Financial crime evolves faster than any rulebook can keep up with.  AI helps us close that gap, but only when we use it with purpose.What truly matters is not only which algorithms we deploy, but how we use AI, which data we trust, and whether we tie every insight back to reality.

There’s one principleI rely on every investigation and every supervisory review:  effective AML starts with facts.

Payment data gives us that reality, offering the clearest view of what actually happened. When you look beyond a single transaction and analyze behavior over time, across accounts, customers, counterparties, and geographies, patterns emerge. Those patterns reveal real risk. 

A single transaction tells you nothing. A sequence is narrative. And narrative is where true risk hides. 

Seeing the Whole Story, Not Just the Checkboxes

For years, AML programs were built around fixed categories: high-risk countries, high-risk products, high-risk sectors.
Useful? Yes.
Sufficient? Absolutely not.

Crime rarely fits into neat boxes. Regulators know this. They don’t define a red flag as isolated datapoints; they describe scenarios, conditions and behaviors that interact. Because laundering is multi-layered. Financing is adaptive. Criminal networks are opportunistic.

This is where AI becomes essential, helping us move from rigid rules to real typologies. It reads across the entire data. It connects behaviors across time and entities. It surfaces patterns no human could detect alone.

AI doesn’t deliver the entire story, it shows us where the story begins. Human expertise determines how that story ends. 

AI Regulation: Move Carefully, But Move Forward 

AI regulation is expanding from the EU AI Act to GDPR to global frameworks on fairness and transparency. These guardrails matter. But compliance teams can’t treat all AI the same. 

AI used for recommending a movie is not AI for detecting money laundering, terrorism financing, or human trafficking. In AML, AI serves a public purpose: protecting people and safeguarding the integrity of the financial system.

So yes, proceed with caution. But caution can’t become paralysis. 

Every year we delay modernization, criminals widen the gap. Responsible innovation is not optional. It’s a duty. 

Why Human Expertise Still Decides Outcomes

There’s understandable excitement around self-learning AI. But in compliance, context remains everything.

Two alerts may look identical. One could point to a shell operation; the other to a  legitimate business with unusual seasonality. AI can detect data patterns, but it can’t yet interpret intent, geopolitical dynamics, or how criminal networks actually operate.

I recently reviewed a risk assessment built  only on two static lists. On paper, it looked complete. In reality, it  missed the broader economic and criminal exposure embedded in certain currency corridors. A trained expert spotted the gap in seconds.

AI accelerates detection. Experts validate meaning. One without the other is incomplete. 

That’s why AI won’t replace compliance teams, it elevates them.
AI brings speed and consistency.
Humans bring judgment and prioritization.

The Future: Shared Human and Machine Intelligence

Criminals adapt fast. Our defenses must adapt faster. The institutions that succeed in the next era of AML will be those that blend machine intelligence with human judgement, continuously, not occasionally. 

• AI adds: speed, scale, full visibility across data
  
• Humans add: context, interpretation, prioritization
  

Together, they turn fragmented data into a coherent risk picture, one rooted in factual behavior, not theoretical assumptions. When both human and machine intelligence work together, AML programs become smarter and organizations more resilient. 

The future of AML is factual, collaborative, and intelligent. Powered by technology. Guided by experts. Aligned with the realities of financial crime, not the comfort of simplified rules.