Pix is one of the world’s most widely used instant payment systems, reaching the large majority of Brazilian adults, and it runs around the clock. The Central Bank of Brazil set completion at up to ten seconds when the system launched. Now Brazilian banks are taking Pix beyond Brazil, with the first cross-border corridor live since March 2026 and more markets expected. That is the point where a domestic success story becomes an operational question for banks that have never processed a Pix payment.
Follow a single payment through and the reason becomes clear. A transfer that starts as an instant payment in Brazil does not stay one for long. It becomes a currency conversion, then a correspondent settlement, then a local payment in another jurisdiction, then a ledger posting somewhere else again. Each leg produces its own record, in its own system, on its own timetable, and each is a chance for the thread tying the payment back to its origin to break. The consumer sees one instant movement of money. Operationally, it is four or five records that have to be persuaded to agree with each other.
The natural instinct is to treat this as more of the same reconciliation work, and that instinct is worth examining. Correspondent reconciliation has always been statement-led: a statement arrives once a day, gets parsed into one record per line against the nostro account, matched automatically, and whatever survives is picked up by an operator. That end-of-day rhythm has worked because the underlying flows moved at roughly the same pace.
An instant rail does not produce a statement. It produces a continuous stream of instructions, confirmations, returns and settlement movements that must be correlated as they happen, which shifts the model from statement-led to event-led. This is a genuine change, not just a faster version of the existing process. Matching typically leans on the end-to-end payment identifier, amount and direction, falling back to timestamp, institution, account and reference where that identifier has not survived a leg intact. One caveat worth stating plainly: transfers between accounts at the same bank can settle internally, so settlement account reconciliation is always narrower than full transaction reconciliation, and the two should never be treated as the same exercise.
Then there are returns. Instant rails generate rejects, timeouts and refunds at a rate that reflects their volume, and each has to be related back to the original instruction rather than processed as fresh activity, or it becomes a break waiting to be discovered later. Handled manually, the exception queue grows in direct proportion to corridor volume, which is the point banks find themselves adding people rather than capacity, and the audit trail becomes something reconstructed under pressure rather than something that already exists.
Volume matters too. Recurring Pix payments have opened up subscriptions, utilities and business invoicing, and those flows behave nothing like one-off consumer transfers. Banks that scaled comfortably around the original use case have already felt the difference domestically. Layering conversion, correspondent settlement and return handling on top as corridors open is not a linear increase in effort.
It is tempting to wait for the architecture to settle before acting, especially with no final published specification for cross-border interoperability. But the flows already exist. The corridor is live, the legs are real, and the reconciliation is happening today in spreadsheets and manual matching more often than anyone would like. None of that gets easier by waiting. Whatever the long-term architecture turns out to be, it will land on whatever infrastructure a bank is running at the time, and the banks best placed will be the ones whose flows are already traceable end to end.
Which brings up the single most useful question a bank can ask itself, and it costs nothing to answer: when a payment originates from an instant rail in one market, does a common transaction identifier survive the conversion, the intercompany leg and the local payment leg? If it does, end-to-end reconciliation across the footprint is largely a matter of bringing the records into one place. If it does not, the first piece of work is building that thread, and knowing that now is far more comfortable than discovering it during an audit.
Two more questions are worth asking alongside it. Which entity actually processes the funding, the conversion and the settlement, because the answer is often not the one people assume. And if corridor volumes doubled next year, would reconciliation scale without proportional headcount, because that ultimately decides whether this is a technology conversation or a hiring one.
The practical priority is to preserve a traceable transaction thread across every leg before corridor volumes make manual work unmanageable. Aquila, Aqua Global’s ISO 20022 native financial messaging and reconciliation hub, supports that approach by sitting between a bank’s core platform and its payment networks. It normalises each rail into a canonical format, continuously reconciles instructions, confirmations, postings and settlements, and manages rejects, timeouts and returns as tracked cases with a complete audit trail, without replacing the underlying core system. If you are assessing what cross-border instant payments mean for your own flows, we are always happy to compare notes.
Learn more
For a closer look at how Aquila handles Pix and other instant rails, see the Aquila Pix factsheet.
