How AI Makes Legacy AML Processes More Efficient

Efficient and accurate data analysis is vital for effective AML/CFT programs – yet AML teams using legacy transaction monitoring programs frequently deal with backlogged systems. Their analysts often experience burnout from processing high volumes of alerts with too many false positives. Without a way to triage incoming alerts, highly-qualified investigators can spend most of their working day on rote tasks such as clearing overloaded systems and low-risk alerts. 

This doesn’t just create frustrations – it wastes company time, financial, and energy resources, overloads personnel, and makes it more likely that teams will miss illegal activity. It can also lead to unwanted organizational costs and losses. For example, team burnout means high turnover rates and costs to recruit and train replacements. Poor screening can result in losses to fraud and resulting disputes. 

And most importantly, if a company is deemed to have insufficient risk management processes, it can face regulatory fines and legal action. In a particularly high-profile instance, one global investment firm was fined over $1 billion in 2022 for – alongside long-term fraud – “failure to implement key risk controls.”

How AI Solves the Cost vs. Risk Dilemma

Despite these rising pressures and risks, many financial institutions worry that a system overhaul would cost even more. But it’s actually possible to keep a firm’s core system in place and while overlaying artificial intelligence algorithms to enhance its abilities. Indeed, competitive firms have spotlighted their reliance on artificial intelligence and machine learning (ML) as key to their success. 

“Effectiveness and efficacy are key for scaling. We can’t grow our team every time we grow our customer base,” explains Valentina Butera, Head of AML and AFC Operations at Holvi, a leading digital bank. And in a recent interview, PwC Luxembourg’s Andreas Braun highlighted the tremendous data processing and analysis possible through AI, which helps solve traditional risk management efficiency and cost dilemmas. 

A 2022 report by Allied Market Research predicted that the market for fintech AI would reach over $61 billion by 2030. Once relegated to speculation, AI and ML are now practical realities – and judging by worldwide regulatory responses, their use is becoming ubiquitous. Key examples include: 

• The European Union’s proposed Artificial Intelligence Act
• US House Committee on Financial Services’ Task Force on Artificial Intelligence
• The United Kingdom Financial Conduct Authority’s AI Public-Private Forum and Artificial Intelligence Discussion Paper
• The Monetary Authority of Singapore’s Project Veritas 

In our annual State of Financial Crime report, 99 percent of surveyed firms expected AI to positively impact financial crime risk detection. Consider the three most-selected use cases for AI in transaction monitoring: 

• Alert Prioritization – 31 percent of respondents expected AI to help rank transaction alerts by risk. This enables transaction monitoring teams to catch more risky activity and do it faster. 
• Flexible Tuning – 26 percent thought they’d use AI to improve their alert system – helping to adjust thresholds and fine-tune alerts responsively. 
• Relationship Identification – 24 percent anticipated artificial intelligence would uncover new relationships between monitored entities and individuals. 

Using AI to Enhance Transaction Monitoring

How might an AI overlay work in practice?

Consider a scenario. A senior analyst, Allison, has been dealing with bloated and imprecise alert queues due to rigid rules and no triage by priority. Every day, she spends hours painstakingly working through individual alerts without an efficient way to tell which are critical and most worth her time investigating. When she does come across a high-risk alert, she has less time to research it because of wasted time clearing false positives. In fact, if the system is backlogged, alerts tied to actual financial crime may sit in the queue for days or weeks before they’re found. The team has lost several members recently, but Allison doesn’t have time to keep up with her queues and train new teammates effectively.

Then imagine her company adding a layer of artificial intelligence over its existing system to handle alerts more intelligently. The new AI overlay combines multiple powerful risk management techniques, allowing it to:

• Automatically triage alerts – The AI knows how to triage incoming alerts by risk level, assigning a high level of risk to those showing the most suspicious activity. It will also continuously improve based on analysts’ feedback. Allison immediately begins looking at the high-risk alerts queue when she comes into work. Meanwhile, the lowest-risk alerts are either resolved in bulk or used to train newer analysts. And when mentoring advancing team members, Allison can use the high-risk queue to illustrate how to handle risky alerts.
• Enable more effective tuning – AI also allows the team to improve and adjust underlying rules’ parameters and thresholds. This enables more risk-responsive alerting, which helps enhance detections and reduce false positives.
• Uncover more bad actors – Weak evidence related to one person alone may not lead to an escalation. But with the new AI overlay, Allison’s team can leverage weak correlates in their data pools to identify and disrupt clusters of criminal activity.
• Identify true actors working behind the scenes using identity clustering to seek out hidden relationships. The team can now see connections and red flags that were previously invisible to them.
• Get greater insights and explainability around the reasons for an alert being generated. Allison is more confident that she and her team can back up their decisions should they be audited or receive an inquiry from their senior leadership.

With minimal upfront cost, AI-enhanced transaction monitoring maximizes Allison’s value to the company, empowering efficient and risk-based investigation while allowing her to train team members effectively. Meanwhile, layered machine learning models improve the effectiveness of her firm’s AML/CFT risk detection process.

Key Takeaways

For years, compliance teams have known that legacy AML software and processes have not met the financial crime challenges faced by their organizations. Rigid rules and tick boxes may capture some egregious behavior, but they miss much of the complexity involved in illicit activity. They also cannot see the bigger picture and broader connections between entities and people necessary to help law enforcement eliminate criminal behavior root and branch. The tools and technologies now exist for banks to meet this moment. 

Iain Armstrong, Regulatory Affairs Practice Lead at ComplyAdvantage, weighs in. “Many firms are already seeing success with AI, so it’s important to be agile, and avoid falling behind competitors who may soon be able to work in a much more sophisticated way without comparable increases in costs.” Indeed, ‘AI’ is no longer simply a buzzword – it’s an umbrella for many actionable programs that firms can implement today. Regulators worldwide recognize this and likely will soon ensure AML regulations reflect the innovations available to firms in their jurisdictions.

An artificial intelligence overlay can provide a simple, cost-effective option for companies that need AI’s benefits but aren’t in a position to make a major overhaul. Using an overlay also involves fewer unknowns, since the algorithms don’t replace existing processes but enhance them. With minimal disruption, firms can improve AML/CFT compliance efficiency with AI-enhanced alert prioritization, risk detection, and escalation – reducing risk and related costs while supporting employee retention rates and remaining competitive in an ever-changing compliance landscape.