Agents in AML
A Production Playbook from the Investigator’s Seat
AML agents are moving from demos into production in real compliance programs. The hard part is no longer getting an AI agent to produce a plausible alert disposition, but is getting that disposition to hold up against SOPs, analyst review, QC, model risk validation, and examiner scrutiny months later.
In this playbook, Seth Sattler ( Head of Compliance/AML at Oscilar) lays out a practical guide for deploying AML agents in production. Drawing on live deployments and active evaluations across banks, fintechs, payments platforms, and digital asset firms, Seth explains how compliance teams can use agents to reduce investigative assembly work while keeping analysts responsible for judgment, accountability, and final decisions.
Inside you'll find:
The build or buy decision: Understand the three main paths for adopting AML agents and why platform-native agents built on a unified risk memory are better positioned to survive operational, regulatory, and model risk scrutiny.
What separates production-grade AML agents from demos: Learn the questions every compliance leader should ask, including whether the agent can access full customer context, reason against institution-specific SOPs, cite its evidence, apply confidence thresholds, and produce an audit trail that examiners can reconstruct.
A practical deployment path for real AML teams: See how to configure agents against your customer population, run a structured one-week POC, measure disposition quality and drift, and phase rollout from L1 alert review into case narratives, SAR drafting, sanctions, PEP screening, adverse media, EDD, and beyond.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><g d="M 0 48 L 0 0 L 48 0 L 48 48 Z M 44.447 0 L 3.544 0 C 1.584 0 0 1.547 0 3.459 L 0 44.531 C 0 46.444 1.584 48 3.544 48 L 44.447 48 C 46.406 48 48 46.444 48 44.541 L 48 3.459 C 48 1.547 46.406 0 44.447 0 Z M 14.241 40.903 L 7.116 40.903 L 7.116 17.991 L 14.241 17.991 Z M 10.678 14.869 C 8.391 14.869 6.544 13.022 6.544 10.744 C 6.544 8.466 8.391 6.619 10.678 6.619 C 12.956 6.619 14.803 8.466 14.803 10.744 C 14.803 13.013 12.956 14.869 10.678 14.869 Z M 40.903 40.903 L 33.787 40.903 L 33.787 29.766 C 33.787 27.113 33.741 23.691 30.084 23.691 C 26.381 23.691 25.819 26.588 25.819 29.578 L 25.819 40.903 L 18.712 40.903 L 18.712 17.991 L 25.537 17.991 L 25.537 21.122 L 25.631 21.122 C 26.578 19.322 28.903 17.419 32.362 17.419 C 39.572 17.419 40.903 22.163 40.903 28.331 L 40.903 40.903 Z" fill="transparent" height="48px" id="adFk4VB3K" width="48px"><path d="M 0 48 L 0 0 L 48 0 L 48 48 Z" fill="transparent" height="48px" id="DghPi7XP3" width="48px"/><path d="M 44.447 0 L 3.544 0 C 1.584 0 0 1.547 0 3.459 L 0 44.531 C 0 46.444 1.584 48 3.544 48 L 44.447 48 C 46.406 48 48 46.444 48 44.541 L 48 3.459 C 48 1.547 46.406 0 44.447 0 Z M 14.241 40.903 L 7.116 40.903 L 7.116 17.991 L 14.241 17.991 Z M 10.678 14.869 C 8.391 14.869 6.544 13.022 6.544 10.744 C 6.544 8.466 8.391 6.619 10.678 6.619 C 12.956 6.619 14.803 8.466 14.803 10.744 C 14.803 13.013 12.956 14.869 10.678 14.869 Z M 40.903 40.903 L 33.787 40.903 L 33.787 29.766 C 33.787 27.113 33.741 23.691 30.084 23.691 C 26.381 23.691 25.819 26.588 25.819 29.578 L 25.819 40.903 L 18.712 40.903 L 18.712 17.991 L 25.537 17.991 L 25.537 21.122 L 25.631 21.122 C 26.578 19.322 28.903 17.419 32.362 17.419 C 39.572 17.419 40.903 22.163 40.903 28.331 L 40.903 40.903 Z" fill="rgb(20, 20, 20)" height="48px" id="Qi9pnIQwR" width="48px"/></g></svg>)
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 34.653 0 L 41.4 0 L 26.659 16.847 L 44 39.772 L 30.422 39.772 L 19.788 25.868 L 7.62 39.772 L 0.869 39.772 L 16.635 21.752 L 0 0 L 13.922 0 L 23.535 12.709 Z M 32.285 35.734 L 36.023 35.734 L 11.891 3.826 L 7.879 3.826 Z" fill="rgb(20, 20, 20)" height="39.77194px" id="A8r2uGwai" transform="translate(2 3.808)" width="44px"/></svg>)
