Nostro reconciliation where every break is explained from the payment's own messages, cited, and signed twice before money moves
Every nostro break explained from its payment messages and cleared with two signatures.
See one case, screen by screen ↓Nostro reconciliation is the daily matching of each correspondent bank's statement against the bank's own ledger for that nostro account, and the clearing of every line that does not agree. Nostro Reconciliation keeps the bank's reconciliation engine and adds seven agents that check the statements, match what the rules cannot, explain each break from its payment messages with every fact cited, and draft the recall, query or posting for an analyst to accept and a team lead to approve.
AI use cases in banking: a buyer's guide to agentic AI for bank and credit union operations →
Updated
Nostro reconciliation is the daily proof that every line on a correspondent bank's statement (MT950, MT940 or, more and more, ISO 20022 camt.053) agrees with the bank's own ledger. The reconciliation engine matches most lines. What it leaves behind, the breaks, is worked by hand: an analyst searches the payments hub, reads the payment messages and the SWIFT gateway log, sends an MT n99 or an email to the correspondent, and waits days for an answer.
Then the posting, the write-off or the recall is prepared, and the approval is chased by message. Aged breaks hide duplicate and lost payments, trap cash in overfunded nostros, and turn into audit findings when the controller certifies month-end from a spreadsheet. Nostro Reconciliation keeps the engine and puts seven agents around the breaks, so the analyst starts the day with each one already explained and cited, and decides.
Estimated split of the 20–30 minutes an analyst spends on a typical break by hand, against the time spent checking the agents' cited explanation and draft; our arithmetic, kept modest. Waiting for a correspondent's reply is not counted.
Seven agents check, match and explain every break from its payment messages; an analyst accepts and a team lead approves.
A worked nostro reconciliation example. Corvane National Bank's euro nostro at Hallerbank AG, Frankfurt, is out by EUR 158,414.60 on value date Friday, October 16. Here is what happens on Thursday, October 22, screen by screen, in the working solution.
Before Dana opens anything, the agents have written her morning, every figure cited. 45 of 46 statements are in; of 58,214 lines, 55,886 matched on their own and 2,146 were matched by the agents for her acceptance. 182 new breaks came in and 177 are already explained. Karoo Commercial Bank's ZAR statement has not arrived, and a request went at 07:10.
Dana opens the Hallerbank EUR nostro. The correspondent's camt.053 statement sits on the left, the ledger on the right, and 1,846 matched lines fold into one count she can open. Only four open breaks stay, each with its amount, its age, a one-line cause and the agents' confidence.
On Fri 16 Oct, Hallerbank AG's statement closes at EUR 41,276,310.62 and our ledger at EUR 41,434,725.22: a difference of EUR 158,414.60 in 4 open breaks.
The first break is a EUR 48,250.00 debit on the statement with no ledger entry, already escalated to the team lead. Hallerbank answered Dana's query at 06:52 with a camt.029: it received two payment orders and executed both. A minute later the investigator ran again, following payment PH-26-1016-0447 from Lakeshore Steel Fabricators to Ostrava Coil Works s.r.o. through the payments hub and the gateway.
Sent twice: the gateway timed out and resent payment PH-26-1016-0447 with a new UETR.
Dana clicks the citation on the gateway log. The payment left at 09:14:07, had no acknowledgement within 30 seconds, and was resent at 09:14:41 with a new UETR under resend rule R-12; both were acknowledged. The two pacs.008 messages sit side by side: different times and UETRs, the same end-to-end id and amount. The customer was debited once.
The recall to Hallerbank is drafted with the original UETR, EUR 48,250.00 value October 16 and reason DUPL, every field traced to a cited message, with the MT 192 equivalent for banks still on MT. A recall is a request: under UCC Article 4A, § 4A-211, canceling an accepted payment order needs the receiving bank's agreement. Sending stays locked until a team lead approves.
Dana accepts the explanation and sends the recall and its posting for approval as one item: Dr 1890-EUR Recall receivable, Cr 1140-2210 Hallerbank nostro mirror, EUR 48,250.00. On the approvals screen the item is hers to read, not to sign.
