Where AI is delivering measurable results in AML

A recent online webinar hosted with NICE Actimize examined practical applications of artificial intelligence across the anti-money laundering lifecycle and where institutions are seeing measurable results.

The session brought together Nicola Eschenburg, EMEA AML Strategy Leader at NICE Actimize; Ashley Bostel, Head of Economic Crime Analytics Strategy at NatWest Group; and moderator Teresa Connors. Panelists discussed how financial firms are moving from pilots to operational deployments and how teams can connect risk signals across systems and workflows.

Speakers identified stages where AI is already in use: customer onboarding and KYC checks, transaction monitoring, alert triage, investigations and case management, and regulatory reporting. Use cases mentioned included adverse media research, automated customer risk scoring, optimisation of screening to reduce false positives, automated triage of alerts, investigator workflow automation and drafting of suspicious activity reports (SARs).

The panel described how different technologies are applied. Machine learning models are used to detect anomalous transaction patterns. Large language models (LLMs) are deployed to summarise open-source media and to structure unstructured documents. Agentic AI tools are being trialled to automate routine investigative tasks and to surface the most relevant evidence for analysts.

Panelists cited measurable outcomes in specific deployments. Examples included shorter investigation times, fewer alerts requiring manual review and improved accuracy of risk assessments compared with rule-only approaches.

Speakers addressed integration challenges. They outlined the need to feed model outputs into case management systems, define clear handoffs between monitoring and investigation teams, and ensure workflow automation fits existing compliance processes. They noted that operational deployment often exposes gaps in data pipelines, model governance and cross-team coordination.

Regulatory and governance issues featured throughout. The discussion covered expectations for transparency, explainability and auditability of AI-driven decisions. Panelists advised maintaining model validation records, tracking data lineage and preserving human oversight of automated outcomes. They said institutions need documented controls and performance metrics to show tools meet regulatory requirements and internal risk thresholds.

When considering build-versus-buy choices, participants recommended assessing internal skills, data maturity and governance capacity. Several panelists described a blended approach in which commercial solutions are combined with in-house customisation and integration into existing case management platforms.

On agentic AI, the panel described a role limited to augmenting investigators rather than replacing them. Examples included automated evidence collection, generation of draft reports and prioritising high-risk alerts. Panelists cautioned that greater autonomy requires robust controls, continuous monitoring and clear rules for when human review is mandatory.

The webinar concluded with a call from participants to focus deployments on clearly defined use cases that reduce workload or improve detection, to align projects with governance and regulatory expectations, and to prioritise integrating risk signals across systems to support faster, more informed decisions.