/ PRODUCT — AML ANALYSIS
Read the whole book of business, not one receipt at a time.
AML Analysis aggregates every transaction the way the regulations do — same-day, per customer, across every agent and license — then links senders to beneficiaries, clusters the names people use to hide, and puts an AI analyst on the patterns. Structuring, layering, and high-risk clusters surface before a report is ever late.
THE DESK OPENS TO FOUR NUMBERS
/ THE PIPELINE
Every source — agents, brands, and Principals — flows into one stream, deduped, scored, and linked into a single ledger you monitor and file from.
/ AML-ANALYSIS
Every license, one ledger.
Transaction analysis across every agent, state license, and reporting agency — normalized into a single stream you can monitor and file from.
/ WHAT IT DETECTS
Six ways money hides. All of them, surfaced.
These aren’t alerts you write rules for — they’re computed across the whole book every time you open it.
Same-day totals, not lifetime totals
We aggregate every transaction by sender within a single calendar day — the way FinCEN thresholds actually work. A customer who never trips a limit on one receipt, but whose same-day activity crosses $3,000, is flagged KYC-required; cross $10,000 and it flags CTR-required — before the report is missed.
The same person under three spellings
Structuring often hides behind near-duplicate names so no single spelling aggregates past a threshold. Our fuzzy matcher clusters near-duplicate sender and beneficiary names — case, accents, punctuation, and word order normalized — and scores each cluster’s similarity. Tune sensitivity Strict / Standard / Loose.
Who’s connected — directly and indirectly
A sender↔beneficiary graph with connected-component analysis: you see not just direct counterparties but everyone reachable through the network. Fan-in (many senders → one beneficiary), fan-out, layering, shared beneficiaries, and circular flows surface as structure, not spreadsheet rows.
Where the dollars concentrate
Every transaction is placed in an amount band, by location, so threshold exposure and volume concentration are visible at a glance — from small-dollar noise to the transactions that carry reporting weight.
Corridors measured against your own baseline
Transactions map by destination country — Mexico, Guatemala, Honduras, and the corridors your business actually runs. A customer’s average to a country is benchmarked against the agent’s own average to that same country, so an out-of-pattern corridor stands out.
Across agents, licenses, and locations
The same customer transacting under different agents or state licenses is aggregated into one view. Multi-license and multi-country activity is flagged, so splitting across licenses doesn’t split the picture.
/ CASE STUDIES
What it looks like when it catches something.
The split that stayed under $10k
A sender ran three receipts the same afternoon, each below the single-ticket limit. Per receipt, nothing. Same-day aggregation summed them and raised a CTR-required flag the same day.
One customer, three spellings
Three near-identical names moved money to the same beneficiary. Fuzzy clustering pulled them into one identity at 88% similarity — revealing an aggregate no single spelling would have crossed.
Twelve senders, one beneficiary
A fan-in pattern hid across two licenses: twelve unrelated-looking senders, one shared beneficiary. Connected-component analysis surfaced the cluster as a single structure to investigate.
/ AML WATCH
Walk the network, hop by hop.
An interactive graph of senders and beneficiaries. Pan and select, marquee a group, or isolate a node to its connected component. Scrub the timeline to watch the network build day by day, and expand outward up to six hops and 150 entities.
AI-generated observations for investigative review — not a determination or legal conclusion. Verify against source records before any action.
/ AI PATTERN ANALYSIS
Select a cluster. Get an analyst’s read.
Marquee-select entities in AML Watch and run Analyze. An AI model, prompted as a senior AML/BSA analyst, reads the selection — using same-day figures, never lifetime totals — and returns structured, hedged observations you can act on:
It names relationships, describes patterns and money movement, calls out risk factors and pending alerts, and leaves watch points — in the careful language of “consistent with” and “may indicate.” It never reaches a conclusion; that’s your job, with the record in front of you.
/ HOW WE ANALYZE OPERATIONS
From raw receipts to a structured read.
Ingest
Pull the full scoped book of transactions — sender, beneficiary, amount, date, country, license, agent.
Place
Bucket each transaction by amount band, destination country, license, and sender-per-day.
Aggregate
Same-day per sender for KYC / CTR thresholds; across licenses and locations into one ledger.
Relate
Build the sender↔beneficiary graph and connected components; cluster near-duplicate names.
Review
Explore AML Watch, select entities, and let the AI analyst return observations for review.
Analysis runs on demand over your full scoped book — recomputed when you open the view or sync KYC — not as a black-box score. Every signal is traceable back to the transactions behind it.
/ FROM SIGNAL TO ALERT
Detect → flag → request → escalate.
Every customer carries a status. Request KYC and the customer moves into AML Alerts in Compliance Alerts — nothing falls through.