top of page
Writer's pictureNguyen Nguyen

Throwback from FinCrime event: AI Integration Took Center Stage


On March 31st, Complidata, Huron, and Oracle hosted a successful networking lunch titled "FinCrime Busted: How AI/ML Fuel Financial Crime Prevention." The event brought together industry experts and leaders from renowned financial institutions and fintech companies for lively debates and discussions on a range of topics, including using AI for ATL/BTL Threshold and Tuning, Effective Risk Detection and Control, and Sanctions and Transactions Monitoring. Despite the diverse perspectives in the room, there was one common question on everyone's mind, bankers and fintech business owners alike: while AI offers significant potential, how do we effectively integrate it into existing systems and operational processes?


The challenges of AI Integration: Regulators, and how to deal with them


The toughest obstacle when integrating AI into banks’ existing systems and processes, as agreed by most in the room that day, was convincing the regulators that it works. The current risk-based guidelines suggest that it is acceptable to “miss” something, as long as it is not “obvious”. What counts as “obvious”, one may ask. The only obvious thing here is the regulators’ expectation: AI is supposed to offer greater insight and cover; therefore, nothing should be missed at all.


Unfortunately, that is not yet the case. The most advanced AI model can increase SAR productivity by 50%, however, it does not guarantee detection of all discrepancies, in fact, it is far from that. Just last year, Germany had a spike in alerts, related to a football club. The AI spotted the discrepancy in the club’s name, Borussia Dortmund. Although the technology is advancing at a phenomenal pace – our AI model has been growing at a similar rate to one of ChatGPT, incidents like these are inevitable, giving reasonable grounds for doubts from the regulators.


With that being said, it does mean that there is no way to get a thumbs-up from regulators to integrate AI into AML Operations, nor does it mean that regulators are all against it. In fact, more and more are starting to jump on board. In Hong Kong and Singapore, two “havens” of money laundering, the governments are mandating the use of AI in AML Operations. The US government takes a lesser stance; however, they still encourage the use of technology in detecting financial crimes. These few examples show that governments are open to innovation and changes, as long as banks can make a case for it. The biggest issue hindering that case is understandability. You cannot present lines of code to a regulator and expect them to understand how it works. Demonstrable progress and effective AI management, using tools like Explainable AI, help to facilitate comprehension between all parties. Regulators also consider the sustainability of a model, how it lasts against model drift. Banks and fintech companies need to pay attention to that when building and integrating an AI model into their existing systems.


It is worth the effort: The indisputable potential of AI


AI Integration is indeed challenging; however, it is surely worth the effort. For a start, AI-based systems can be implemented at a much lower cost than traditional AML platforms, saving banks possibly millions of euros per year on wages for big teams of senior compliance officers. It is necessary to point out that AI models are not looking to replace human compliance officers – they rather act as co-pilots that help to “unlock” more data, and reduce human errors. It saves time, so fewer people are needed. With advanced, explainable AI models, for example, the one developed by Complidata NV., the system can be operated, and the predictions understood by junior compliance officers, so the bank does not need to spend money on more experienced, and thus more expensive staffs. Apart from the obvious benefit of reducing the number of false positives, a reduction in human errors and document processing time also helps the bank to increase its customers’ satisfaction. All these benefits, and more, can be seen after as little as 12 to 14 weeks of implementation.


Another area that could use some great help from AI is investigating terrorist financing. Spotting a red flag in these cases usually requires a lot of time going through and investigating a ton of documents and information about each individual case, to find one terrorist among the thousands of suspects. The machine’s AI algorithm will be able to organise the unstructured data into useful information, taking away “the noise” to reveal only the most relevant and crucial details. If we can teach the machine to spot a pattern in the suspects’ habits, it will help us to spot more terrorists before they end up in jail or dead.


Moving forward: AI Investment Considerations, and Focus on Microservices


The shift to focus more on Microservices is a good start to easier AI Integration. Fintech solutions that do that will require no new tech platforms for the banks to install and implement, and the system will also be no longer tied to a large data lake, making the operations leaner.


Another step forward would be creating an industry standard for AI like ISO 9001 and 20027. The benefits are substantial – it would create a standardised approach to AI model design, implementation, and adoption, which will not only make it easier for both fintech and banks, but also will lend credibility to the AI model industry. Efforts to educate about AI are also crucial. Pushing the Association of Certified Anti-Money Laundering Specialists (ACAMS) to include AI adoption in the syllabus would reduce regulators’ speculations and fear of it, making it a “norm” in AML Operations.


AI holds enormous potential in helping with the fight against financial crimes, however, it has a long way to go. Continued investment in research and development, along with collaboration between industry, regulators, and law enforcement, is essential to fully harness the power of AI and achieve our shared goal of a safer and more secure financial system. As we navigate this complex landscape, one thing is certain: AI is a critical tool in our fight against financial crimes, and we must work together to unlock its full potential.

228 views
bottom of page