AI Agents serve a very specific purpose in financial crime compliance (FCC) operations. Not only do they drive massive efficiency gains, but they also help to minimize risk in anti-money laundering (AML) programs. In fact, these digital workers have been working behind the scenes in AML compliance operations at some of the World’s largest banks and other financial institutions for several years.
Unfortunately, the newfound hype surrounding AI Agents has muddied the waters around the definition of what they are. Suddenly, it seems that nearly every technology vendor has sprinkled ‘AI Agent speak’ into their marketing programs to the point that non-practitioners have become utterly confused about what AI Agents are, specifically:
- How they differ from bots
- Whether or not they are just another term to describe GenAI
- How much they incorporate the latest advances in AI, such as LLMs
To help bring clarity around AI Agents, recently, WorkFusion’s Chief Technology Officer Peter Cousins, VP of Financial Crime Compliance David Caruso, and VP of AI Solutions Kyle Hoback discussed how to define AI Agents and explained why they are so important in the battle against financial crime today.
AI Agents do collaborative ‘people work’
When asked to describe the difference between bots and AI Agents, Peter replied, “We usually bristle a little bit when people refer to our AI Agents as bots, because they are so much more than that.” He values bots for the speed and helpfulness they bring to step-by-step tasks that can be locked down and inflexible. But he points out that AI Agents operate in a completely different realm, mimicking the way people work (see the 30-second video below).
Peter also explained in the webinar how GenAI—in a complete contrast to bots—are very flexible and interesting but can go off the rails and put reliability, consistency, and explainability at risk in the context of FCC. He then delved into the dangers of equating AI Agents to LLMs and GenAI, highlighting the various ways that AI Agents operate more like people, even though they sometimes leverage LLMs and GenAI.
Explainability sets AI Agents apart
With WorkFusion AI Agents performing FCC operational work at 10 of the top 20 banks and many other financial institutions and FinTechs, David underscored the importance of regulators being able to understand the decisions and actions taken by AI Agents. After all, he speaks from experience as a professional who designed, built, and ran AML programs in the years prior to joining WorkFusion. “You have to provide all of the documentation that explains how these agents work,” he said. He then noted how there is absolutely no chance of using an AI Agent if regulators cannot understand it.
For their part, regulators are expecting banks and other FIs to leverage AI Agents now and in the coming years. David explained how different regulators in multiple regions of the world are evaluating FCC programs for their familiarity and usage of AI as a way to keep pace with criminal organizations that are using advanced technologies to perpetrate financial crimes. For details, watch the webinar now.
Peter then delved into how to demonstrate to regulators that your AI Agents are accurately and correctly making decisions, adjudicating alerts, and escalating to people as needed. He also disclosed the telltale signs of when an AI Agent-supported process is regulator-ready. He closed the regulator discussion by stating that it is very satisfying to know that field examiners have evolved to the point where their old question of “are you sure you want to use AI in such an important process?” has been replaced with their new question: “are you sure you don’t want to use AI in such an important process?”
When Agents use and don’t use AI
Despite their name—AI Agents—the agents which WorkFusion customers leverage each day to resolve millions of alerts and to perform KYC diligence do not always use AI. Surprising, right? Well, not surprising at all to people who have run FCC programs for years. In the webinar, Peter explained that each AML or FCC process has multiple phases, and some phases simply do not require the use of advanced AI. In fact, intelligent automations often deliver large percentage gains in operational efficiency long before advanced types of AI enter the picture. That’s not to say AI is never used in the early phases. It is, but in the longstanding and well-known forms such as machine learning, natural language processing, optical character recognition, etc.
Listen to the webinar as Peter explains where and how the different types of AI get phased in as a program’s use of AI Agents progresses. To illustrate his point, Peter describes the specific processes which WorkFusion customers first prioritize and how they use these ‘wins’ to progress to more complex AML processes. You will see that by starting small and gaining familiarity with the workings of AI Agents, the mysteries behind them melt away, and you begin to appreciate their intended uses and how they truly transform FCC operations.
Watch the webinar. It’s 30 minutes well spent to gain a solid understanding of what AI Agents actually are and why they are important to FCC.