Back to site

Quality assurance

A stratified sample of the agent’s dispositions is re-worked by a human reviewer. The sample is weighted toward auto-closed alerts, because an unjustified clearance is the failure a regulator will find.

Concurrence
80.0%
reviewer agreed with the agent
Sample size
5
cases re-worked this batch
Overturned
1
returned as training signal
Sampling rate
8%
auto-closed · 100% of escalations
ALT-2026-88104Marcus D. OyelaranStructuring
Escalate to L368%concur with note

Agree with the escalation. Would have cited the Brazil wire's invoice reference explicitly in the findings section — it is in the evidence but not in the prose.

R. Okafor · QA lead

ALT-2026-88063Ana Sofía RestrepoPEP exposure
Cleared72%concur

Source of wealth evidenced, RCA tier correctly applied. Narrative is filing-ready as drafted.

S. Lindqvist · L3

ALT-2026-88088Grover Lane Café LtdRound-amount patterning
Cleared97%concur

Clean false positive. Contract on file resolves the round-value pattern.

R. Okafor · QA lead

ALT-2026-88075Oksana PetrenkoSanctions nexus
Escalate to L373%concur

Correctly refused to clear on an unresolved alias match. Registration-number discriminator noted but appropriately not treated as dispositive.

MLRO office

ALT-2026-88052Lagos Swift Remit LtdHigh-risk geography
Cleared97%overturned

Disposition reversed to RFI. Corridor growth is explained, but the agent under-weighted the absence of originator data on the aggregated pool. Retraining sample.

S. Lindqvist · L3