By Adam Famularo, CEO of WorkFusion, a UiPath company
Today, financial crime compliance has become one of the most resource-intensive and scrutinized functions in banking. Each process produces enormous volumes of alerts. Each alert must be reviewed, documented, and resolved in a way that can withstand regulatory examinations.
Most of those alerts are false positives – upwards of 99 percent. But the work still must be done. And this reality is exhausting for teams.
For the past 25 years the process has barely evolved, banks have tried to manage financial crime compliance the same way: add people, outsource work, and lay new tools on top of old systems. This approach has increased cost and complexity without fundamentally changing the workload. The result is a model in which analysts do the same labor-intensive work of reviewing alerts while highly trained investigators spend much of their day gathering data from multiple systems, organizing information, and writing case narratives to satisfy documentation requirements.
This is where Agentic AI is making a real, measurable difference at top banks and leading financial institutions around the world.
A surprising amount of financial crime compliance work is what I would call “programmable.” Regulations require step-by-step procedures for investigating potential money laundering or financial crime. Analysts follow defined processes to gather data, compare names and identifiers, check transactions, search news sources, and document findings. These steps are necessary, but they do not always require human judgment.
AI is very good at following procedures.
In payment sanctions screening, for example, systems flag potential matches to sanctions lists. Analysts then determine whether the match is real or a false alert by reviewing names, addresses, dates of birth, transaction details, and historical data across multiple systems. AI agents can perform these steps in real-time, assemble the relevant information, and present it to a human reviewer with a suggested resolution. In many cases, this removes more than 80 percent of the manual effort that used to come before the decision.
The same is happening in transaction monitoring investigations. Historically, an eight-hour investigation might involve four or five hours of searching systems, gathering documents, and organizing information before analysis even begins. Agentic AI can automate most of this preparation. Investigators start with a complete picture of the customer, the transactions, and relevant external context. Investigation times can drop by 50 percent or more, and investigators can handle several times the volume they once could.
Adverse media monitoring is another powerful example. One of the most reliable ways to identify financial crime risk is to monitor global news for mentions of customers involved in fraud, corruption, or criminal activity. The challenge has always been scale. The volume of daily global media made continuous monitoring impractical for most institutions. Agentic AI changes that by reading tens of thousands of articles in minutes, identifying which ones relate to a bank’s customers, and presenting only the relevant information to analysts.
This is already happening inside many banks. These changes are doing more than improving efficiency. They are beginning to redefine what standard practice looks like in financial crime compliance.
Regulators have historically expected institutions to remain in the middle of the pack relative to peers of similar size and risk profiles. That worked when change happened slowly. Today, banks that adopt AI for these processes are moving forward much faster than those that do not. The gap is widening, and so is the operational and regulatory risk of falling behind.
Importantly, in many cases, regulators are encouraging institutions to explore AI as a way to improve outcomes, provided the systems are explainable, auditable, and well governed. The same concerns about transparency that existed when transaction monitoring systems were introduced 25 years ago are appearing again and will likely follow the same path toward familiarity and acceptance.
The critical requirement is explainability. Institutions must be able to show exactly how information was gathered, what logic was applied, and how a conclusion was reached. When Agentic AI is designed to mirror the procedural steps that analysts already follow, explainability becomes not only possible, but also easier to document consistently.
This shift is also changing how compliance teams are structured. Traditional models relied on large groups of junior analysts to manage volume, with senior investigators handling complex cases. As Agentic AI absorbs much of the initial review work, human expertise is shifting toward higher-judgment analysis. Compliance teams are starting to rethink how they staff, train, and develop talent since the nature of the work itself is changing.
This is something we’re seeing more broadly as banks modernize financial crime operations. Compliance leaders don’t view Agentic AI as a replacement for investigators, they describe it more as an AI co-worker. “They” prepare cases, gather and organize information, and document findings with consistency, while human analysts remain responsible for judgment and final decisions. This framing resonates because it reflects how the technology is actually being used in practice.
In many ways, AI is beginning to establish new norms in financial crime compliance, similar to how sanctions screening and transaction monitoring systems became standard over time.
For banks that still see AI as an innovation exercise, the reality is more practical: it’s about fixing a financial crime compliance model that just doesn’t scale anymore.
About the author
Adam Famularo is the CEO of WorkFusion, a UiPath company, where he leads the development of AI agents designed for financial crime compliance in banking and financial services. He is a visionary business leader dedicated to customer and partner success, building world-class teams and delivering great returns for shareholders and investors, with prior leadership roles at erwin Inc., Verizon, and CA Technologies. Adam is a graduate of the General Management Program and Harvard Business School and holds an MBA from Dowling College and a bachelor’s degree in business economics from SUNY Oneonta.
Sponsored feature
