WorkFusion’s financial crime compliance leader explains how AI and agentic automation are transforming AML, fraud detection, and regulatory efficiency.
Welcome to Business Insights Journal, David. We’re delighted to have you. To start, could you share a bit about your professional journey and what led you to your current role leading Financial Crime Compliance at WorkFusion?
My journey in Financial Crime Compliance started in 1996, and since then I have held various roles, including Chief AML Officer and Chief Compliance Officer at a national bank, founder and CEO of an anti-money laundering (AML) advisory and consulting firm, and Head of Sales and Chief Client Officer for a successful Regtech startup in the Risk Intelligence Data space. I’ve always sought ways to improve how financial institutions and organizations can improve and strengthen their AML and Fraud compliance programs, and I believe AI is the best way to do that. At WorkFusion, I can bring my three decades of experience in building and managing operations, along with my interest in modern technology, to help our team develop the best products and ensure our market teams share our vision and message with FinCrime executives and regulators.
Artificial intelligence is rapidly transforming financial crime compliance. How is AI modernization helping financial institutions strengthen their anti–money laundering (AML) frameworks and stay ahead of increasingly sophisticated threats?
Potential money laundering and fraud overwhelm compliance teams at financial institutions. Systems that detect possible financial crimes generate numerous alerts. Each alert must be analyzed to determine if it indicates actual money laundering or fraud. Fortunately, most of these alerts are false alarms, yet countless hours are still spent gathering, analyzing, and documenting evidence to meet regulatory requirements. This is where AI offers significant help. AI automates much of this record collection, sorting, and analysis, filtering out most non-suspicious alerts and saving time, money, and, most importantly, the attention of skilled compliance investigators. This allows their time to be focused on investigating real suspicious activity and the more sophisticated threats. The positive impact of AI in this area cannot be overstated.
Agentic AI—AI that takes autonomous actions—is being viewed as the next frontier in financial technology. How do you see this evolution changing the way banks and fintechs manage risk and compliance?
Much of risk and compliance work is what I call “programmable.” Picture a set of procedures that instruct a worker on what data is needed for activities like risk assessments, customer reviews, or determining if a flagged transaction involves a sanctioned entity. Many of these procedures involve simple yet essential steps to gather, organize, and analyze information. This is the type of work AI is designed to do – follow steps, process data and information, understand context, and present it in an easy-to-understand format. There are strong arguments that AI can reason and apply intelligence. But set that aside for a moment and just think of the benefit of AI completing all the initial work before a human needs to make a risk or compliance decision. AI can guarantee consistency in how data is collected, organized, and presented. This fundamentally changes the current state by freeing risk and compliance professionals to think more critically, focus more on important issues, and, when necessary, gather additional information.
Many organizations struggle with balancing automation efficiency with the need for transparency and accountability. How can AI be deployed in AML processes without compromising accuracy or regulatory integrity?
There is a misconception that AI operates independently of programming, logic, and controls. It is understandable because many AI applications people use in their non-work lives, such as creating fun videos, generating new recipes, or planning vacations, have a sense of “magic” to them. While this is also not accurate, it’s easy to see why the perception exists. Financial Crime Compliance AI Agents are built with the understanding that they must be auditable and withstand the rigor of a regulatory examination. This means there must be transparency about how they operate, including comprehensive documentation that explains all the logic and programming, as well as all the machine learning and AI models. No AI Agent should ever be implemented and deployed if how they work cannot be explained to a regulator. For an institution that is looking to purchase AI Agents, if whoever is selling that agent cannot provide an easy-to-understand explanation and supporting documentation, then don’t buy it.
With global regulators tightening their expectations around AI governance, what trends do you foresee shaping the future of regulatory frameworks for AI-driven compliance systems?
As AI adoption becomes more widespread and early adopters are subjected to rigorous examinations, regulators will grow more comfortable and accepting. Although AI is more advanced than rule-based systems from a decade ago, the same concerns about explainability existed when those systems were introduced 20–25 years ago. As more institutions purchased and implemented transaction monitoring and screening systems to detect potentially suspicious activity, regulatory examiners saw the same or similar systems at bank after bank. This created familiarity and eventually confidence. The same process will occur with AI-based systems. So, while strong governance is needed now and will remain so – ensuring that decisions to build or buy AI are sound and make sense for each institution’s risk profile – the level of concern about AI will subside.
