The Customer Context Checklist to Run Before You Close an Alert
Most AML reviews fail at the same point. The analyst has the flagged payment, the rule name and an account number, and starts pulling threads. Some of those threads lead somewhere. Some run out because the working day ran out. A short checklist at the front of the case removes the guesswork about what "enough context" means.
This is a practical guide for UK and EU teams. It assumes you already have monitoring, screening and a case tool. It does not assume they talk to each other. If they do not, that is the point of the missing layer in the AML technology stack. The checklist follows the same logic as our customer-first investigation model: assemble the customer, then judge the payment.
1. Identity and onboarding
Who is the customer, and what did they tell you at onboarding? Read the original note, not the current dropdown. Occupation, source of funds, expected geography and the stated purpose of the account are the baseline against which everything else is measured.
2. Risk rating trajectory
Not the current rating. The path. A customer who moved from low to medium after a prior review is a different proposition from one who has never been touched. A static rating hides the reason the alert exists.
3. Prior cases and dispositions
Every closed alert, with the note that closed it. "Consistent with documented income" fourteen months ago is context, not noise. It tells you what changed since. Firms that fail to carry that reasoning forward end up re-deciding the same question each time.
4. Connected accounts
List related accounts and check whether their inflow patterns correlate. A connected account is not automatically suspicious, but it changes what "normal" means for the customer. If this step takes four system switches, record that too.
5. Counterparty indicators
Who is on the other side, and do they carry their own risk signals? One unusual counterparty is a fact. Five unusual counterparties sharing a structure is a typology.
6. Typology fit
Which known pattern does the behaviour resemble, and which does it not? Being able to say why a typology does not fit is as valuable as saying that it does.
7. Decision record
Write the reason for the outcome before you write the summary. The record is what survives a QA review, a handover, or a supervisor's question two years later.
Run this against five closed cases this week and time each step. The hidden cost of manual AML investigations shows up in exactly those minutes. If most of your time is steps one to five, you have a retrieval problem, not an analyst problem. The fuller case-file structure is in the evidence hierarchy guide and the operating model is set out in this longer article; I also posted a condensed version on Mastodon.
More checklists and guides sit in our knowledge series and the main blog. When the checklist keeps taking longer than the decision itself, it is worth looking properly at how an investigation platform assembles this material before the case opens rather than asking analysts to move faster.
General commentary, not legal or regulatory advice.
Comments
Post a Comment