5 top tips on adopting agentic AI for AML due diligence

Agentic AI – or artificial intelligence (AI) systems that carry out tasks autonomously without direct human supervision – has become one of the hottest topics in financial crime compliance. With forward-thinking businesses looking to expand their automation capabilities, the application of AI agents to anti-money laundering (AML) workflows is top of mind for compliance leaders. 

The most recent session of our AML Unplugged event series saw an expert panel take on this topic, including Iain Armstrong (FCC Strategy Executive Director, ComplyAdvantage), Laura Lehane (Head of Financial Crime & MLRO, AJ Bell), and Will Lawrence (CEO & Co-Founder, Greenlite). 

Our panel discussed the revolutionary impact of agentic AI on customer due diligence (CDD) processes, exploring how firms can use it to strengthen risk detection and enhance efficiency at the same time. Here are five key points that emerged from the discussion: 

1. Agentic AI lets firms do more with less 

Compliance officers are used to constant pressure on their time, budgets, and teams, which invariably means most compliance functions cannot work at their full potential. During the know your customer (KYC) and CDD processes, alert reviews to identify and clear false positives are often a particular drain on compliance resources. This means genuine risks can get lost amid backlogs of unnecessary alerts, while a lack of capacity can delay ongoing checks on higher-risk customers. 

Agentic AI systems can automate various CDD tasks that suffer from slow, manual workflows. They conduct initial customer screening checks against essential risk data, including sanctions, politically exposed persons (PEPs), and adverse media, and generate alerts for any matches. 

AI agents can also review and triage alerts, removing false positives with a higher efficacy rate than manual reviews. This also lets higher-risk cases go straight to human analysts, who can preserve their time and energy for the cases that actually matter. Finally, AI agents monitor for risks continuously, updating customer profiles as soon as they detect changes in their information and allowing firms to move from periodic reviews towards perpetual KYC (pKYC). 

2. You can tailor your AI adoption to your risk appetite 

While agentic AI can work across the spectrum of CDD operations, you don’t need to take an ‘all or nothing’ approach to its adoption. As with any element of AML compliance, your existing tech stack, business-wide risk assessment, and risk appetite will factor into your AI implementation. In any case, you should: 

1. Establish a proof of concept.
2. Test the possibilities of agentic systems.
3. Build a range of use cases as you move up the compliance maturity curve. 

Depending on your risk appetite, this might involve only using agentic AI for initial screening checks and setting a lower threshold for human reviews, while including the option of using it for full alert remediation in your tech roadmap. 

Business brief: Agentic AI and AML compliance

Learn how to test and implement agentic AI across your organization to scale your business while streamlining compliance.

Download your copy

3. Agentic AI is about scaling your business, not replacing your team

Fundamentally, the role of agentic AI is to automate tasks, not eliminate human teams. Trying to find out customer information by carrying out manual searches is a waste of time for compliance analysts when this work could be done by AI agents, leaving them free to concentrate on more complex, higher-risk investigations. 

In the longer term, this provides a significant boost to firms trying to scale, allowing them to take on more customers without having to expand their compliance team just to keep up with data collection requirements when onboarding customers. With agentic AI in place, compliance teams have the potential to become more proactive as their companies grow, taking on a more strategic role across businesses as compliance becomes a driver of growth. Meanwhile, as AI expertise becomes essential, not optional, for compliance leaders, newer roles around agentic AI testing and management will become increasingly important. 

4. Always prioritize AI explainability 

While regulators recognize the potential of AI to enhance compliance, explainability has become a key consideration in how they assess firms’ use of AI tools, and failing to demonstrate transparency carries the threat of enforcement action. 

The success of your agentic AI systems depends partly on the data you use. Understanding the lineage of your data, having data-cleaning practices in place, and keeping records of investigations all ensure transparency while minimizing the possibility of AI mistakes or hallucinations. Model risk management practices and regular testing can also help to balance efficiency with explainability. 

One way to achieve this is to combine financial crime and AI expertise, ideally by partnering with a vendor whose solutions are specific to AML rather than an AI generalist. These vendors can provide a rigorously tested solution, comprehensive documentation, and tailored support for full auditability. 

AI regulation and the future of AML: Navigating compliance in a changing landscape

Watch our on-demand webinar for expert insights on your AI strategy as global regulations catch up with the AI revolution.

Watch on demand

5. People and policies remain paramount 

As transformative as agentic AI systems are, technology tools remain just one part of your compliance operation. Having an effective team of analysts and written compliance program is critical to efficient AML operations. Agentic AI is most effective when putting an already-robust compliance program into practice, while its resource-saving benefits only matter if your team can redeploy these resources elsewhere. 

These factors will also ensure smooth integration into your existing setup. Firms sometimes assume that adopting AI tools means creating a governance framework around their use from scratch, but this tends not to be the case in well-designed compliance programs. Ensuring you have sound, detailed structures around issues like testing, oversight, and data protection will allow you to add agentic AI-specific policies without overhauling existing rules.