By: Parvez Shaikh, Head - AML Practice, LTI

The AML transaction monitoring is based, among other criteria, on customer segments. These are derived from a wide range of information, including the customer demographics, transactional behavior, historical AML investigation outcomes, past law enforcement actions, negative news, etc.

Customer Segmentation

Traditional AML Compliance programs rely on time-consuming and manual processes to keep the customer segments up to date, where the subject matter experts collect and analyze relevant information and adjust the customer segments on quarterly or semi-annually basis.

The segmentation review and adjustment process can be automated through careful selection and preparation of data, to be used to train the cognitive models in order to identify the customer segment adjustments, which can be applied systemically. The automated customer segmentation process presents two advantages to the traditional process:

• It cuts down the review and adjustment cycle from three to six months to a single month or less
• It avoids the judgmental mistakes and inconsistencies from the subject matter experts

Faster and more consistent customer segmentation translates to reduced number of false positive alerts generated by the AML monitoring detection engines, which in term, translates to higher operational efficiency and lower AML compliance program cost.

Data collection for AML Investigation.

AML TM case investigation process begins with the collection of customer information from various sources, both internal and external to the financial institution. Traditional AML investigation process relies on large teams of analysts manually login to the source systems, perform searches, download the search results to files, extract and upload the relevant information to the AML case investigation system.

Such manual data collection process is time consuming, inconsistent and error prone. Automated data collection can be implemented using specialized BOTs configured to perform repetitive browser tasks, like accessing an URL, login with security credentials, navigating the pages, entering search criteria, extracting and downloading information from the search results, etc.

The automated data collection process presents a number of benefits:

• It is faster and more consistent
• It requires much less preparation time as opposed to the time needed to train the analysts to perform the same tasks manually
• It is more efficient, because:
• BOT’s can be run off business hours, holidays and weekends
• BOT’s can be scaled up and down according to the number of cases

It is more consistent and less error prone, as BOT’s perform same tasks same way, as opposed to a large team of analysts performing the same tasks.

AML Transaction Monitoring (TM) Case Risk Scoring

AML TM case investigation process is typically tiered and iterative, and it goes through multiple levels of research and assessment, depending on factors like customer risk, case complexity and other information that determine the likelihood of ending an investigation with SAR filling. By regulations, AML TM investigation is required to be completed within certain period, typically 60 days from the time when suspicious activities were identified, before SAR is filled with the regulators.

Traditional AML TM case investigation runs all new cases through level 1 review, that determines which cases would require further research and assessment by level 2 or Financial Investigation Unit (FIU).

The large number of new cases generated every month, combined with longer time needed to complete the research and assessment on high risk and high complexity cases, makes it paramount to identify the higher risk cases as quickly as possible. Automated case scoring solutions feed cognitive models with customer and case information, in order to identify cases, from a new batch, that are more likely to require level-two involvement. The benefit of automated case risk scoring is allowing more time to work on high risk cases and, consequently, improved ability to complete the review within the window mandated by the regulators.

AML TM Case Closure

Post AML TM case risk coring, certain cases may be systemically closed using automated process, which can be rule or model based. Both types of case closure process automatically close newly created cases that are similar to past cases.

The main difference between the rule and model-based methodologies are:

Rules

• Rules are provided by subject matter experts.
• Rules are specific and can be translated into definite logic and thresholds.
• Rules tuning is manual and time consuming.

Models (Algorithms)

• Models learn from historical case investigation outcomes and continue to improve and adapt when fed with new investigation outcomes.
• Models don’t involve specific thresholds.

In the case of rule based AML case closure, new cases created on customers with similar cases in the past, based on a wide range of criteria, involving customer risk, transactional and counterparty characteristics, past case dispositions, etc., are automatically closed.

In the case of model based AML case closure, relevant information is collected, prepared and fed to the models during the training phase. Upon completing the training, models apply the learned algorithm to identify the low risk cases.

The benefit of automated rule or model based case closure is improved process efficiency and cost avoidance, allowing valuable investigative resources to focus on higher risk cases.