AML transaction monitoring is software that checks every transaction in a bank as it happens, and raises an alert when money moves in the way criminals usually move it. AML stands for anti-money laundering. Laundering means hiding where criminal money came from, so that it looks legal.
Every transaction in the bank reaches this software. It receives deposits and withdrawals from the core banking system and every UPI, NEFT and RTGS payment from the payment hub. It compares each one with what is normal for that customer. It uses two kinds of checks.
The first kind is rules. A rule describes a known pattern. Such a known pattern is called a typology. For example: "more than ₹10 lakh deposited in cash within a week, by a customer who usually deposits ₹50,000 a month." Some banks share typologies with each other.
The second kind is models. A model learns from past cases what suspicious behaviour looks like, and gives each transaction a score. Models can find new patterns that no rule describes.
Consider a small trading firm's account in Surat. In one afternoon, it receives forty UPI payments from forty different people, each just under ₹50,000. Within an hour, almost all of it is sent to three other accounts. No single payment breaks a rule. But together, the pattern matches a known typology. The software raises one alert, with all the payments attached.
The alert goes to an investigator in AML case management. Many alerts turn out to be harmless, so the software tries to keep false alerts low. Special checks look for mule accounts, which pass stolen money along. If the investigator confirms the suspicion, the case goes to regulatory reporting, and the bank reports it to the Financial Intelligence Unit, India (FIU-IND), on FIU-IND's online portal, FINGate, within 7 working days of confirming the suspicion.
The models themselves must be checked regularly by model risk management. The software depends on good KYC data, because it needs to know who each customer is and what is normal for them.
How is it different from payment fraud detection? Fraud detection protects the customer from a thief, and it must decide in milliseconds whether to block a payment. AML monitoring looks for the customer who is moving criminal money, often across many payments and days.