IllustrativeThis case study is anonymized and modeled after a real prospect conversation. No customer is identified or quoted. Real, named stories will replace these as pilots go live.
Switched from a vendor screening tool with a 23% MLRO escalation rate to a weighted-factor reasoning engine that explains every match.
23% → <0.5%
11 min → 90 s
OFAC + EU + UN + UK + OpenSanctions
A virtual-asset service provider was screening ~10,000 deposit addresses per day against OFAC SDN + OpenSanctions + the firm's internal adverse-wallets list. The legacy screening vendor returned hit / no-hit only; analysts had to research every match by hand. 23% of matches escalated to the MLRO, and the MLRO had no way to triage the queue.
The fintech turned on RiskSonnar's sanctions reasoning engine. Each match is scored against six weighted factors (name similarity, country overlap, attribute overlap, watchlist authority, age of the listing, internal corroboration). The LLM produces a one-paragraph rationale grounded in those factors, and the analyst sees exactly why a match scored where it scored. Crypto-specific factors include chain-analytics overlap and on-chain heuristics from the wallet-screening module.
Match volume stayed the same; MLRO escalations dropped to effectively zero because L1 analysts could now dispose 'no match' confidently with a defended rationale. Mean disposition time per match fell from 11 minutes to 90 seconds. The audit chain captures the factor weights at the time of disposition, which the firm's external auditor accepted as evidence under FATF R.10.
“Our MLRO used to wake up to a 200-deep escalation queue. Now she sees three. The factor breakdown means the L1 analyst can defend every disposition against the audit pack — we don't lose context on the way up.”
The synthetic-data sandbox runs the same surfaces this story used, against deterministic test data. No customer setup needed — the session auto-expires after 60 minutes.