Skip to the main content.

3 min read

Jurisdiction Derivation, Powered by AI, Helps Financial Institutions Reduce Risk and Their Number of AML Investigations

Financial institutions are held accountable by regulators to ensure they are taking a risk-based approach in their AML/BSA compliance operations. As such, institutions must consider AML risk based on certain types of customers and transactions, including risky jurisdictions impacted by political or economic unrest.

Financial institutions struggle to identify transactions and transacting parties related to risky jurisdictions because jurisdiction/country code information in SWIFT messages and international ACH payment transactions is often malformed or missing altogether. This reality is further frustrated because each country’s addressing system has its own set of unique conventions, making it difficult to map to geographic fields.

Current Approaches to Jurisdiction Derivation Are Failing

Numerous other issues plague financial institutions grappling with jurisdiction risk. For example, when reviewing data located in fields within SWIFT messages, there is often no delineation of where the name stops, and the address starts. In some instances, a name will be on the first line with the address on the second and third lines. Other times, the name is on the first and half of second line with the address starting in the middle of line two. Name-address disambiguation is required to simply isolate the address.

Furthermore, the financial services domain has its own complexities/conventions that make an address difficult to determine.  For instance, rather than entering the entity (originator or beneficiary) address, the SWIFT code or the ABA routing number of the entity’s bank may be used. There are also transaction-type message formats that must be considered.  While the SWIFT code or ABA routing number can aid in properly identifying the country code, they must be coded explicitly to ensure complexities and issues do not arise.

Even when the address in a message or transaction can be identified, the problem can be further complicated if the address is malformed when manually entered.  Omissions, misspellings, extraneous data, abbreviations, and other issues make determining the country code challenging.

There are also inconsistencies in language-specific representations. Take, for example, that a French bank may write “ETAS UNIS” for United States and “Cote d’Ivorie” for Ivory Coast. Even when you have something that is standard, the codes used (such as ISO 3166-1 Alpha 2, ISO 3166-1 Alpha 3, ISO 3166-1 Numeric, FIPS 10-4, etc.) must be deciphered. What’s more, standards evolve over time as regional geopolitical changes take place. Is the West Bank part of Israel or Palestine, or is it being treated as a separate region altogether? While there may be a technically correct answer to this, bank employees who are manually entering this information are making decisions about which code to use.

Human Remediation is Expensive and Time-Consuming

When jurisdiction fields are unknown due to any of these issues, a human analyst is often brought in to scour Google and other resources in an attempt to identify the location of the transacting parties based on the information that is available.

If no jurisdiction or country code is discernible, or there is no human resource available to do this work in a prescribed amount of time, the transaction goes into a bank’s TMS with the jurisdiction/country code field blank. Transactions missing these vital details are required to automatically default to the highest risk jurisdiction resulting in TMS alerts that trigger time-consuming and costly AML alert investigations on transactions that are often legitimate. According to industry estimates, approximately 95 percent of all TMS alerts are false positives and missing jurisdiction information or country code designation contributes significantly.

AI Offers Answers

In order to deal with the diversity and complexity of issues that stand in the way of efficiently identifying jurisdiction, QuantaVerse offers an AI-powered solution that is based on a multi-layer approach where a number of models and techniques are employed to determine jurisdiction/country code. As there are many hurdles to accurately identifying jurisdiction, there isn’t a single magic algorithm to address the issue, but rather a series of models developed that address the challenges described above. A number of models and techniques (among them is fuzzy matching) are critical to meaningfully cutting the cost and time spent running down country codes.

In aggregate, this model-of-models solution delivered by QuantaVerse is able to derive country codes for nearly all entities analyzed.  QuantaVerse recently performed country code analysis for a large U.S.-based bank and found that 80% of its dedicated jurisdiction fields were blank.

“Prior to working with QuantaVerse, we conducted country code derivation manually, representing thousands of transactions,” the bank’s Chief Compliance Officer explained. “Analysis and quality control usually take three full days to complete in the traditional manner and, due to time constraints, covers only 75% of the wires that are missing the country code. QuantaVerse’s multi-layered approach to country code analysis helped us derive correct country codes for nearly all entities, saving us an incredible amount of time while cutting false positives.”

QuantaVerse is now offering the country code derivation capability of its Pre-TMS Entity Resolution & Risk Scoring solution for trial. The offering gives financial institutions an easy way to trial AI-enabled automation.

For more information on QuantaVerse’s country code derivation capability, please visit: