AI in Financial Services: An FNA Perspective

Insights

16 Dec

By Dr. Carlos Leon (Director of Central Banks and Financial Market Infrastructures)

At DC Fintech Week 2025, FNA’s Carlos Leon joined a panel discussion exploring how AI is reshaping financial services. The conversation made one thing clear: AI’s impact is real, but only when we understand the complexity of the systems we are trying to improve. Carlos sums up the discussion from an FNA perspective in this article.

As financial institutions, infrastructures, and markets interact in ways that simple models cannot capture, the discussion at DC Fintech Week offered an opportunity to explain why this matters and how AI can help us move from reactive oversight to proactive system-wide resilience.

Financial Systems are Complex Adaptive Systems – AI Must Reflect this Reality

Financial systems behave as complex adaptive systems. They are shaped by intricate interactions, nonlinear effects, and path dependencies that make traditional analytics tools insufficient.

Because of this, modelling these systems requires a combination of techniques. At FNA, the Digital Twin solution integrates simulation, machine learning, graph theory, network analytics, and optimisation. AI plays a vital role here, not by replacing other methodologies but by complementing them, especially in clearing and characterising institutional payment behavior.

This integration is essential because payment systems provide real-time footprints of economic activity. When we compare observed payment behaviour to what our models expect, authorities can detect anomalies, early signs of stress, or liquidity issues long before they appear in traditional supervisory data.

This shift from delayed reporting to continuous, real-time oversight is one of AI’s most transformative contributions. Yet the value of real-time analytics becomes even more evident when we consider a second major issue we discussed: the rise of fraud in increasingly interconnected payment networks.

Institutional-level Fraud Controls are No Longer Enough. We need System-Level Defences:

The same interoperability that makes modern payments efficient also makes them vulnerable. Fraudsters exploit the ability to move funds instantly across banks, jurisdictions, and currencies. This reality means that institution-level compliance alone cannot protect consumers. Once the funds leave a bank’s ledger, opportunities to intervene diminish rapidly.

To counter this, we need to elevate fraud management to the system level. A National Anti-Scam Utility, where banks, payment operators, and authorities share data and intelligence, offers the most effective approach. By analysing national-level data with tools such as GraphAI, authorities can identify networks of mule accounts and patterns of concealment that would be invisible to any single institution.

The connection to Digital Twins is direct: both rely on the idea that network behavior reveals the system-level risks. Just as real-time payment data allows supervisors to spot liquidity stress early, shared fraud data enables authorities to detect and disrupt fraudulent flows before they spread.

The ultimate form of consumer protection, however, is rapid fund recovery. With a national utility capable of tracing across all systems in real time, authorities can freeze and recover funds more efficiently, even as criminals exploit AI and social engineering to scale their operations locally and across jurisdictions.

How Can Supervisors Use GenAI to Make Oversight More Natural, Efficient, and Scalable

When asked about the role of GenAI in Suptech, the most important contribution is enabling more intuitive, accessible interaction with complex data. Supervisors must navigate large volumes of raw financial information and dense regulatory frameworks. GenAI helps by allowing natural language interaction, automated summarisation, and efficient analysis.

Importantly, GenAI doesn’t work in isolation; its impact is greatest when paired with the kinds of systemic models described earlier. A supervisor who can ask a GenAI interface to “explain liquidity risk changes observed today” and receive an answer grounded in a Digital Twin’s analysis gains a level of insight that was previously unavailable.

This brings the narrative full circle. Whether we are modeling liquidity stress, tracing fraudulent funds, or navigating regulatory complexity, the common thread is that AI enhances our ability to understand and manage complex systems. It amplifies expertise, accelerates analysis, and provides the situational awareness needed to intervene early and effectively.

The panel’s title, “Fact or Fantasy?”, highlighted a common misconception about AI in finance. Some expect too much from AI, others underestimate its value. What emerged from our discussion is that AI is neither a magical solution nor a distant possibility. It is a practical, powerful set of tools that, when used responsibly and embedded in robust analytical frameworks, can materially strengthen the financial system.

The key is to deploy AI in ways that: