AI in AML: practical uses and governance

A recent webinar with NICE Actimize examined how artificial intelligence is producing measurable results across the anti-money laundering lifecycle and what controls firms need when moving from pilots to live systems.

Panel participants included Nicola Eschenburg, EMEA AML Strategy Leader at NICE Actimize, with Teresa Connors acting as moderator. They outlined specific areas where firms are deploying AI: customer onboarding and KYC, transaction monitoring, investigations, case management and regulatory reporting.

Speakers described concrete use cases now in production. Automated adverse media checks can surface relevant negative news faster. Customer risk models that combine structured data and free-text sources refine risk scores. Screening workflows are being adjusted to reduce false positives. Alert triage tools prioritise higher-risk cases for investigators. Investigation automation can assemble timelines and produce document summaries. SAR preparation can be accelerated by generating draft reports and consolidating evidence.

The panel discussed vendor, in-house and hybrid approaches to building AI capabilities. The choice was linked to an institution’s data maturity, available skills and regulatory obligations. Clear success metrics for pilots were recommended, including reductions in false positives, shorter triage times and improvements in the quality of regulatory filings.

Governance issues featured throughout. Regulators expect documented model validation, ongoing performance monitoring and traceable decision records. Panelists highlighted the need for explainability for models that affect customer outcomes, testing to detect model drift and controls that ensure human oversight where required. Data quality and lineage were described as necessary for audits and supervisory review.

On agentic AI-systems that autonomously perform sequences of tasks-speakers advised a cautious, use-case focused approach. They outlined possible benefits, such as coordinating routine investigation steps and synthesising information across sources, and flagged risks including unintended actions, unclear accountability and challenges in explaining autonomous behaviour. Narrow pilots, rollback mechanisms and human-in-the-loop checkpoints were proposed as practical safeguards.

The session also covered operational integration. Panelists said linking alerts, case records and external screening results helps create a joined-up view of customer risk. They noted that interoperability and data exchange standards reduce fragmentation between front-office, compliance and investigations teams and support faster decisions.

Background context included pressure on financial crime programmes to improve efficiency and detection, and heightened regulatory attention on the use of advanced analytics. The webinar set out where AI is producing measurable outcomes today and what governance steps firms need to scale those outcomes while meeting supervisory expectations.