SAR dispositions that hold up under examination

Every suspicious activity alert ends in a decision. The disposition records what the institution concluded: the alert is closed with a rationale, or it moves toward a Suspicious Activity Report (SAR). The narrative is the written account of the activity and of the reasoning behind that decision.

SAR disposition oversight is the practice of checking those decisions and narratives against policy and against the evidence in the case. It asks whether the conclusion is supported, whether the rationale was recorded clearly enough to be understood later, and whether similar alerts were treated in the same way.

For financial crimes investigations, this is where quality is won or lost. A disposition that cannot be explained, or a narrative that leaves out what a reader needs, is a weakness that an examination can find long after the decision was made.

Quality control by sampling reads a small share of dispositions and narratives and assumes the rest look the same. Weak rationales, inconsistent decisions and incomplete narratives that fall outside the sample go unseen, and the first reader to find them can be an examiner.

Reviewing every disposition changes the question, from how a sample looks to how every decision was made. Problems are found while they can still be corrected, and patterns that no sample would reveal become visible across the whole population of alerts.

Prisma evaluates each disposition and each SAR narrative against the policy and regulatory expectations your institution has set, detecting violations and ambiguity, with the reasoning behind each finding. A violation is recorded with the reason it breaks policy. Where the evidence does not allow a clear answer, Prisma flags the ambiguity and escalates it for a human decision instead of forcing certainty.

The evaluation looks at whether the decision is supported by the evidence, whether the rationale is recorded, whether the narrative answers who, what, when, where, why and how in line with FinCEN's narrative guidance, and whether the decision is consistent with how similar alerts were handled.

Oversight is the first step: every disposition and narrative is read and judged. Intervention follows when something falls short. The weak disposition is escalated for review, and consequences and feedback are applied, to the person or the AI that produced it, in time to fix it instead of being discovered after the fact.

Remediation closes the loop and improves behavior. Consequences and feedback are fed back into how future dispositions are made, so the same weakness does not repeat and quality improves and stabilizes over time, for people and for AI alike. The evidence of how each decision was judged is kept, giving you an audit trail for every disposition, ready when it is asked for.

Narratives are increasingly drafted by AI agents, and alerts are increasingly triaged by them. Prisma applies a single governance standard to work produced by people and by AI, so an AI-drafted narrative is read against the same policy and checked for violations and ambiguity in the same way as one written by hand. As automation takes on more of the work, the oversight stays the same.