TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

How AI Helps Financial Institutions Perform Customer Due Diligence

By utilizing AI-powered solutions to delve deeper into customer transactions and relationships, financial institutions can prepare for a more regulated future.

• By David McLaughlin
• August 30, 2017

Expensive and highly manual processes for conventional KYC (Know Your Customer) and customer due diligence (CDD) programs continue to challenge financial institutions. The process typically requires the financial institution to review and verify the account opening information and contact the customer to request additional information and documentation. If necessary, this is forwarded to a team to conduct appropriate risk-based checks on the customer that include negative news searches, employment verification, document authentication, etc.

For Further Reading: AI, Deep Learning, and Financial Services How Banks Can Use Their Data to Compete with Fintech Start-ups Understanding the Health of Financial Markets Through Semantic Technology

However, even when these processes run smoothly, traditional methods of verifying the beneficial or true owner(s) of an account routinely fail to identify connections between customers and their underlying motives.

Facing New Regulations

On May 11, 2016, the Financial Crimes Enforcement Network (FinCEN), a bureau of the U.S. Department of the Treasury, published a Final Rule designed to formalize new and existing CDD requirements for financial institutions. It codified four anti-money laundering (AML) provisions, or “pillars,” found in Section 352 of the USA Patriot Act and added a “Fifth Pillar” that requires covered institutions to establish risk-based procedures for ongoing CDD, to include (but not limited to):

• Understanding the nature and purpose of customer relationships for the purpose of developing a customer risk profile
• Conducting ongoing monitoring to identify and report suspicious transactions and, on a risk basis, to maintain and update customer information

Financial institutions utilize transaction monitoring systems (TMS) to address the existing requirement of providing sufficient controls and monitoring systems for timely detection and reporting of suspicious activity. However, the new Fifth Pillar requirements will compel financial institutions to revamp their current AML transaction monitoring procedures to allow for account-specific transactional review and analysis. Covered financial institutions must comply with these rules by May 11, 2018.

Understanding the Nature and Purpose of Relationships to Develop Risk Profiles

Most financial institutions take a client-centric view to understand the nature and purpose of their relationships but don’t often review account activity from both a client and individual account perspective. With the Fifth Pillar final ruling, financial institutions are now required to ascertain the purpose of the account, account operating behavior, and which products and/or services are in use, as well as the volumes and values of transactions. Account-level analysis enables financial institutions to improve their risk management while better understanding client needs.

Conducting Ongoing Transaction Monitoring to Maintain and Update Customer Information

The second requirement entails financial institutions to conduct ongoing monitoring to identify and investigate suspicious account activity and to maintain and update, from a risk perspective, client information. Although financial institutions do a respectable job with rules-based TMS, this requirement demands additional analysis of individual accounts’ actual activity against a benchmark of expected activity. This is called account activity review (AAR). It is not a substitute for TMS, but rather a complementary tool to provide additional financial-crime risk management.

Both of these requirements add a significant burden to financial institutions’ efforts to understand clients’ businesses and ensure that their account activity is aligned with the nature and purpose of each client’s relationships.

New Technologies to the Rescue

With new data science technologies -- including artificial intelligence (AI), machine learning, and big data -- financial institutions can reduce their risk of money laundering while supporting these new Fifth Pillar requirements. New AI-powered AML solutions leverage automated methods to analyze massive amounts of structured or unstructured data, then extract knowledge from the filtered results.

AI can help financial institutions reduce their AML risk -- including addressing Fifth Pillar obligations -- by facilitating the data collection process using external sources, such as public registrars, private databases, social media, and other unstructured data, to prepopulate data and streamline customer interaction. AI validation and verification solutions can then cross-reference customer-provided ownership and identity information with a variety of these sources to ensure its veracity.

Useful data science and AI strategies for supporting the new requirements include, but are not limited to:

Link analysis: The analysis evaluates relationships or connections between various types of objects (nodes), including people, organizations, and transactions. Link analysis is employed to augment traditional KYC and KYCC (Know Your Customer's Customers) processes, creating multilayer and hierarchical networks for relationships between customers, their organizations, suppliers, and business partners.

This relationship network is dynamically assembled from structured and unstructured data obtained from proprietary, open, and deep-Web sources. This augmented approach to link analysis can enable financial institutions to infer customer risk from the overall connectivity risk. Link analysis is also widely used in search engine optimization (SEO) as well as in investigations, intelligence, security analysis, and market and medical research.

Transactional analysis: Through a risk-based approach, data science evaluates a predetermined amount of transactional data and identifies suspicious activity related to the KYC subject. Data techniques “learn” what activity is expected and alert the KYC teams to anomalies before traditional TMS would detect the out-of-bounds transactions.

One transactional analysis technique, known as Triple Exponential Smoothing (TES), breaks down historical transactional data by trend, level, and seasonality components. Requiring as little as five months of transaction history, TES can determine whether future transactions are within the expected parameters or if they represent an anomaly (a significant and unexpected change in behavior).

Outside investigative sources: Using AI to examine email messages, phone numbers, addresses, and other customer information, anomalies can be quickly identified in customer activity through fraudulent account activity, social media monitoring, open source data, and other data points.

Unsupervised machine learning: This technique can be employed to identify customer segments and anomalies in customer data. Customers are grouped based on their jurisdiction, line of business, and expected net worth, which allows for better customer understanding and risk assessment.

AI enables financial institutions to go above and beyond traditional CDD efforts and predict what customers might do. Once a client portrait emerges, it becomes possible to predict or model future customer actions. By utilizing AI-powered solutions to delve deeper into customer transactions and relationships, financial institutions can prepare for a more regulated future. AI and data science technologies are uniquely suited to address the increasing country, customer, and product-specific regulations from state, federal, and international regulators.