You prepared this, so you cannot approve it. A team lead approves.
Marcus sees the entries, the cited evidence and both signatures together, and approves. The camt.056 goes to Hallerbank through the SWIFT gateway at 09:34 and the suspense posting is made in the ledger.
Back on the account, the duplicate's break reads Recall sent. The ledger now closes at EUR 41,386,475.22, and the difference on October 16 falls to EUR 110,164.60 in 3 open breaks, each with its own proposal waiting.
The liquidity agent compares each nostro with its target buffer from the treasury forecast: $212.4M sits above buffer on six accounts. Elena proposes releasing GBP 36,700,000.00 from the Meridale Bank account, and it waits for Victor Han, head of treasury funding, to approve.
The day's trail shows Dana's acceptance at 09:31, Marcus's approval at 09:34, and the correspondence drafter (v1.9) sending camt.056 CNB-RCL-26-1022-0019. Every row carries its evidence, and the trail exports for examiners.
The same break, seen by the five people who carry it — what their day looked like, and what it looks like now.
Seven agents check, match and explain every break from its payment messages; an analyst accepts and a team lead approves.
Receives MT950, MT940 and camt.053 statements, checks the sequence, that the opening balance equals the last close and that the entries prove the close, quarantines exact duplicates and requests a late statement.
Matches what the rules cannot: one-to-many splits, netted charges, references truncated between MT and ISO 20022, and rounding, for the analyst to accept in bulk.
Explains each break from the statement, the ledger, the payments hub, the message archive, the gateway log, the payment tracker and the correspondent's replies, names the cause and cites every fact.
Drafts recalls, claims and queries in each correspondent's standard (camt.056, camt.027, MT n99 or MT 192) and links the reply back to its break.
Proposes balanced postings, reversals, suspense moves and write-offs with reason codes, and routes each by amount to the right approver.
Compares each nostro's balance with a target buffer built from the treasury forecast and proposes releases, with the forecast cited.
Runs the 5- and 30-day clocks, escalates, writes the morning brief and keeps the month-end certification pack current.
Accept, approve, certify. The agents propose; a named person decides.
The analyst can ask in plain words; each answer cites its source and opens the screen it is about.
Why is the Hallerbank EUR nostro out by EUR 158,414.60 on 16 Oct?
Four open breaks make up the whole difference: a EUR 48,250.00 debit sent twice after a gateway retry (recall drafted), a EUR 120,000.00 ledger credit on the 19 Oct statement with value 16 Oct, a EUR 9,870.40 credit that names no customer and needs a query, and EUR 35.00 of Hallerbank charges. 48,250.00 + 35.00 + 120,000.00 − 9,870.40 = EUR 158,414.60.
Which breaks reach the financial controller by Monday?
Three cross 30 days by Monday 26 Oct: PLN 22,400.00 at Bank Wisłoka (standing order never loaded), HKD 156,000.00 at Victoria Peak (inflow sent to another nostro) and DKK 18,900.00 at Kattegat (FX difference). Two have a ready proposal; the Hong Kong one needs a second query.
How much can treasury release from overfunded nostros today?
$212.4M across 6 accounts sits above the target buffer. The largest are the Meridale GBP account (£36.7M, about $48.9M) and Wattle Bay AUD (A$50.1M, about $33.0M). Each release is proposed by Elena Sokol and approved by Victor Han.
Show every duplicate after a gateway retry this month
Three in October, all after a timeout and a resend with a new UETR: EUR 48,250.00 at Hallerbank on 16 Oct (recall drafted), GBP 7,415.00 at Meridale on 14 Oct (recall waiting for approval) and MXN 412,000.00 at Cuatro Vientos on 7 Oct (recovered 9 Oct). The cause is resend rule R-12 in the gateway; a note to Payments Operations is drafted.
13 screens from the working solution, on its sample data. Pick one to see it large.
