What none of them can do is know that the record has become wrong. A treasury system stores what it was told and applies it faithfully. The evidence that an instruction has drifted sits in payment confirmations, correspondent charges and counterparty correspondence, and those live outside the static data record in message archives and mailboxes.
Reading confirmations and correspondence and comparing them against stored records is document work plus matching, which is what ChatFin does alongside treasury automation.

Reviewing the population, not the failures
Repairing failed payments is reactive and endless. Reviewing the population is bounded, and it is smaller than teams expect because most instructions are dormant.
Rank by exposure and activity
Value paid through each instruction over twelve months. A small number of records usually carry most of the value, and those are reviewed first.
Flag the never verified
Any record with no verification date or source. These are unproven regardless of whether they have worked.
Flag the drift indicators
Records associated with delayed payments, unexpected correspondent charges or name mismatches on confirmations.
Verify independently and record it
Confirm through a channel established beforehand, and attach the source and date to the record so the next review starts from evidence.
Retire the dormant
Instructions unused for a long period should be deactivated rather than left available, since an unused record is an unmonitored one.
The last step matters more than it appears. A dormant instruction is a live payment route that nobody watches, which is the same risk profile as a dormant bank account.
How ChatFin surfaces the drift
ChatFin reads the payment confirmations, bank advices and counterparty correspondence that hold the evidence, and compares what actually happened against the stored settlement record for each counterparty.
The document processing agent extracts beneficiary names, accounts, correspondent routing and deductions from confirmations. The autonomous reconciliation agent matches those against the instruction that was used and reports mismatches: a name that differs from the record, a correspondent that is not the one stored, a deduction pattern that indicates unintended routing. The pattern recognition agent ranks counterparties by payment value and by intervention frequency, which produces the review list.
| Document processing | Reads confirmations, advices and counterparty correspondence in bank and message formats |
| Reconciliation agent | Compares actual routing and beneficiary detail against the stored instruction used |
| Pattern recognition | Ranks records by value and by how often payments against them required manual repair |
| Data query agent | Answers questions such as which instructions have never been verified and carry material value |
| Governed write back | Prepares the static data change with the verification evidence attached, for dual approval |
Static data changes are never applied automatically. ChatFin prepares the change with the evidence and routes it for the second approval the control requires, which is the same governed pattern described in adding AI without replacing your ERP.

A first pass in two weeks
The scope is bounded by value, so the work is finishable.
List instructions by value paid
Twelve months of payment value per settlement record. The top slice will cover most of the exposure.
Check verification provenance
For each, whether a verification source and date exists. Expect most to have none.
Pull the confirmations for those records
Compare actual beneficiary and routing against what is stored. Mismatches are your priority list.
Verify and deactivate
Confirm the material records independently, record the evidence, and deactivate anything dormant.
Teams running this typically find that a small number of records carry most of the value and that almost none of them have any verification history at all, which makes the first pass both quick and worthwhile.
A dormant settlement instruction is a live payment route that nobody watches. That is the same risk profile as a dormant bank account.
ChatFin
Frequently Asked Questions
What is a standing settlement instruction?
It is the stored record of how to settle with a particular counterparty: the beneficiary details, the account, the correspondent banking chain and the settlement conventions applicable to a specific currency and instrument. It is static data that governs live payments, which is why it should carry a verification source and date rather than simply existing.
How do settlement instructions become incorrect?
Four ways. The counterparty changes bank and tells a commercial contact rather than whoever maintains the record. A correspondent bank in the routing chain changes or exits a corridor. An entity is renamed or restructured so the beneficiary name no longer matches the account. Or an instruction created for one currency is reused for another, routing through a conversion or a correspondent nobody intended.
Why do stale settlement instructions go unnoticed?
Because the failure is soft. Payments on a stale instruction are usually delayed at an intermediary or charged unexpected deductions rather than rejected outright. Operations repairs the individual payment and moves on without correcting the stored record, and no report exists showing how many payments needed intervention or why, so the underlying problem stays invisible in aggregate.
Can ChatFin verify settlement instructions?
ChatFin compares what actually happened against what is stored. It reads payment confirmations, bank advices and counterparty correspondence, extracts the beneficiary and routing detail, and reports where the actual routing diverges from the instruction used. It ranks records by value and by intervention frequency to produce a review list, and prepares any static data change with the evidence attached for the dual approval the control requires.
See ChatFin compare your confirmations against stored instructions.
ChatFin reads payment confirmations and bank advices, extracts the beneficiary and correspondent routing that was actually used, and reports every divergence from the stored settlement record.
Static data changes are prepared with the verification evidence attached and applied only after the dual approval your control requires.
Book a DemoBring your settlement records and a month of confirmations.
ChatFin will compare actual routing against stored instructions, rank records by value and intervention frequency, and show which have never been verified.