AI in AML: moving from pilots to live operations

NICE Actimize hosted an online webinar featuring Nicola Eschenburg, EMEA AML Strategy Leader, to examine where artificial intelligence delivers measurable results across the anti-money laundering lifecycle and how firms can move from pilots into operational use.

The session covered practical applications of machine learning, large language models and agentic AI across customer onboarding, know-your-customer checks, transaction monitoring, investigations, case management and regulatory reporting.

Speakers identified specific use cases that firms are testing or deploying. Automated adverse media research and customer risk scoring were highlighted for reducing manual review time and surfacing more relevant leads for investigators. Screening optimisation and alert triage were shown to prioritise work queues and concentrate analyst effort on higher-risk activity. Investigation automation and natural language generation were discussed as methods to speed drafting of suspicious activity reports and to standardise narrative sections.

Panelists described technical and organisational barriers to scaling pilots. Integration across teams, legacy systems and existing workflows is required to connect separate risk signals and produce a single view of customer risk. Many pilots do not reach production because organisations underestimate the data engineering, change management and governance needed to deploy models in live systems.

The discussion addressed agentic AI and large language models. Agentic tools can automate routine tasks, while human review is needed for complex judgements and regulatory submissions. LLMs are used for document review, entity linking and draft narratives but are not presented as replacements for human decision-makers.

Speakers also covered options to build, buy or blend AI capabilities: some firms with strong data teams develop in-house models, others deploy vendor solutions, and many combine vendor tools with internal work on governance and integration.

Regulatory and governance topics included transparency, explainability, continuous model performance monitoring, validation, audit trails and documented human review. Presenters advised aligning AI use with existing risk frameworks and keeping records to meet auditor and regulator scrutiny.

The webinar presented operational details and governance considerations for embedding AI into AML controls. Nicola Eschenburg participated to discuss those implementation and oversight issues.