Rules, tolerances, clocks and approval limits are settings. Every change is saved as a new version and approved by a second person before it runs.
| Setting | Default | Choose from |
|---|---|---|
| Matching rules | R-01 to R-05 on: exact, UETR, one to many, netted charges up to EUR 50.00, reference repair | R-06 across value dates: the agents propose, a person accepts; describe a new rule in plain words |
| Amount tolerance for a match | EUR 0.01 | Stepped; applies to rounding on converted amounts |
| Escalation | Team lead after 5 days, financial controller after 30 | Stepped, per policy |
| Matching runs | 23:00 and 06:00 ET | A statement after the cut-off waits for the next run |
| Correspondent query reply due | 2 days | A reminder when it passes; the team lead hears after a second miss |
| Statement late after | 3 hours | Measured from each correspondent's usual arrival time |
| Approval limits | Write-offs over $10,000 and adjustments over $250,000 also need the financial controller | A person never approves what they prepared |
| Retention | 7 years | BSA records at least 5 years; deletion needs the controller and the administrator |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
Sources: SWIFT, CBPR+ roadmap (ISO 20022 for cross-border payment instructions since November 2025; MT9xx statements continue for now) · UCC Article 4A, § 4A-211 (Uniform Law Commission; cancellation of a payment order) · 31 CFR 1010.430 (BSA record retention) · "demo" = seen in the working solution, on its sample data · "target" = design goal, measured in the live solution · "estimated" = our arithmetic from the demo bank's volumes, kept modest. Corvane National Bank, its correspondents, customers and accounts, and the people named on this page are fictional — characters and sample data in the working solution.
Nostro reconciliation is the daily matching of each correspondent bank's statement against the bank's own ledger for that nostro account, and the clearing of every line that does not agree. Nostro Reconciliation keeps the bank's reconciliation engine and adds seven agents that check the statements, match what the rules cannot, explain each break from its payment messages with every fact cited, and draft the recall, query or posting for an analyst to accept and a team lead to approve.
A break is a statement line with no matching ledger entry, or a ledger entry that is not on the statement. Common causes are duplicate payments after a gateway resend, correspondent charges not booked, value-date differences, rejected or returned payments still booked, unapplied credits and FX differences. Each break here shows its amount, age, cause and the agents' confidence, cited to its sources.
Statements arrive overnight; the intake agent proves each one, and rule matching runs in the engine. The matching agent proposes matches for splits, netted charges and repaired references, accepted in bulk at 0.97 or above. The break investigator explains what is left, the correspondence drafter and ledger correction proposer draft the action, the analyst accepts or rejects with a reason, and a team lead approves before anything posts or is sent.
Yes. In the working solution 46 nostro accounts in 19 currencies and 12 vostro accounts, the accounts other banks hold with Corvane, are reconciled the same way, the vostro side against the statements the bank sends its respondents.
Yes. It reads MT950, MT940 and camt.053 statements and shows each correspondent's format: 22 of the 46 nostros in the working solution send camt.053 and 24 send MT950. Recalls, claims and queries are drafted in whichever standard the correspondent reads, camt.056 or MT 192, camt.027 or MT n99, and the agents repair references truncated between MT and ISO 20022.
No. The reconciliation engine stays and is a connection, with its own rule versions. The agents work on what the engine leaves: they match the harder items for a person to accept, and explain and draft the action for every break.
Yes. The agents may not post, send, approve, move funds or contact a customer. An analyst accepts every explanation, a team lead approves every posting, write-off, recall and claim, and write-offs over $10,000 also need the financial controller; the person who prepared an item can never approve it.
The Agentic Solution Engine builds and deploys it from your requirements and documents, such as your reconciliation policy, sample statements, ledger extracts and payment messages, and it goes live once every quality gate has passed. We will walk you through it on your own breaks first.
Examined twice, cited to the rule, decided on banking day two of five.
1-min demoBank Guarantee & Standby LC IssuanceEvery clause tied to a model form and a rule, and no instrument issued without an end date.
1-min demoTrade-Based Money Laundering DetectionEvery trade alert decided in days, with the evidence that decided it.
1-min demoWe'll run Nostro Reconciliation on a sample of your own statements, ledger extracts and payment messages.