Agentic AI for AML Investigations: Matching Analyst Accuracy and Identifying More True Positives

Agentic AI for AML Investigations: Matching Analyst Accuracy and Identifying More True Positives

Agentic AI for Investigations: Challenge in Summary

  • Alviere had already benefited from significant efficiency gains with Hawk’s AML Transaction Monitoring solution 
  • They now wanted to determine if Hawk’s AML Investigative Agent could match analyst-level accuracy and increase those efficiency gains
  • Alviere tested the Hawk Agent against historical alerts, comparing its output to decisions that analysts had already reached on the same cases

How Did the AML Investigative Agent Perform: Results in Summary

  • Hawk’s AML Investigative Agent correctly escalated every confirmed true positive, ensuring genuine risks were not overlooked by the Agent
  • The Investigative Agent also closed 98% of the alerts that human analysts had identified as false positives
  • More significantly, it flagged a number of additional potentially suspicious alerts

Alviere: Using Agentic AI for AML Alert Efficiency

Alviere operates across a broad range of markets and customer segments. As a result, Alviere is responsible for managing regulatory risk across a diverse range of customer types, business lines, and transaction behaviors.

The alerts that reach its AML team can vary significantly depending on the business context, customer risk profile, and financial activity involved. What looks like a financial crime threat within one business line might be normal customer behavior within another.

To adapt detection, the team relied on rule threshold tuning. However, optimizing rules across diverse use cases created an ongoing resourcing challenge. To maximize analyst coverage, Alviere wanted its team to focus on high-value manual reviews, not fine-tuning detection.

Alviere needed a more efficient way to triage and investigate alerts. The Hawk AML Investigative Agent helps to achieve these goals, enabling Alviere to automate data collection and documentation, investigate more cases, reduce investigation times, and make reviews more consistent.

What Did the Hawk Agentic Solution for AML Investigations Deliver?

Deployment Across Business Lines

Alviere deployed Hawk’s AML Investigative Agent across four business lines. The workflow was configured around Alviere’s existing AML policies and procedures, so the Agent followed the same investigative process.

Consistent Investigative Workflow

The Agent applied a consistent process to every alert: read the alert, gather customer information, assess the source of funds and whether the activity was consistent with the customer’s profile, and investigate counterparties – going as deep as required.

Business-Context Configuration

The same investigative process was applied to each alert, with business-specific context added to the red-flag investigation stage. This gave the Agent an understanding of normal activity for each business line.

Network Graph Visualization

The Agent used network analysis to trace the flow of funds and identify connections between accounts and transactions. This helped the Alviere analysts identify potential risk patterns, including circular transfers, hub-and-spoke activity, and shared addresses or employers.

Results

Across three months of AML alerts spanning four business lines, Alviere found that Hawk’s Investigative Agent replicated analyst outcomes almost perfectly.

  • The Hawk Agent correctly escalated every confirmed true positive alert
  • It recommended closure for 98% of the alerts Alviere’s analysts had previously deemed false positives
  • It also flagged a number of other genuinely suspicious alerts

Examples of other suspicious activity flagged by the Hawk Agent included:

  • A 400x step-up in transfer size led the Agent to a counterparty account operating as a fan-in/fan-out hub, pulling funds from multiple sources and rapidly distributing them to unrelated beneficiaries abroad
  • A subject account that looked clean on its own, the kind a single-account review would reasonably close, turned out to be transacting with a counterparty running same-second round-trip transfers with a prior AML history 

  • A small, easy-to-overlook $500 transfer led to a cash-fed hub anchoring a larger layering network, fanning funds out to Mexico and to a broker with no relationship basis to the subject

     

See how Hawk's AI AML Investigative Agent matched analyst accuracy for Alviere — clearing 98% of false positives and catching 13 missed risks.
Share this page