AI advances, and in particular AI Agents, have gained broad acceptance among banks and other financial institutions to the point where we are witnessing already a second generation of AI agents that deliver new levels of collaboration – both between AI agents and between people and agents.
In this recent episode of the What the FinTech Podcast, Paul Hindel, editor of FinTech Futures, and David Caruso, vice President of Financial Crime Compliance at Work Fusion, discuss the pace of AI acceptance in FinCrime compliance, what it is achieving, and where it is headed in its use of AI agents.
A noticeable attitude shift in the market
David explained that, in the last two years, there’s been a noticeable shift in banks’ acceptance of AI. “We know that because we’re in the business of developing AI agents and we have a lot of clients, we have a lot of customers, and we have no trouble having conversations with banks of all sizes all over the world who have shown real interest in it,” David stated. In fact, WorkFusion counts 10 of the top 20 banks as customers (as well as many others)—and that’s specifically within the realm of AI for financial crime compliance.
David explained that, as AI technology has improved, so has the perception of it among financial institutions. This is what has driven acceptance of first-generation AI agents in anti-money laundering (AML) compliance teams. AML leaders appreciate how AI agents have driven efficiency in their operations and changed the processes of gathering of information, organizing it, and presenting it via the creation of artifacts. To delve into how first-generation AI agents have brought efficiency to AML compliance ops, read our recent blog post How AI Agents Drive Efficiency In Financial Crime Compliance Operations.
The second generation of AI agents
Paul and David discussed how we’re now moving into a second generation of AI agents, which are doing more of the investigative reasoning as they analyze alerts that indicate potential crimes. David explained what he means by “reasoning” here: “Are there indications that these transactions are out of the norm for this customer? Is that indicative of just a change in their business or lifestyle, or is that indicative of perhaps financial crime, something nefarious that we need to look at?”
The second generation of AI Agents by WorkFusion can take the thousands of results that might come through a search on a company and narrow it down to just two or three things of value to examine at closely. This saves people (typically L1 analysts) from having to review hundreds of things. That cuts down their mundane, repetitive work by 80%.
Second generation also involves AI agents collaborating with one another. David explained that is the reason why WorkFusion puts people names to its current lineup of six AI Agents. Just like office workers, they speak with each other when required to work with information that only the other agent has. That’s why having them all work in complementary FinCrime compliance roles makes ultimate sense. These currently include:
- Tara: Transaction screening alerts review
- Evelyn: Name sanctions and PEP alerts review
- Evan: Adverse media screening
- Isaac: AML transaction monitoring
- Kayla: Customer ID verification and customer due diligence
- Edward: Enhanced due diligence
David also explained why all of these AI Agents must be great at collaborating with people. He underscored how each agent does a great job of narrowing down the volumes of information to surface only the most suspicious of items to be investigated by people. So, this elevates people to the higher-value role of analyzing only the alerts and situations that are deemed most risky to the bank and/or the overall financial system.
A good example of this is how Edward collaborates with people during enhanced due diligence. “He” uses IDP to automate the extraction and structuring of data from complex documents, then applies ML to analyze transactional data to identify patterns, anomalies, risks, etc., and leverages NLP to process and summarizes textual data from news articles, media reports, or corporate filings. Edward then uses GenAI to provide people (analysts) with written summaries, visualizations, and draft rationales based on analyzed data.
This type of agent-people collaboration is a hallmark of second-generation AI Agent workflows.
Where AI in FinCrime compliance is headed from here
With his technology expertise and constant involvement with regulators and compliance professionals, David has keen insights into what will come next for AI Agents in FinCrime compliance. This is where the interview turns truly fascinating for compliance professionals who want to plan for keeping ahead of nefarious actors in the age of AI. The conversation touches on the coming pace of change, bringing regulators into the fold, balancing the varied paces of the conservative banking world with Silicon Valley-like technology advances, and more. David puts it all in context and makes it easy to digest for non-technology compliance professionals. To hear everything David had to say and put your finger into the winds of change in FinCrime compliance operations, listen to the podcast here at FinTech Futures.