Thousands of institutions are already reaping the benefits of migrating to ISO 20022. But some upcoming ISO 20022 deadlines require careful planning, including the move from unstructured to structured postal data. Here’s how we're helping financial institutions make this transition as smoothly as possible.
As the ISO 20022 roadmap continues to unfold, financial institutions are working to meet the standard’s requirements. One mandatory change that may catch many Swift users off guard – both from a compliance and operational perspective – is the move to structured postal data. By November 2026, banks must stop using unstructured postal addresses in certain ISO 20022 payment messages to avoid message rejection, fundamentally changing how address data is captured, stored and transmitted.
For example, Swift customers and participants in Market Infrastructures that are adopting ISO 20022 HVPS+ guidelines will need to complete a mandatory transition – from free-text address fields notably in debtor/creditor and agent fields in the MT 103 and pacs.008 payment messages, to the structured ISO 20022 CBPR+ format with field options for town and country.
The importance of structured postal data in the ISO 20022 migration
Structured address data refers to the standardised format used in financial transactions to ensure clarity and accuracy in address information. Structured addresses are vital for financial institutions as they ensure higher data integrity, streamline payment processing, and strengthen adherence to regulatory requirements, particularly in anti-money laundering efforts. This structured approach improves data quality and facilitates efficient processing of financial transactions.
Many existing systems still rely on free-text address fields, which are inconsistent and difficult to interpret. Converting these into structured formats (e.g., separating street, town and country) is not always straightforward, especially at scale. And there’s growing time pressure too: with the deadline approaching, users must update systems, processes, and data – an effort that can be both resource-intensive and technically complex. The risks of noncompliance are equally significant, ranging from rejected messages and payment delays to potential regulatory penalties.
For financial institutions, the migration offers an opportunity to enhance data quality, strengthen compliance and improve interoperability across the global payment ecosystem. The transition to structured postal addresses in ISO 20022 payment messages is a key step toward achieving these goals.
How AI is helping the financial community transition to structured postal data
Because structured address data is complex and time-consuming, it’s especially important that institutions get the transition right. We’re here to help the financial community meet ISO 20022 postal address migration deadlines. For the first time, we’re delivering an open-source AI solution, made available free of charge for the whole community.
The model is accessible to Swift users, service bureaus, vendors and Swift partners to use and build on. Integration is flexible – it can be built into internal systems, standalone tools, or used as a batch-processing engine to pre-process and clean unstructured address fields in legacy message formats.
Our Natural Language Processing-based AI solution enables the automatic extraction of mandated structured address elements – specifically town and country – from legacy unstructured data that can still be found in corporate and client databases, or in payments files received through corporate channels. It offers high-confidence predictions for the countries and towns found in the free format content, including confidence scores and multiple resolution suggestions for manual or automated review. And, by leveraging network-wide information that spans more than 200 countries and territories, it offers comprehensive geographic coverage, which has been successfully validated through pilot customers.
How it works
The model is a lightweight NLP-based AI system that can interpret structured address data (town and country) from unstructured fields (MT 103 fields 50 and 59). The output is limited to town and country predictions and customers have the option of configuring the system for regional tuning.
Built around a probabilistic graphical model, it’s well-suited for structured prediction tasks, where outputs are interdependent. Its primary objective is to extract town and country entities from unstructured address strings. Input data consists of tokenised address components enriched with linguistic and contextual features enabling accurate, consistent predictions at scale.
