AI in AML: why the hardest investigations often pay off first

TL;DR: AI in AML delivers the most value when teams start with high-handle-time work such as enhanced due diligence, escalated investigations, or large fraud rings, according to Unit21, because sanctions screening is usually too quick to produce meaningful return. The harder test is less about model capability than about whether the underlying process is written clearly enough to automate safely.

Editorial analysis by NHI Mgmt Group, based on content published by Unit21: “Where to start with AI in AML, and what actually decides whether it works”.

Key questions

Q: Where should AML teams start with AI if they want meaningful return?

A: Start with the workflow that consumes the most analyst time, not the easiest one.

Q: Why is sanctions screening often a weak first use case for AI in AML?

A: Sanctions screening is usually too fast to produce large savings, even if AI shortens each review.

Q: What breaks when AML processes are not written down precisely enough for automation?

A: AI surfaces the inconsistency immediately.

Practitioner guidance

• Prioritise high-handle-time workflows first Start with enhanced due diligence, escalated investigations, or large fraud-ring cases where AI can remove hours rather than minutes from review time.
• Test AI on your hardest real case Ask vendors to run the most complex workflow on your own data, not a clean demonstration case, so you can see whether the system holds up under real review pressure.
• Document analyst steps at execution level Write the workflow closely enough that different analysts would follow the same sequence, because any variance will surface immediately when AI tries to automate it.

Bottom line: The central issue is not whether AI can help AML teams, but which workflow creates enough leverage to justify the change.

Explore further

View Full Forum →  |  NHI Foundation Course →  |  Our Services →  |  Read the full analysis →