Traditionally, financial crime compliance executives aim to ensure their programs sit comfortably in the pack of their peers. Regulatory examiners like to see one institution keeping pace with others of similar size and risk profile. However, institutions must realize that those in their “pack” that implement AI are moving forward more quickly. In the past, change was slow, and the pack could easily stay together. Now, if an institution falls behind because it is not yet implementing AI, the gap widens quickly, and so does the risk.
Data quality and explainability are major concerns in financial crime prevention. What steps should institutions take to ensure their AI systems operate responsibly and remain auditable?
Financial institutions can take several steps in this area. First, ensure AI systems used for financial crime prevention are auditable and, as I mentioned earlier, easy to explain. Strong data governance to ensure data quality is essential, as is the use of API-based integrations, which, when implemented properly, protect data integrity. Maintaining detailed documentation, including data sources, model logic, and audit logs, is necessary for regulatory review.
WorkFusion has been at the forefront of intelligent automation in compliance. Could you share a few examples of how your platform has helped clients modernize their AML operations and reduce manual inefficiencies?
Among the impacts of AI is that it creates new standards. Just as prior AML technology like transaction monitoring systems established the standard for detecting suspicious activity, AI is now establishing new standards for identifying AML risk. For example, it is known that one of the most reliable ways to identify AML, fraud, and even credit risk is by monitoring news. AML compliance history is full of instances where banks were unaware they were banking with clients involved in fraud or corruption allegations. If a regulator or law enforcement reads a news report about a bad actor and connects it to a bank, it can be a serious problem. With our AI Agent, called Evan, adverse media can be monitored in real time for every customer. Due to the large volume of customers and daily news articles, this was not feasible until now because it would have overwhelmed teams with too many alerts. Evan allows AI to read thousands of articles in minutes, identify if they concern a bank’s customer, and flag them for an AML analyst to review. This is just one example of how AI Agents are creating new standards. One new perspective on AML compliance is that anything once impossible due to volume is now manageable.
As a leader in the intersection of technology and regulation, what personal strategy or guiding principle helps you navigate the constant evolution of both AI innovation and compliance demands?
When speaking with Financial Crime Compliance executives, I constantly urge them to learn about AI and become conversant in the topic. It is rare for a head of AML or Fraud Compliance to have a purely technical background, yet they still need to understand how their teams’ systems work. Some aspects of AI and its related technologies, like Machine Learning and Natural Language Processing, can be complex, but they are understandable by everyone. It is the responsibility of these executives to acquire that knowledge and explain it to their teams, management, and regulators. In the past, it might have been acceptable to leave technology discussions to system owners or the IT team, but that is not advisable in today’s world of modern software. Having this knowledge is crucial, especially for management and executives.
What advice would you offer to financial institutions that are still hesitant to adopt AI due to perceived risks, legacy infrastructure, or lack of expertise?
At this point, if institutions are not yet implementing or are close to implementing AI as part of their financial crime compliance programs, they are falling behind their peers. When I hear of institutions that are not yet using or about to use AI, it sounds like they see the topic as overwhelming. While this is not new advice, when something feels that way, start small. The good thing about AI in the financial crime space is that there are applications targeting specific processes that allow for a small start. For example, sanctions screening is an excellent place to begin. AI in this area is explainable, transparent, and already adopted by many institutions. Importantly, when implementing AI, institutions should approach it like hiring new staff—until they are confident in the work produced, they should check everything.
Finally, David, as AI becomes more deeply embedded in financial systems, what are your final thoughts on how organizations can responsibly leverage agentic AI to create safer, more efficient, and future-ready compliance programs?
Regarding AI safety, as long as institutions have strong processes for reviewing, testing, and auditing new technology, they can be confident. It’s important to recognize that we are now at least three years into the “AI Agent” era for financial crime compliance. Many banks, including the largest, regional, and smaller local ones, use AI Agents for AML and fraud detection. Regarding efficiency, this is a fundamental aspect of AI – it brings unprecedented efficiency to AML and fraud operations. For operations that traditionally relied on large domestic staffs and, in some cases, even larger offshore outsourcing teams, AI is upending conventional operating models. By freeing people from much of the routine, manual, and repetitive work, this approach may be the best way to future-proof compliance programs – allowing staff more time to think, assess, and make critical decisions.
David Caruso
Vice President of Financial Crime Compliance
David Caruso is Vice President of Financial Crime Compliance at WorkFusion, where he is helping redefine how financial institutions fight financial crime using AI. With more than 25 years of experience, David has built and led AML and sanctions programs at banks including JP Morgan, HSBC, Wachovia, and Riggs Bank. He has also led major corruption investigations and founded and scaled RegTech companies. David is a former U.S. Secret Service Special Agent and a graduate of The George Washington University.
