Anti-money laundering desk
How is an anti-money laundering alert review judged?
Did you name the pattern and measure it against who the customer is? Did you pull the full evidence and make a call it supports? Did you escalate at the right level, write a report a regulator can follow, and meet the deadline? And did the customer never learn they were being looked at?
- Each line restates an item in the desk's scoring guide. Your graded report on the desk carries the full detail.
- 9 graded areas, each one below
Graded area 1
Whether you spotted the pattern in the payments
Names a specific laundering pattern or red flag, by the name FinCEN or FFIEC material uses, not a general suspicion.
Points to at least two signs of it in the case file.
Rules out the closest other pattern, and shows how it differs from innocent activity that looks the same.
Claiming a pattern the cited evidence does not support.
Graded area 2
The money trail, checked against who the customer is
States what the customer's file says to expect, and the field the activity contradicts.
Measures the change with an amount, a multiple or a share, never an adjective.
Allows for how long they have banked and for seasonal swings, and tells a bigger volume apart from a new pattern.
Calling activity "unusual" with no figure anywhere.
Graded area 3
Whether the evidence you pulled was complete
Opens the customer file before writing the decision, and looks beyond the dates the alert fired on.
Screens the other parties, not only the customer; checks news coverage, earlier reports and alerts, and related accounts.
Records a check that found nothing as done and empty.
Before calling a watch-list hit a true match, compares two or more details and records both values.
Skipping a step this kind of alert requires.
Calling a true match on a name alone.
Graded area 4
The call you made on each alert, and the risk in it
Quotes at least one specific payment: amount, date, other party or channel.
Names the alternative explanation considered, and what would have changed the decision.
Weighs suspicion against the legal standard rather than asserting a conclusion, and does not treat amounts just under a reporting limit as enough on their own.
When it changes the customer's risk rating, shows the working behind it.
A claim about the case that is not in the case file.
A changed risk rating with no reason recorded.
Graded area 5
When you passed an alert up, and when you held it
Tests the innocent explanation and says how; where it held, closes the case with reasons.
Escalates with a recommendation and a proposed action (monitor, restrict, exit or file), the evidence summarised for someone who has not seen the case, and the open questions listed.
Sends sanctions matters the same day.
Filing to be safe, case after case.
Holding back real suspicion out of fear of being wrong.
Graded area 6
The report you wrote, in words a regulator can follow
A suspicious activity report that says who, what, when, where, why and how, in order, readable without the case file open.
Keeps attempted and completed activity apart, and draws no legal conclusion.
Every decision has its own note, with dates, amounts and names that match the file exactly, and open questions stated.
Adjectives where facts should be.
The same template text pasted across unrelated cases.
Graded area 7
Whether you met the deadline
Decides inside the time limit counted from the detection date, with the arithmetic right.
Works urgent items ahead of routine ones, and flags a clock at risk rather than quietly missing it.
Gives high-risk items full treatment and low-risk ones a recorded, proportionate one, not silence.
Leaving low-risk items with nothing written at all.
Graded area 8
That nothing private left the desk: no tipping off, ever
Gives no hint of a report, an investigation or an alert, and deflects a probing question in normal service language.
Logs the probing question itself as a red flag in the case.
Any disclosure of a report or investigation. It caps this part for the whole shift.
Graded area 9
Checking the AI helper
Fixes the actual defect in a flawed AI draft, not just its wording, and names what it missed when overruling it.
Checks at least one claim in a draft before accepting it, and does not overrule a sound draft without cause.
Accepting a flawed draft unchanged.
Questions
Asked about this desk
Why is tipping off treated so harshly?
Because telling a customer they are under review is a serious breach in the real job. On this desk any disclosure caps that part of your grade for the whole shift.
See it on the desk itself: the Anti-money laundering desk, or work a real one free. Every guide is listed on the guides index, and if something here is wrong or unclear, tell us.
Free to start · no card
You do the work. We hand you the record.
Real tickets and cases, with AI at your side. Your first ticket is free and graded on the core checks. Every shift after it is reviewed part by part.