top of page

Payments Compliance Just Lost Its Batch Window

8 hours ago
4 min read
Payments Compliance Just Lost Its Batch Window

By Jay Dembani, Founder at FinSaAIstra


A head of payments compliance at a mid-sized European bank used to have the luxury of time. A new monitoring rule could be drafted, tested against last quarter's data, and rolled out over weeks. The transaction itself was reviewed after it settled, or at worst, inside that evening's batch window.


Since 9 October 2025, payment service providers (PSPs) across the euro area have had to offer Verification of Payee (VoP) before a credit transfer is authorised. The check compares the payee's name against the IBAN and returns a match, a close match, or no match, and it has to happen in seconds, inside the same screen the customer is already looking at (European Central Bank, Instant Payments Regulation guidance). The same regulation moves EU sanctions screening from the individual payment to the customer: PSPs must check their customers against EU sanctions lists immediately after each new listing and at least once every calendar day, and must not screen the payer and payee of each instant transfer for those measures in addition (Article 5d).


Verification of Payee was not written as an AI mandate. It functions as one. A system that has to return a verified answer in seconds, on every transfer, at full transaction volume, is no longer a monitoring layer that flags things for a human to look at later. It is a decisioning system that has to be right, explainable, and fast, with no batch window to hide behind. Moving sanctions checks to the customer level takes one obligation out of the payment path. It leaves VoP and real-time AML and fraud monitoring carrying the rest.

This is where the two timelines diverge.


The first timeline is the one on the screen: seconds, per transaction. The second is institutional, and moves on a completely different scale. The EU's Anti-Money Laundering Regulation becomes directly applicable across member states on 10 July 2027, eight months from now, closing the gap that let member states interpret due diligence rules their own way (Herbert Smith Freehills Kramer, "What's done, what's due: the AML countdown," 2026). More than a dozen regulatory technical standards under that regulation already fell due in July 2026, so institutions still treating 2027 as the starting gun are behind. The new Anti-Money Laundering Authority, operational since mid-2025 and based in Frankfurt, starts selecting roughly forty high-risk institutions for direct supervision in July 2027, with supervision beginning January 2028 (Jones Day, "Direct Supervisory Powers of the New European Anti-Money Laundering Authority," 2025; eucrim.eu, "Tasks, Powers and Structures of AMLA").


A payments compliance function has to pass both timelines at once. Most transaction-monitoring stacks were not built for either demand, let alone both.


This is where the assumption breaks. Transaction monitoring has long been treated as a detection problem: find the suspicious pattern, raise the alert, let a human decide.


Verification of Payee is not a detection problem. It is an authorisation problem, built into the payment rail itself, with no later moment at which a human reviewer quietly fixes what the system got wrong. A false "no match" that a customer overrides anyway, or a fraud score that arrives after an instant transfer has settled, does not get caught downstream. The payment has already gone.


A monitoring engine that was designed to run in batches, overnight or hourly, cannot be made real time by adding a faster server underneath it. The logic itself has to change: matching has to happen inline with the payment message, risk scoring has to finish before the funds move, and the system has to produce a reason code a regulator can read, not just a score a compliance officer trusts. Institutions that treat this as a performance upgrade to the existing stack are solving the wrong problem.


Procurement conversations in payments compliance have mostly been about alert quality: fewer false positives, better case management, faster investigator throughput. That is useful, but no longer the whole brief. The real question for 2027 and 2028 is whether a vendor's matching and monitoring engine can hold up under a continuous, per-transaction obligation and explain each decision in terms a harmonised rulebook now makes checkable, with no human in the loop to catch what slips through.


Boards approving payments infrastructure spend over the next two budget cycles are not funding a better monitoring tool. They are funding whether the institution can answer correctly, every single time, with no second pass.


Compliance is becoming something a payment has to pass through, not something that happens after it has already gone.


That changes what "good" looks like in a monitoring system. Does real-time verification make transaction monitoring faster, or does it just move the backlog from the alert queue to the investigation team, one step later in the chain?



Sources

 
 
bottom of page