5 Steps Banks Can Take To Implement Next-Level AI-Driven AML

Financial crime is evolving at a pace never seen before. This is being fuelled by the rapid adoption of AI by both banks and bad actors. Today’s criminals are leveraging AI to create synthetic identities at scale, automate money mule recruitment, and generate flawless deepfakes. These are all designed to evade traditional AML controls. As these threats become more sophisticated, it is not enough for banks to just keep up, they need to get ahead. To do this, financial institutions need trusted, explainable AI that is grounded in context-rich data.

Without context, even the most advanced AI can only see isolated data points, missing the real-world picture of connected behaviors and relationships. And, with banks facing mounting pressure from regulators, boards, and customers, they need to act decisively.

Here are five steps that banks should take.

1. Build a strong data foundation

   A strong data foundation is the bedrock of effective AI in AML. But rather than just collecting more data, you need to be creating context. By unifying fragmented data sources and resolving entities and relationships, banks can build a dynamic contextual fabric. This context transforms isolated data points into a connected, real-world view of customers and counterparties, enabling AI to identify complex patterns and risks that would otherwise go undetected.
   

2. Leverage context in your data

   Context is what separates truly intelligent AML from legacy approaches. AI models trained on siloed data often miss the subtle, network-driven behaviors of sophisticated criminals. By embedding context into every stage of detection and investigation, financial institutions can move from static, rules-based monitoring to adaptive, intelligence-led strategies.
   
   A contextual approach means connecting fragmented data sources to build a holistic, real-world view of customers, counterparties, and their relationships. This network perspective enables both human analysts and AI systems to identify complex patterns that would otherwise go undetected. Context also enhances the performance of machine learning and graph analytics, allowing models to distinguish between genuine risk and noise more effectively.
   
   Additionally, teams will be able to carry out more targeted investigations and their reporting can become far more accurate. The result is not only improved detection rates and reduced false positives, but also greater efficiency and confidence in decision-making.
   

3. Use various domain-specific models trained for specific outcomes

   Your AI program can’t be a one-size-fits-all solution. Relying on one AI model, technique, or approach can lead to limitations in perspective, adaptability, and overall performance. For example, the methods and approaches used to build effective models in Retail AML, where there are many recorded outcomes, false positives, and most importantly, disclosures to the regulator (e.g., training data) might differ from approaches used in trade finance or correspondent banking processes. In these cases, training data may be scarce. The scarcity requires a completely different approach and model that may rely on enhanced domain knowledge and subject matter expertise.
   

4. Prioritize explainability and transparency

   Ensuring transparency and explainability of predictions and decisions is of utmost importance in AML and financial crime programs. As alerts are investigated by human analysts, it's critical to show them where the risk is and why the analytics triggered the alert.

   A “white box” approach ensures you can meet regulatory and compliance requirements while customizing and tuning your models where needed. Both technical and non-technical teams should be empowered with the ability to easily explain the obtained results and validate the effectiveness of their models. Explainable outcomes will promote adoption and trust in your AI while also mitigating security, privacy, and ethics concerns.

5. Ensure responsible governance
   Effective governance frameworks ensure that AI-driven decisions are transparent, and aligned with regulatory expectations. This includes establishing clear data quality standards, maintaining comprehensive records of model behavior, and enabling human oversight at key decision points. By prioritizing governance, organisations can foster trust, support compliance, and create a resilient foundation for ongoing innovation in financial crime prevention.

“By replicating your best investigators’ approach across networked data, the AI will alert far more accurately and then provide the investigator team with a picture and wider, more detailed explanation of the fraud."

Jennifer Calvery
Group Head of Financial Crime Risk and Compliance, HSBC

Read the full customer story here

AI-driven AML in practice at HSBC

HSBC is one example of a global bank leading the way in using next-generation technologies such as AI to improve financial crime compliance. Back in 2017, HSBC deployed Quantexa’s Decision Intelligence Platform to improve its AML programs across complex areas of the bank.

Quantexa’s Decision Intelligence Platform has helped HSBC to:

• Build a strong data foundation

“Knowing who is transacting with whom is critical for HSBC investigators who want to view entire counterparty networks as they search for illicit activities,” explains Jennifer Calvery, HSBC’s Group Head of Financial Crime Risk and Compliance. For HSBC, that means putting together a picture of how individual customers behave and operate using Quantexa’s Entity Resolution and Graph Analytics technology.

“Decision-makers didn’t always have a direct line of sight into their risk exposure to see all of the connections between their counterparties…and their counterparty’s customers, vendors, and suppliers,” explains Kai Yang, Global Data CIO at HSBC. Now Quantexa is HSBC’s enterprise-wide vendor of choice for Entity Resolution.

• Gain a contextual view of global customer and counterparty risk.

In 2017, Quantexa created a Global Social Network Analytics (GSNA) platform for HSBC, so the bank could analyze data with “… a global analytics solution that identifies potential financial crime by contextually analyzing customer, transactional, and publicly available data to understand a customer’s global network…”  This view allows the bank’s human investigators to better conduct informed analysis of potential criminal activity. The GSNA platform replaced several of the bank’s legacy data analysis tools and won HSBC the Celent Model Bank Award for Risk Management.

• Simplify complex banking AML practices such as Trade Finance.

HSBC’s Global Trade and Receivables Finance (GTRF) business processes over 5.8 million trade transactions a year for signs of financial crime — vastly more transactions and connections than on banking’s retail side. In an industry first, in 2019, HSBC deployed Quantexa’s AI-enabled customer surveillance system that uses big data, advanced analytics, and automated Contextual Monitoring to detect and disrupt financial crime in international trade. 

• To automate screening processes for complex investigations.

When cases are reviewed by bank investigators, Quantexa supports investigations by highlighting resolved entities and networks – reducing false positives, and shortening the time needed to close cases.

“The vast majority of what’s going on is completely legal and legitimate, so we’re looking for a very small sub-segment in this big group, and so we need that contextual picture,” says Calvery. “All of these factors are causing banks to take a second look at how they use AI to automate the screening process and flag suspicious activities that may be lurking in the dark corners of this highly complex transactional maze.”

The results of HSBC’s reliance on Quantexa contextual AML solutions are clear: The bank has continued to reduce false positives year-over-year and is finding new risks not detected previously.

“If I wanted to boil it down into really simple terms, we find more of those financial criminals who are exploiting our products and services and we find them faster, through tools like Quantexa.”

Jennifer Calvery
Group Head of Financial Crime Risk and Compliance at HSBC

See how HSBC is using Quantexa’s Decision Intelligence Platform to fight financial crime here.

HSBC also uses Google AML to provide the transaction alerting solution in its retail and commercial operations, helping the bank help clients in the financial sector meet compliance regulations for screening and reporting potentially suspicious money laundering activity. This is a great example of how a bank can use multiple complementary solutions for different areas of the bank.

Resilience in the age of AI-driven financial crime

As criminals harness AI to outpace traditional defences, banks must respond with trusted, explainable AI grounded in a connected data foundation. Quantexa’s Decision Intelligence Platform empowers banks to detect, investigate, and prevent financial crime with confidence.

The collaborative capabilities of Quantexa are not locked to a specific provider of AI anti-money laundering solutions. Quantexa can work directly with banks that are building their own AML models. The future promises multiple pathways for banks looking to work with AI, which include, working with best-in-breed technology to building models in-house and integrating with technology from companies such as Quantexa.

Are you ready to move from fragmented data and reactive compliance to proactive, intelligence-led AML? Discover how Quantexa can help you win the war on financial crime.