The migration to ISO 20022 has given banks richer and more structured payment data, but standardising a message does not automatically make the information actionable. Data can remain trapped between schemes, payment systems, fraud engines and front- and back-office applications even after the format itself has been modernised.

That distinction is becoming more important as payments accelerate. Banks are expected to improve straight-through processing and transparency while detecting fraud earlier, supporting more payment rails and maintaining business continuity when infrastructure fails.

Natasha Lapierre, Global Head of Product Strategy and Innovation, Financial Messaging at Bottomline, described ISO 20022 as a foundational layer rather than the end of modernisation. The next stage, she said, is to break down the silos around that data and use it across payment, fraud and operational workflows. Her broader argument is that optionality becomes useful only when the operating environment can connect the choices banks have created.

ISO 20022 is the foundation, not the finish line

Lapierre said the industry has reached an important milestone in standardising payment data, but banks still have to turn that consistency into operational value. Structured data can support compliance, fraud detection, pre-validation and straight-through processing only if systems can access and act on it.

She pointed to pre-checks such as whether an account is open or able to receive funds as examples of information that can improve payment execution before a transaction is sent. At the same time, market infrastructures can retain jurisdiction-specific requirements, such as purpose-of-payment codes, so banks still need the ability to transform and enrich data as transactions move between systems.

Data silos weaken both payments and fraud controls

Lapierre identified several forms of fragmentation. Fraud indicators may sit in different systems for different payment schemes, preventing a bank from seeing activity associated with an account. Information embedded in a payment message may also fail to reach a separate fraud engine even when it could provide useful context.

She used an ISO 20022 request-for-cancellation message as an example: a reason code can indicate that a cancellation is related to fraud, but the value is lost if the messaging environment and fraud system do not share the information.

The operating challenge is therefore not only to collect more data. Banks need an environment in which payment connectivity, fraud controls and other systems can interoperate so that information can form a feedback loop rather than remain inside the application that generated it.

AI can make checks better rather than simply adding more of them

Lapierre identified fraud detection as one of the strongest near-term use cases for AI because it is fundamentally an anomaly-detection problem. AI can process large amounts of information quickly, identify patterns and pre-triage lower-risk cases so investigators can focus on decisions requiring human judgement.

Lapierre said fraud could already move beyond a purely human-in-the-loop model towards tightly bounded semi-agentic decisions, particularly in pre-filtering lower-risk cases. Moving further towards fully agentic decision-making would depend on governance and trust, which she said institutions are developing incrementally as they validate individual use cases. Explainability remains essential: a risk score is of limited operational value if an investigator cannot understand what produced it.

The combination of richer data and AI also changes the old trade-off between safety and speed. Historically, safer payments often meant more checks and therefore more time. Lapierre’s formulation is that safer payments can increasingly mean better checks rather than simply more checks.

That requires parallel progress in data and AI. Institutions need broader information, including appropriate data-sharing arrangements, while respecting privacy and governance obligations. AI can then improve alert quality and reduce the time required to handle individual alerts.

Optionality only becomes resilience when banks can switch

Lapierre said disaster recovery and dual-site arrangements remain essential, but the harder question in an always-on environment is business continuity when several dependencies fail at once.

Diversifying infrastructure can reduce concentration risk, but it can also create fragmentation if a bank cannot switch seamlessly between alternatives. In her view, optionality becomes operational resilience only when institutions have enough agility within a single operating environment to move between those options.

Traditional messaging and on-chain settlement are entering a coexistence phase

Lapierre said the question around blockchain-based settlement is moving from whether it will coexist with ISO 20022 messaging to how that coexistence will work. Banks face a fragmented set of on-chain networks and forms of digital money and may be reluctant to commit early to one model.

At Sibos, Lapierre announced that Bottomline and Chainlink had demonstrated interoperability between ISO 20022 messaging and on-chain settlement through Global Pay Connect, a platform banks already use. She described this as a key step as ISO 20022 and on-chain settlement begin to coexist. The design remains agnostic across on-chain networks and digital-asset types, allowing banks to avoid committing to a particular network or asset model before the market matures.

The demonstration provides a practical example of Lapierre’s optionality argument. Banks do not have to commit immediately to one on-chain network or form of digital money if the operating layer can connect different models without creating another silo.

The next test is execution

Lapierre’s closing point was deliberately practical. By Sibos 2027 in Singapore, she said the challenge would be to show concrete results and move from talking to doing.

That is also the test for investment in ISO 20022, AI and new settlement infrastructure: data has to move across systems, AI needs governance and explainability, and multiple rails need to provide resilience without creating more fragmentation.

For Lapierre, the next phase of payments modernisation is therefore less about accumulating more options than making existing and emerging capabilities work together.