Curriculum / Module 04
core2.5 hours5 lessonsTimed assessment

AI in the AML Fight

Monitoring, network analytics and agentic laundering

How machine learning actually does transaction monitoring, where it fails, and how criminals now use automation against it. The module covers supervised and unsupervised detection, graph and network analytics for structuring, the false-positive problem, and the emerging threat of agent-driven laundering — all grounded in the South African FICA regime and the Risk Management and Compliance Programme obligation. You leave able to read, challenge and supervise a model-driven alert, and to articulate the human role that FATF expects to remain.

Create an accountBack to curriculum

By the end of this module, you can
  • Explain how machine-learning and network-analytics models detect suspicious activity and why they generate false positives.
  • Read and challenge a model-driven alert, and reason about structuring patterns a rules engine misses.
  • Locate these tools inside FICA obligations and the Risk Management and Compliance Programme, and articulate the human role that FATF expects to remain.
Skills taught
  • Alert triage under model-driven monitoring
  • Graph and network reasoning about laundering typologies
  • Supervising and challenging an AML model
  • FICA and RMCP literacy

Lessons in this module

L1
From rules to models: how machine learning monitors transactions
Explain the difference between rules-based, supervised and unsupervised transaction monitoring, and identify what each can and cannot detect.
25 min
L2
The false-positive problem: triage without rubber-stamping
Explain why monitoring systems generate overwhelming false-positive volumes, and apply a triage discipline that neither rubber-stamps the model nor drowns in its alerts.
25 min
L3
Graph and network analytics: seeing the scheme, not the transaction
Reason about laundering typologies — structuring, mule networks, layering — as network patterns, and explain what graph analytics detects that transaction-level monitoring cannot.
25 min
L4
Agentic laundering: when the adversary automates
Describe how agentic automation changes laundering — speed, coordination, adaptation — and identify the detection signals that survive it.
20 min
L5
FICA, the RMCP, and the human the law expects to remain
Locate model-driven monitoring inside FICA's obligations — due diligence, monitoring, reporting, the RMCP — and articulate where automation may assist and where human judgment is legally load-bearing.
25 min
Case 04.1
Case file: The alert the model cleared
An applied fact pattern worked against a model resolution, followed by the timed assessment (35 minutes, pass mark 70 percent).
25 min

How it lands across the four desks

Fraud & AML

Your core module. You learn what the monitoring model is actually doing when it scores a transaction, how to triage its alerts without either rubber-stamping or drowning, and how to reason about the network patterns — mules, structuring, layering — that no single-transaction view reveals.

Compliance & Risk

You own the RMCP and the model-governance layer. This module gives you the vocabulary to challenge a monitoring vendor, the documentation posture FICA requires, and the reasoning for why a human decision-maker must remain in the reporting chain.

Wealth & Advisory

Your onboarding and client instructions feed the monitoring chain. You learn what downstream models do with the data you capture, why accurate source-of-funds and expected-activity records matter, and how laundering typologies touch advisory products.

Credit & Underwriting

Credit products are laundering instruments too — loan-back schemes, early settlements in cash, credit-balance abuse. You learn the typologies that arrive through lending, and how your application data feeds network analytics.

Key literature · 6 sources

Every module rests on a verified scholarly and institutional evidence base. The full core and further reading lists open with the module.

  • Chen, Z. et al. (2018) 'Machine learning techniques for anti-money laundering (AML) solutions in suspicious transaction detection: a review.' Knowledge and Information Systems 57(2) — the supervised/unsupervised monitoring toolkit surveyed.
  • Weber, M. et al. (2019) 'Anti-Money Laundering in Bitcoin: Experimenting with Graph Convolutional Networks for Financial Forensics.' KDD '19 workshop (Elliptic dataset) — the demonstrated principle of neighbourhood-aware detection.
  • Altman, E. et al. (2023) 'Realistic Synthetic Financial Transactions for Anti-Money Laundering Models.' NeurIPS 2023 — why labelled AML data is scarce and synthetic data matters for model development.
  • FATF (2025) 'Horizon Scan: AI and Deepfakes — Impacts on ML/TF/PF' — the automated-laundering and detection-lag findings framing Lesson 4.
  • Financial Intelligence Centre Act (FICA) and published FIC guidance on the Risk Management and Compliance Programme — the statutory frame of Lesson 5 (consult the current consolidated Act and live FIC guidance).
  • SARB Prudential Authority & FSCA (2025) 'Artificial Intelligence in the South African Financial Sector' — local adoption of AI in fraud detection and the skills-gap finding.
Create an account