Skip to Content

Challenges and opportunities for aml & sanctions screening: modern technology is the only answer

Jeffrey F. Ingber
05 July 2024

The process of screening natural persons, legal entities, and transactions applies in a variety of AML contexts—including adhering to sanctions and identifying adverse media and politically exposed persons (PEPs). It’s integral to a satisfactory AML and sanctions program, but rife with errors, backlogs, improper decisioning, and outsized costs, and increasingly difficult for financial institutions to manage properly. These institutions are struggling to bring more efficiencies and better risk management to their screening systems, understanding that throwing human resources at the problem is not the answer.

The benefits of employing modern artificial intelligence (AI)-based technology to enhance screening processes are compelling, including retrieving relevant information, executing researches, analyzing data, making initial decisions on alerts, and generating and publishing detailed reports and an audit trail. Of note is that AI can be used to derive a very curated matching logic, one that’s far more advanced than the fuzzy logic method that’s been used for many years and allows for a better ranking of the probability that a match is a real one.

But how do institutions identify the modern, innovative tools that are best to enhance their screening systems, and then acquire and implement that technology in a seamless manner that doesn’t create additional risk? To help understand and address the challenges and opportunities in the screening arena, Capgemini, together with its partners Hummingbird and WorkFusion, hosted at the Harvard Club in New York City on June 20 an industry roundtable that included senior representatives from a range of financial institutions.

“It was a frank, valuable exchange of views. I was so impressed by the thoughtfulness and candor of these industry leaders in sharing information about the challenges they’re facing regarding their screening processes.”

Supriyo Guha, Senior Director & FCC Practice Lead, Capgemini

In the initial portion of the roundtable, the key issues related to existing screening processes were reviewed. They include:

  • The frequency of updates to the OFAC and other sanctions lists;
  • The strict liability associated with OFAC violations;
  • The complexity of recent activity-based Russia sanctions;
  • Dealing with various regulatory obligations and supplementary considerations across geographies;
  • Not having sufficient information on counterparties;
  • Budgeting and staffing constraints;
  • The extremely high level of false hits;
  • The huge amount of “noise” (i.e., insufficient or immaterial data – a particular problem with adverse media monitoring and dual use goods) that surrounds alerts.

Several participants mentioned the significant operational challenges in the trade finance space alone, including manual inputs from letters of credit and other documents resulting in numerous errors, and the lack of ability to identify and analyze key informational items in an efficient fashion. The attendees also noted that, in addition to sanctions, the handling of export controls poses an increasingly larger burden.

The discussion then moved to how financial institutions use, or plan to use, AI-based tools. It was pointed out that currently, AI intervention is applied primarily on the alert adjudication side and not to up-front screening, where it’s also needed. AI also is being looked to address the problem of duplication of due diligence reviews and to promote better sharing of information among teams and throughout the organization.

The group discussed how human productivity can be enhanced by AI, given that the traditional performance of screening reviews and alerts is an arduous process that, over time, can wear down and demoralize human analysts.

“Enabling individuals to collaborate with an AI-based system frees them from performing menial tasks such as copying, pasting, and data gathering and review, allowing them to work on higher-value investigations and, thus, be used in a more productive, strategic way.”

Art Mueller, Vice President—Financial Crime, Banking, and Financial Services, WorkFusion

Indeed, several organizations have replaced first level human review of screening alert hits with purpose-built algorithms for screening names and payments and analyzing them in real time. In complex situations, there’s a hand off to a human analyst, with alert review and transaction history provided in one place for efficient and easy review.

Finally, the conversation turned to lessons learned as to how best to introduce and implement modern AI-based tools. The process of incorporating AI into a screening system includes a number of steps, such as model selection, training, testing, and validation; integration of models with existing systems; user training and adoption; and ensuring continued compliance with all applicable laws and regulations. As with any AI system, a huge consideration is ensuring comprehensive, quality data. Data accessibility, sourcing, consistency, privacy, and security all are critical, along with integrating end-to-end workflows to allow for a seamless stream of information.

Implementation challenges identified by the roundtable participants included model governance, ensuring the ability to trace where the large language model is receiving its information from, resistance to change, skills gaps, legacy system compatibility, data security concerns, and user training and adoption.

Given the highly regulated nature of the financial industry, another key consideration is ensuring regulatory acceptance.

“The good news is that financial regulators globally have, in recent years, embraced AI-driven innovation as an appropriate if not necessary development in addressing financial crime.”

Joe Robinson, Co-founder & CEO, Hummingbird

In this regard, important aspects to regulatory acceptance were identified, including ensuring explainability, transparency, accountability, and proper model risk and data management.

In sum, the benefits of employing modern AI-based tools to enhance screening processes are compelling, and have been embraced by financial industry regulators. However, implementing these tools presents challenges that require careful planning, internal support, collaboration between IT and business units, attention to regulatory imperatives, and a strategic approach to ensure a smooth integration.

Meet our experts

Manish Chopra

Manish Chopra

Global Head, Risk and Financial Crime Compliance
Manish is the EVP and Global Head for Risk and Financial Crime Compliance for the Financial Services Business at Capgemini. A thought leader and business advisor, he partners with CXOs of financial services and Fintech/payments organizations to drive transformation in risk, regulatory and financial crime compliance.

Jeffrey F. Ingber

Senior Advisory Consultant, Risk and Financial Crime Compliance
A former ex-Senior Fed Official, Jeff runs Capgemini #RegDesk that helps clients stay abreast of developments in the FCC landscape and demystifies complex regulations into clear actionable insights. He provides a rage of advisory services to clients across the FCC lifecycle and helps them tackle the ever-changing global risk landscape.
Supriyo-Guha

Supriyo Guha

Senior Director, Financial Crime ComplianceCapgemini
Supriyo is the practice lead for financial crime compliance at Capgemini. He leads strategic industry-first initiatives to help clients transform their anti-financial crime functions and heads Go-to-market for Capgemini’s marquee FCC clients.

Peter Weitzman

Practice Lead, FCC Compliance and Risk Analytics