AI Agents have been helping large and mid-sized banks and other financial institutions, or FIs (i.e. FinTechs, payment providers, etc.) to drive cost out of their AML operations for several years. This post discusses the three most salient areas where AI Agents have been making the biggest impact:
-
- Transaction monitoring alert reviews
-
- Sanctions screening alert reviews
-
- Adverse media monitoring
Transaction monitoring alert reviews
Most of the “people power” that goes into AML operations happens in the area of transaction monitoring (TM). This is due to TM requiring more complex analysis than do other compliance processes, such as customer due diligence. Transactions can introduce risk when they are performed in a way to intentionally obfuscate parties to the transaction, the purpose behind the transaction, and how the transaction may be related to multiple other transactions. The latter occurs in layering—wherein bad actors break down a large financial transaction into smaller, individual transactions to avoid investigator scrutiny. TM can gain complexity quite fast, particularly when it involves many parties (people and/or entities) to a transaction or set of transactions, with money passing across borders and spreading to multiple jurisdictions.
TM also involves high costs because of compliance leaders’ need to miss nothing. After all, if they miss a true crime, regulators can slap them with very public fines that cause brand damage and impose harsh remediation measures that add tremendous costs over several years. As the new WorkFusion eBook, Start Small, Win Big points out, the cost of TM operations has skyrocketed as banks and other FIs act on the “fear of missing something.” At the same time, the eBook discusses how TM systems that are tuned to miss nothing typically end up sending huge volumes of alerts to L1 analysts, with 95 percent of those alerts being false positives. This represents a massive chunk of the AML budget going to people investigating non-issues.
Where an AI Agent for TM drives out cost
Whether a transaction is simple or complex, banks and other financial institutions look to WorkFusion AI Agents to solve the back half of the Detection>Alert Generation>Alert Review>Decisioning process. That’s because the front half of the process, Detection>Alert Generation is already addressed by many tools. The majority of resource spend occurs in the back-end steps that involve labor-intensive Alert Review>Decisioning. That’s where all the people power has traditionally been dedicated, and that’s where banks and other FIs allow AI Agents to step in to make people resources more efficient.
Sanctions screening alert reviews
As a new conflict rages across the Middle East at the time of this writing, we are reminded that geopolitical issues and new/evolving sanctions put pressure on AML/CFT budgets. Which people and entities represent sanctions evasion risk, and how can people keep track of them all. Once again, being highly cautious, banks tend to tune their sanctions screening system in a conservative manner so that nearly everything receives a sanctions alert that requires review and decisioning. And like with TM alerts, a 95 percent false-positive rate is the norm. So, like with TM alerts review, an AI Agent represents the ideal solution to alleviate the burden placed on AML compliance staff. By shifting compliance teams away from low-value alert triage, AI enables investigators to focus on genuine risks while allowing financial institutions to optimize resources, eliminate backlogs, accelerate payments, improve onboarding, and drive business growth.
Adverse media reviews
AI Agents combine many aspects of AI (ML, NLP, OCR, GenAI, etc.), and they bring them all to bear in the adverse media review process of KYC (know your customer) operations.
KYC work involves suspicious activity or sanctions detection, investigation, and reporting. It is made up of five major activities:
-
- Gathering information: customer records, third-party data like adverse media or background information on a company or business.
-
- Organizing information: Organized information is easier for ops team to work with.
-
- Assessing the information: What do I have here? What are the relationships or their patterns of their activities?
-
- Reasoning: Attempting to decipher relationships and patterns.
-
- Judgement / Decisioning: Deciding if something seems suspicious. Does it need to be referred for investigation? Can the matter be closed? Etc.
AI Agent performs all of these activities by accessing data from a standard adverse media screening tool and reviewing it to prioritize news articles and other information about your customers and potential customers. Evaluating relevance, demographic data, level of material significance, and other attributes, the agent (prebuilt with deep industry and prior FinCrime knowledge) can comprehensively investigate a person or entity with as little information as just a name. The agent then highlights articles/information that indicate risk and passes the results to a team member for review. WorkFusion’s AI Agent, named Evan, also provides a detailed justification of the risk evaluation in written, auditable form.
The adverse media cost savings are substantial
A typical adverse media search yields five or more results, and each one takes a person between 10 and 20 minutes to read and analyze. AI Agent Evan automatically and rapidly reads them, identifies the people, parties, dates, amounts, and all other material information, and Evan does it in just 2-3 minutes. On top of reducing review times by 80-90%, Evan refers the riskiest cases to team members via human-in-the-loop technology. Once again, people are relieved of the tedious work that drives up budgets and they can focus on the riskiest cases.
In the end, whether it’s TM alerts review, Sanctions Screening alerts review, or Adverse Media review, every AML analyst or investigator has an AI Agent colleague that is alleviating the burden of tedious, costly work, freeing people to perform high-value investigations.
To learn more, request a demo of WorkFusion AI Agents in action.
