Financial institutions have begun to implement AI agents to enhance the quality of financial crime investigations. These agents can conduct investigations sequentially, document them comprehensively, and verify continuously – rather than over months with a small sample. Christopher Phillips, Director of Financial Crime Industry Engagement at Flagright, noted in an interview with PYMNTS: “AI doesn’t get tired, doesn’t get bored, and doesn’t forget to insert a reference for the search.” The result is “reliable, which is easy to verify.”
Traditional quality checks in the financial crime sector mainly work through sampling: compliance or audit examines a certain percentage of cases and extrapolates conclusions. Agent-based AI makes another model possible. “With AI, you can do everything,” said Phillips. “You can literally conduct a quality check on every case.” According to him, it is necessary to “rethink the investigation process entirely – from start to finish,” including documenting the workflow and determining where controls should be placed along the way.
As banks shift AI from recommendations to execution, the main question shifts. “The technology is mature,” said Phillips. “If applied correctly, it will change the game.” But maturity does not mean unlimited autonomy. “If you are going to close an account, recommend filing a suspicious activity report, or take any action with consequences for the client, that must be verified by a human,” he emphasized. Machines can investigate, search, identify patterns, and gather evidence, but consequences require a different standard. A human must ensure what Phillips calls “effective challenge”: understanding why the agent came to a conclusion and documenting why they agree or disagree. “AI agents are essentially employees. They need to be monitored like employees, and they should have the same limitations.”
Related articles
