AI task spotlight | Edition no. 06: Custom Structuring Detection
Every two weeks, we spotlight an AI task from Unit21's task library, something many compliance and fraud teams are configuring and running inside their workflows today.
This edition is a little different. Instead of a single shipped capability, it covers a live build: one specific structuring check, described in plain English, backtested against real alerts, and put to work, the same way any team can build one for their own program. The task is Custom Structuring Detection, and the problem it solves isn't that structuring is hard to define. It's that structuring means something different at every program, and a single pre-built check can't hold all of those definitions at once.
Why Structuring Detection Isn't One-Size-Fits-All
Structuring is the practice of splitting a large transaction into smaller ones to stay under the $10,000 threshold that triggers a Currency Transaction Report. It's one of the most common typologies behind a SAR filing, and one of the most time-consuming to document: an investigator has to review a transaction window, calculate how close each amount comes to the reporting line, check whether deposits on the same or consecutive days add up to a reportable total in aggregate, and write a justification for whatever they conclude.
What makes this harder than it looks is that the pattern doesn't look the same everywhere. A remittance company watches for repeated small amounts moving through the same corridor; that repetition itself is the signal. A payroll company already has repeated payments by design; paychecks recur on a schedule, so the same detection logic would flag every pay cycle. There, the signal isn't repetition; it's an anomaly on top of the expected pattern. A single structuring task, built once and shipped to everyone, has to either miss one of those patterns or generate false positives chasing the other.
That's the gap a custom task closes.
Introducing Unit21's AI Task: Custom Structuring Detection
Custom Structuring Detection: What it does
Custom Structuring Detection was built from a plain-English description of exactly what structuring means for one specific program, then backtested against five real alerts before it ever touched a live queue. It reviews deposit and withdrawal activity, checks how close each transaction is to the $10,000 reporting line, and looks for same-day or consecutive-day activity that, in aggregate, clears that threshold even when no single transaction does.
The agent automatically reviews:
- Deposit and withdrawal amounts against the $10,000 CTR reporting line, banding results by proximity (near-threshold, below-threshold, well below)
- Same-day and consecutive-day deposits that don't trip a single transaction but add up to a reportable total combined
- Whether a backtested alert actually has usable transaction data before drawing any conclusion from it
- Full transaction history per entity, not just the flagged transaction in isolation
Custom Structuring Detection: What the agent outputs
- A structured pattern assessment naming exactly which transactions land in a near-threshold band, with the dollar gap to $10,000 for each
- A flagged combined total when same-day or consecutive-day deposits clear the reporting threshold in aggregate, even if no single transaction does
- An explicit "no usable data" call-out for any backtested alert that can't support a conclusion, instead of a forced or fabricated result
- A full reasoning trail, every step the agent took to get from raw transactions to its conclusion, visible end to end
Custom Structuring Detection: Why this matters
Backtested against real alerts, this task found what it was built to find: a transaction $253 under the $10,000 line, and $18,085 in same-day deposits that individually stayed under the limit but cleared it combined, exactly the kind of pattern a SAR gets filed on. It also did something a forced result never would: on one alert without usable transaction data, it said so, instead of guessing.
That combination- a check scoped to exactly what one program means by structuring, and an agent that reports what it doesn't know as readily as what it does- is what makes a custom task different from a template. A remittance company and a payroll company can both build a structured check with the same tool and end up with two different tasks because they're not actually looking for the same thing.
Because it's built with Unit21's Agentic Task Builder, this runs inside your existing workflow, in plain English, with no engineering ticket and no waiting on a template that doesn't quite fit. And because every step of its reasoning is visible, investigators can act on its findings, challenge them, or escalate them with the backing to defend that decision under examination.
About the AI task spotlight series
The AI Task Spotlight runs every two weeks. 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.
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.
