Today’s financial institutions regularly collect, store, and maintain customer information in a series of files, systems and databases. Employees operating throughout the financial institution then update and use the data across various lines of business. These data manipulations result in a mix of redundant, non-standardized, incorrect and incomplete versions of the information spread throughout the financial institution.
An anti-money laundering (AML) transaction monitoring system’s (TMS) ability to function efficiently depends heavily on a financial institution having clean and consistent data. However, when you consider the immense volume and variety of data flowing into today’s institutions, addressing incomplete and duplicate data presents a challenge that, while time-consuming, is essential for the reduction of false positives.
The process of front-ending the AML monitoring process with a data cleansing and enrichment solution increases the accuracy of an institution’s data matches, therefore making investigative efforts more effective. Additionally, financial institutions struggle with the quality of external data sources which may having missing fields or partial info. The process of data cleansing and enrichment can help institutions update these fields with correct data.
So how can financial institutions utilize new technologies to streamline their data collection and AML process?
Advanced artificial intelligence (AI) algorithms have been found to excel at identifying missing or erroneous information and can be used to correct a mistake, or locate missing information. For instance, a common issue found in data sets is the omission of a country name or country code in an address field. AI algorithms can quickly detect that the country is missing and determine the code from the address and other information provided.
When evaluating suspicious behavior, a key facet of detecting a potential financial crime is the AML investigators ability to review the full set of activities an entity in question has participated in. Unfortunately, data quality issues often make straight forward matching on certain fields, such as name and address, insufficient to properly group all activity related to an entity.
For example, a human will recognize that William Smith and Bill Smith could be the same person, whereas unintelligent matching software will treat these as separate entities. AI agents can account for these nuances in addition to dealing with issues such as typos/misspellings, word order (Bill Smith vs. Smith, Bill), and missing or redundant information. This work facilitates the process of entity resolution which can uniquely identify the transacting parties. Once complete, a full and complete picture of each entity’s activity can be provided for TMS analysis.
Tackling Data Enrichment with Artificial Intelligence
In terms of data enrichment, AI systems have the essential capacity to create observables or data points that rules-based TMS are simply unable to create on their own. Many of these observables center around detecting anomalous behavior. In order to do this, a baseline of normal behavior must be created for each transacting entity, something that a TMS can’t do. AI systems are able to develop individual profiles of “normal” behavior for each entity and can detect when activity differs notably from the norm.
For example, being able to detect that an entity moved three times the normal amount of money through their account in the last month than they have in the past year is an invaluable piece of information that AI algorithms can unveil. Other techniques allow the AI agents to determine if the transaction activity is consistent with known patterns of business relationships, thereby facilitating the identification of anomalous activity. Supplementary techniques allow the creation of a network of relationships between transacting entities and other parties which can reveal further indications of risk. These “flags” can be provided to the TMS for consideration in the overall assessment of entity and transactional risk.
Artificial intelligence-based systems are capable of drawing on a wide array of data sources to create these “flags,” including unstructured web-based resources such as articles and social media data. These sources can support the collection and correction of missing information, or the standardization of erroneous information as well as provide significant insights that would otherwise be unavailable with the sole use of traditional TMS.
Reducing False Negatives with Data Cleansing and Enrichment
In addition to reducing false positives, a data cleansing and enrichment solution can have a significant impact also on decreasing false negatives, or missed financial crimes. Data cleansing allows financial institutions to see all activity on an entity so that nothing is missed while data enrichment provides new risk observables around anomalous behavior that can’t be accomplished with a TMS.
Artificial intelligence solutions exist today that enable institutions to cleanse and enrich their data through a comprehensive entity resolution and entity verification process that normalizes the data before it enters a TMS. The fundamental takeaway for financial institutions to have when considering the use of AI is that placing AI before their TMS allows them to avoid disrupting their entire current AML process. AI is a cost-efficient solution built to enhance AML investigative efforts and thwart more global crime.