Financial crime compliance has become one of the most expensive operational commitments in financial services, and one of the least effective. North American financial institutions alone now spend more than $60 billion annually on compliance obligations. Despite that heavy investment, approximately $3.1 trillion in illicit funds circulated through the global financial system in a single year. At the same time, law enforcement agencies seized or froze less than 1% of those flows.
The explanation lives inside the alert queue. Only approximately 19% of AML alerts ever become investigative cases — and only a fraction of those result in a SAR filing. The cost isn’t just financial: it’s structural. Every hour an analyst spends clearing noise is an hour not spent on the cases that genuinely warrant investigation. But the problem isn’t the program design. It’s what the program can see and how insights are converted into intelligence.
This is not a willingness problem. The industry continues to invest heavily and compliance teams are not standing still. But the way most firms have built their technology stacks, with fraud and AML as separate point solutions, separate data, and separate workflows, means a valuable source of customer risk intelligence stays locked on one side of the house. FATF’s Annual 2024/2025 Report bears this out. Across nearly 200 jurisdictions assessed, technical compliance with the forty recommendations now exceeds 75%, yet the majority still show significant gaps in investigation, prosecution, and asset recovery.
Fraud is cited as a predicate offense in 89% of mutual evaluation reports, making it the second most common, yet coordinated responses linking financial intelligence to enforcement remain inconsistent. Being compliant and being effective are two different things, and regulators are increasingly measuring the latter.
SEON’s 2026 AI Reality Check report, which surveyed fraud and AML leaders, highlights exactly where this breaks down:
“Most organizations report that their fraud and AML workflows are only partially integrated, and 80% indicate that obtaining a unified view of data and insights remains a significant challenge.”
The signals that reveal how someone is actually behaving, not just who they claim to be, are already captured in fraud operations across most regulated firms. They are live, running and driving real-time decisions every day. The question is how to get them working for compliance teams, too.
Fraud and AML have traditionally been run as separate disciplines, with different teams, tools, and reporting lines. For many firms,s that still makes sense, and regulators often expect it. Segregation of duties matters. SAR confidentiality matters. But the people moving illicit funds know exactly where the gap between these two functions sits, and they are very good at exploiting it. The job now is to close that gap.
Detection and Analyst Productivity Can Be Comprehensive
It is worth being precise about where the challenge lives, because misdiagnosing it leads to the wrong solutions.
BSA/AML programs are built around a sound framework: internal controls, designated compliance leadership, ongoing training, independent testing, and customer due diligence. That architecture holds. The opportunity now lies in both enhancing detection by enriching data around the customer and improving how investigations are conducted to increase team productivity through the careful application of automation and AI.
The regulatory signal on this is now unambiguous. In its April 2026 proposed rulemaking, FinCEN stated directly that financial institutions have, for too long, been asked to measure success by the volume of paperwork rather than their ability to disrupt illicit finance threats. That is not yet a finalized rule, but it is a clear statement of regulatory intent. Institutions that wait for it to be codified before responding will already be behind.
This presents a clear opportunity to leverage fraud-grade signals to enhance AML transaction monitoring and customer due diligence. Providing a deeper context for actual behavior helps analysts quickly identify normal versus illicit activity. Device fingerprinting, digital footprint analysis, linked account mapping, and velocity anomalies across entity clusters were developed for fraud detection precisely because fraud operates in real time. This intelligence, built to identify bad actors in motion, can provide exactly the behavioral context compliance teams need to make faster, more accurate decisions.
By enhancing investigation and case review with supporting AI capabilities to summarise what has happened and identify patterns in the data to support judgment and decision-making, institutions open up huge opportunities to increase overall effectiveness. This is the direction we are building towards at SEON. A unified platform where fraud signals, AML intelligence, and a single view of customer risk sit in one place, with the governance layer to match. The goal is not to merge fraud and compliance into one function. It is to ensure the intelligence each function produces reaches the other at the moment it matters.
Fraud Is a Predicate Offense. So, Why Do We Treat Them Separately?
The relationship between fraud and money laundering is foundational, not incidental. Fraud generates proceeds. Those proceeds need to be moved, layered, and integrated into the legitimate financial system. The detection of that movement is, by definition, an AML problem. Yet the teams responsible for detecting the fraud and the teams responsible for detecting the subsequent laundering of its proceeds routinely operate without shared data, shared tooling, or even a common definition of what a suspicious customer profile looks like.
