

Each month, we spotlight a real-world AI task in Unit21, something many compliance and fraud teams are configuring and running inside their workflows today. This month: human trafficking, one of FinCEN's AML/CFT National Priorities, and one red flag from FinCEN's advisory taken through two AI tasks, Rule Writer Agent and AI Investigation Agent.
Most red flags in an advisory are written for people, not systems. This edition starts with one sentence from FinCEN and ends with a documented case.
Victims of human trafficking often have little contact with anyone but their traffickers, and FinCEN's advisory notes that a visit to a financial institution may be the only outside contact they get. The signals in the data are quiet. A wire here, a transfer there, each one unremarkable on its own.
FinCEN's advisory FIN-2020-A008 lists 20 red flags and describes four ways traffickers move money: front companies, exploitative employment practices, funnel accounts, and alternative payment methods such as prepaid cards and virtual currency. FinCEN estimates human trafficking generates around $150 billion a year worldwide. It is also clear that no single red flag proves anything.
That leaves two jobs. The first is detection: turn a red flag into a rule that catches the next case. The second is investigation: take an alert that has already fired and work out whether the pattern holds up. They need different tools, so this edition runs both, starting from red flag 19.
Red flag 19 reads: "Frequent transactions using third-party payment processors that conceal the originators and/or beneficiaries of the transactions." The hard word is "frequent." A person reading the advisory knows roughly what it means. A monitoring program needs a number.
Rule Writer Agent closes that gap. You paste the red flag in as written, and the agent asks five follow-up questions to pin down what frequent means in your data: which signal identifies a processor, how to count, what time window to apply, and which direction of money flow to watch. Then it builds the rule. No SQL, no schema knowledge, no engineering ticket.
AI Investigation Agent picks up from there. Once an alert fires, the agent works the case: it reviews the transactions, writes up the findings, and drafts a narrative for the analyst to edit.
Here is the rule Rule Writer Agent built from red flag 19:
The rule follows the same path as any other rule in Unit21. You can edit it, validate it, shadow test it, and then take it live.
From red flag to rule to documented case, in under a minute and a half.
The agent recommends. The analyst decides. Every finding links back to the transactions behind it, so any claim can be checked. The narrative is editable, and a full audit trail records each step the agent took.
Rules get the same care. Nothing goes live until you have reviewed, validated and shadow tested it.
Both tasks are available in the platform today. To try the flow on your own red flags:
No. FinCEN says no single red flag is a clear indicator, and the advisory asks institutions to weigh a customer's history and other red flags together. The rule surfaces activity worth a look. The analyst decides what it means.
Red flag 19 is about frequency and concealment, not size. Repeated wires through processors can add up to a pattern even when each one is small. You can adjust the rule before it goes live if your program needs a threshold.
The steps are the same: paste the red flag as written, answer the follow-up questions, and review the rule.
No. When an analyst does file, FinCEN asks filers to reference the advisory with the key term "HUMAN TRAFFICKING FIN-2020-A008" in SAR field 2 and in the narrative, and to select SAR field 38(h).
Human trafficking is hard to see in transaction data because each piece looks ordinary. A wire through a processor is a wire through a processor. The assembly work, putting a customer's history and several weak signals together, is where analyst time goes.
Here, detection started with one sentence from FinCEN's advisory, and the investigation ended with a narrative the analyst can edit and a complete audit trail. The same approach works for the rest of the advisory.
Read it. Detect it. Investigate it.
The AI Task Spotlight runs every month. Each edition covers one task from Unit21's library, covering what it does, how it works, and who it's for. If a task is solving a real problem for one team, it can probably solve the same problem for yours.
Want to learn more? Sign up for a demo of our AI. Alternatively, stay informed of our AI by signing up for our next AI Task Spotlight. Missed last month? Read Edition No. 12: Metrics on Demand.

Gal Perelman is the Product Marketing Lead at Unit21, where she spearheads go-to-market strategies for AI-driven risk and compliance solutions. With over a decade of experience in the fintech and fraud sectors, she has led high-impact launches for products like Watchlist Screening and AI Rule Recommendations.
Previously, Gal held marketing leadership roles at Design Pickle, Sightfull, and Lusha. She holds a Master’s degree from American University and a Bachelor’s from UCLA, and is dedicated to helping banks and fintechs navigate complex regulatory landscapes through innovative technology.