Reclaiming 18,000 Hours with AI Agents

The challenge: Scaling review capacity with alert volume

For a payments company serving millions of customers globally, reviewing potential matches placed a recurring demand on compliance resources. Each alert required approximately four minutes of manual review, making analyst workload directly dependent on alert volume.

The client needed a more efficient way to clear routine false positives while maintaining the company’s compliance standards. Adjudications had to align with the organization’s policies and procedures, and supporting audit trail documentation had to stand up to regulatory scrutiny. 

The new approach also needed to fit the team’s existing screening environment, LexisNexis Bridger Insight® XG. Replacing systems or retraining analysts on new workflows would introduce additional complexity.

The solution: AI alert adjudication within their existing screening workflow

Castellum.AI has a family of AI agents (Arbiter) that  adjudicate screening alerts directly inside of LexisNexis Bridger Insight® XG, the company’s existing screening platform. It allowed the team to add AI alert adjudication without overhauling their screening workflows or retraining analysts on a separate interface. 

For each alert, Arbiter compared customer and watchlist records against the company’s review policies, procedures and escalation thresholds to determine whether a false positive could be cleared or required further review. Arbiter reached decisions in approximately one second, compared with four minutes of manual review, reducing decision time per L1 alert by over 99%.

The company controlled the criteria for automated clearance and defined which cases require human judgment. Potential matches and cases with insufficient information were referred to analysts within their established workflow.

Arbiter enabled the compliance team to resolve alerts faster with clear controls and oversight:

• Institution-defined decision criteria. The company’s existing procedures governing sanctions, PEPs and adverse media alert reviews set the rules for clearance and escalation, aligning automated review with existing  compliance standards.

• Documented reasoning for every decision. Each decision included a documented audit trail within Bridger with plain-language rationale and the policy rule applied. 

• Context for further investigation. Escalated alerts included supporting reasoning so analysts could understand what required review and where information is missing, streamlining investigations.