When a payment looks wrong, a chain of automated checks decides in milliseconds whether to let it through, flag it or freeze it. Understanding how anti-money laundering works means following that chain from the moment an account opens to the day a suspicious report lands on a regulator desk. The stakes are large: Nasdaq Verafin counted more than $3.1 trillion in illicit funds moving through the global system in 2023, per Nasdaq.
How anti-money laundering works in the US financial market comes down to four building blocks: verifying who customers are, watching what they do, reporting what looks wrong, and screening everyone against sanctions lists. Each block runs on a mix of rules, data and increasingly machine learning.
How anti-money laundering works at account opening
The process starts with know your customer checks. Before a bank or fintech app lets someone transact, it verifies identity using documents, biometrics or database matches, and it assesses how risky that customer is likely to be. A salaried worker opening a checking account is low risk; a cash-heavy business in a high-risk industry gets more scrutiny.
This onboarding step feeds everything that follows. A customer risk score set at the door decides how closely the system watches them later. Get it wrong and either a criminal slips through or an honest customer is buried in checks. Digital-first firms have turned this into a near-instant flow, a convenience that customers now expect from apps like those that manage money and crypto together.
Risk scoring is not a one-time event. Good programs refresh a customer profile as behavior changes, raising the score if transactions suddenly look unusual. A dormant account that springs to life with large international transfers, for example, should trigger a fresh review rather than coast on its original low-risk label.
Transaction monitoring and suspicious activity reports
Once an account is live, transaction monitoring takes over. Software watches the flow of money for patterns that suggest laundering: sudden large deposits, rapid transfers in and out, structuring payments just under reporting thresholds, or money moving to high-risk regions. When a pattern trips a rule, the system raises an alert for a human analyst.
If the analyst agrees the activity is suspicious, the institution files a suspicious activity report with the authorities. In the United States these reports go to the Financial Crimes Enforcement Network. Banks also file currency transaction reports for cash movements above a set threshold. The table below shows the scale of crime these reports try to catch.
The volume is the hard part. A large bank can generate tens of thousands of alerts a month, and most turn out to be harmless. Tuning the rules so they catch real crime without flooding analysts is a constant balancing act, and it is where most of the human effort in AML actually goes.
| Metric | Figure | Source |
|---|---|---|
| Illicit funds through global system, 2023 | $3.1 trillion | Nasdaq Verafin |
| Drug trafficking proceeds, 2023 | Nearly $800 billion | Nasdaq Verafin |
| Human trafficking proceeds, 2023 | Nearly $350 billion | Nasdaq Verafin |
| Global fraud and bank fraud losses, 2023 | Nearly $485 billion | Nasdaq Verafin |
| AML market, 2025 to 2035 | $2.07B to $9.14B (16% CAGR) | Precedence Research |
Sources: Nasdaq Verafin 2024 Global Financial Crime Report; Precedence Research.
Sanctions screening and watchlists
Running alongside monitoring is sanctions screening. Every customer and payment is checked against lists of sanctioned people, companies and countries maintained by bodies such as the US Office of Foreign Assets Control. A single match can block a transaction instantly. Cross-border payments face the heaviest screening, which is why international transfers can take longer, as we note in our guide to cross-border payment solutions.
Screening is harder than it sounds. Names are spelled many ways, and criminals deliberately use variations. Good systems use fuzzy matching to catch near-misses without drowning analysts in false hits.
Lists also change constantly. When a government adds new sanctioned parties, institutions must rescreen their entire customer base against the update, sometimes within hours. That demand for speed is one reason sanctions screening has become heavily automated.
The technology that runs modern AML
The plumbing behind these checks has moved from simple rule engines to machine learning. Models trained on past cases can spot subtle patterns and reduce false alerts, freeing analysts to focus on real threats. This is the same wave reshaping the rest of finance, which we cover in our look at agentic AI tools in finance.
Crypto adds a new layer. Blockchain analytics firms trace funds across public ledgers, helping investigators follow stolen money, a process we break down in our article on recovering stolen crypto. The tools are powerful, but criminals adapt, so detection is a constant race.
Data quality decides how well any of this works. A model is only as good as the records feeding it, so banks spend heavily on cleaning and connecting customer data before a single alert is generated. Poor data produces both missed crimes and false alarms, which is why data engineering has become a core AML skill.
Where the workflow still breaks down
Even a well-built AML system has gaps. False positives waste analyst time, legacy data sits in silos, and information rarely flows between competing banks. That is why the industry is moving toward shared intelligence. Verafin reports 2,500 institutions holding $6 trillion in assets now use its consortium model to pool signals, a sign of where the workflow is heading.
How automation reshapes the workflow
The newest AML systems try to handle the routine work without a human. They cluster related alerts, pull the supporting data automatically, and draft the first version of a suspicious activity report for an analyst to approve. Done well, this cuts review time sharply and lets skilled staff focus on the cases that matter.
It also changes the skills banks hire for. Compliance teams now want data analysts and model validators alongside traditional investigators. The risk is over-reliance on a model no one fully understands, so regulators expect firms to explain why their systems flag what they flag.
For fintech firms without legacy systems, this is an advantage. They can build modern, model-driven AML from the start rather than bolting it onto decades-old software, which is part of why specialist compliance vendors have grown so fast. That demand shows up in the numbers: the global AML market is set to grow from $2.07 billion in 2025 to $9.14 billion by 2035, per Precedence Research.
How anti-money laundering works is ultimately a story of speed versus accuracy. The institutions that win will be the ones that move detection closer to real time while keeping honest customers out of the crossfire.
