In a recently released webinar by WorkFusion, Putting AI to Work for Sanctions Screening, compliance operations professionals at banks and other financial institutions (FIs) can gain a quick understanding of how AI can transform their work – and the work of their teams – for the better.
Grant Vickers, Head of Financial Crimes Strategy at WorkFusion, noted that L1 sanctions alert review teams are under a lot of stress in today’s compliance operations environment as they struggle to mitigate the risk of missing something that really should be escalated. After all, more than 9,000 individuals and entities have had sanctions leveled against them since the start of 2022, and financial institutions (FIs) incurred fines of nearly $5 billion in 2022 alone for AML infractions and sanctions breaches.
“It’s a really big problem in this space,” said Grant. “I think it’s a really good use case for us to look at as we discuss Evelyn and Tara, two of WorkFusion’s AI Digital Workers that adjudicate level one (L1) alerts.” Tara adjudicates sanctions alerts for real-time payments and other transactions, while Evelyn performs sanctions screening alert review as well as adverse media monitoring.
The difference between AI Digital Workers and sanctions screening tools
Evelyn and Tara (and all WorkFusion AI Digital Workers) are not screening tools. Instead, Evelyn and Tara integrate with an FI’s screening tools. “For example,” he noted, “Looking at PEP adverse media and sanctions as a part of that screening, Evelyn can actually adjudicate those alerts.” Thanks to the pre-trained machine learning (ML) models that allow AI Digital Workers to disposition alerts, Evelyn can handle an alert in the exact same way a human analyst does. Evelyn picks up the alert, performs research around it from internal and external data sources, and quickly understands if the alert is a false positive or a potential true match that is worthy of escalation. “So, when WorkFusion says ‘Digital Worker,’ this is a personified software automation, fully software, and not people,” Grant clarified.
AI Digital Workers are further differentiated by the automated, intelligent, two-pronged justification they provide for alert dispositions. Evelyn and the other AI Digital Workers provide justification for dispositions in two different ways – one mathematical and the other a human-readable justification. Here’s Grant again: “We’re actually parsing the alert and feeding the information through a machine learning model and ultimately coming up with a mathematical confidence score as to whether or not something is a false positive or should be escalated.” That mathematical score accompanies a human-readable justification.
A deeper dive into how AI Digital Workers integrate with third-party sources
Both Evelyn and Tara integrate with commonly used third-party systems, whether they be paid sources through screening tool add-ons or internal sources to gather additional information – all in an effort to increase the likelihood that they can achieve a false positive decision. “We pipe in third-party enrichment data into the decisioning process,” said Grant. The system writes a narrative around an adjudicated alert, viewing it within the workflow and highlighting how it is ready for human use.
Evelyn and Tara both work with sanctions screening tools by:
- Integrating with a screening tool
- Ingesting alerts from the screening tool
- Using machine learning to process alerts
- Putting information back into the screening tool
Much of the work that AI Digital Workers perform happens behind the scenes, and it gets placed right back into your screening tool. “So, whether you have a 4-eye check at level one, or you allow WorkFusion to make a straight through processing (STP) decision for false positives, we’ll pass everything back to the screening tool or the case management tool,” noted Fergal Clarke, WorkFusion’s Senior Product Manager.
You can read more about how WorkFusion’s AI Digital Workers automate sanctions screening alert review in our eBook, The Risks of Not Automating Sanctions Screening.