If you want a concrete example of why fraud and AML need to see each other’s signals, synthetic identity fraud is as clear as it gets. A synthetic identity is built, not stolen. Fragments of real data are combined with invented data and nurtured for months or years, slowly building a credit file and a transactional footprint until it appears to be a real customer. Then it is monetized, either through a credit bust out or by serving as a conduit for laundering. By 2030, synthetic identity fraud is projected to generate at least $23 billion in losses in the U.S. alone, driven by fabricated identities sophisticated enough to clear static verification.
The fraud team usually sees the early signals. Device reuse across accounts opened in tight windows, shared digital footprints between supposedly unrelated customers, and thin-file behavior that does not match the stated occupation or income are all classic fraud tells. The AML team sees the later signals. Structured inbound payments, rapid outbound transfers to newly introduced counterparties, and velocity patterns that do not align with the customer’s declared life are clear AML red flags.
Neither function handles the full lifecycle on its own. Synthetic identities are built to exploit exactly that gap. They are designed to look clean for each function in isolation. The only reliable way to catch them is to connect the signals across the customer’s entire journey.
This gap has persisted for structural reasons that go beyond org charts. In North America, BSA/AML programs are examined by prudential regulators — the OCC, Federal Reserve, FDIC, NCUA — and FinCEN. Fraud sits under operational risk or consumer protection, examined through an entirely different lens. The examination frameworks don’t converge, so the executive incentive structures don’t either. Compliance leaders are measured on exam outcomes and on the quality of SAR filings. Fraud leaders are measured on loss rates and detection speed. Nobody is currently being held accountable for the gap between them, and that accountability gap is where financial crime operates. With 67% of financial institutions reporting rising fraud rates, the cost of maintaining that division is no longer theoretical. It’s measurable and it’s growing.
One Customer Journey = One Risk Picture
Fraud and AML officers do not think in features. They think in terms of questions, and those questions change as the customer moves through their lifecycle.
It starts before a single document is checked. When someone first registers, the question is whether they should be there at all. Device intelligence, connection patterns, and digital footprint can surface risk at this point that traditional checks will never catch. Then comes KYC and onboarding: are they who they say they are, and can we legally do business with them? Sanctions, PEPs, and adverse media form the traditional gate, but behavioral signals can support the customer risk assessment and inform whether someone goes straight into enhanced monitoring.
Once the customer is active, the questions come faster, and the stakes rise. Every transaction raises two questions at once: is the counterparty clean, and is the behavior legitimate? Payment screening and fraud monitoring have to work in parallel, and the outcome has to be real time: hold or release.
Then there are the longer-horizon questions. Has this customer’s risk profile changed? Are higher-risk counterparties appearing that were not there before? Do the patterns over the last quarter look different from the first? Does the activity show behavior associated with money laundering, and do I need to file? Should this account be escalated for enhanced review?
No single system, fraud or AML, answers these alone. They need a unified view of risk that pulls on behavioral transactional, and identity intelligence across the full customer lifecycle. When that view exists, every question gets a better answer. When it does not, the gaps are exactly where illicit funds move.
The Gap is Fragmentation, The Solution Must Be Unified
The conversation about fraud and AML convergence has been running in the compliance community for well over a decade. The reason it has not translated into structural change at most institutions is not a lack of awareness. It is that the incentives, examination frameworks, and vendor ecosystems have all reinforced the status quo.
The direction of regulatory travel is now explicit on both sides of the Atlantic. FinCEN’s proposed rulemaking signals a clear shift in supervisory expectations toward evaluating whether programs are genuinely effective at detecting illicit finance. The EU’s Anti-Money Laundering Authority is centralizing supervision across member states. FATF’s revised 2025 methodology points in the same direction internationally. The question for institutions is not whether this shift is coming; it is whether they are prepared for it. It is whether they will be ready to demonstrate effectiveness when it arrives.
The organizations that respond most effectively will not be the ones that simply add headcount or layer in more rules. They will be the ones who surface intelligence to the right people, providing the context needed to make more informed, quicker decisions. By leveraging technology to automate workflows and prioritize the investigation experience, these organizations eliminate the manual effort of data collection, freeing their teams to focus entirely on critical thinking and complex risk analysis, delivering faster, better outcomes.
The evidence to detect financial crime more effectively already exists inside most regulated businesses. The gap is not intent nor effort. It is fragmentation, meaning fraud and compliance systems that were never designed to talk to each other, and the processes that grew up around them. When the tooling is disconnected, the workflows stay disconnected, and the critical intelligence never reaches the people making the decisions.
The gap between fraud and AML is where financial crime currently lives. The institutions that close it first will not just be better at compliance. They will be significantly harder to exploit.
