One system of record for financial crime compliance.
From the first login to the last goAML filing, Sentinel connects detection, investigation and reporting, so nothing falls between a spreadsheet and an inbox.
Every movement of money, checked within seconds.
Send transfers, cash-outs, card payments, wallet movements, bill payments and reversals as they happen. Sentinel evaluates them against your rules and models and raises an alert the moment something fits a typology.
- Typologies that match local crime: structuring under the CTR threshold, rapid pass-through, fan-out to mules, POS cash-out spikes, KYC tier breaches.
- Per-customer baselines. "Unusual for this customer" is measured against their own history, not a one-size threshold.
- Logins and onboarding too. Brute force, impossible travel, VPN and TOR use, and identities reused across accounts.
Allow, review or block before the debit.
For the riskiest flows (payouts, withdrawals, first transfers to a new beneficiary) your backend asks Sentinel first. The answer comes back in milliseconds with a score and the reasons.
- Fail-open SDKs with a 300 ms timeout, so a slow network never blocks a genuine customer.
- Configurable thresholds for review and block, per institution.
- Shadow mode records what would have happened, without touching customers, until you're ready.
- Review at score
- ≥ 40
- Block at score
- ≥ 80
- Client timeout
- 300 ms · fail open
- Latency target
- p95 < 150 ms
- New rules start in
- Shadow
Write the rule your examiner asked about, without a ticket to engineering.
Combine conditions on amount, channel, counterparty, time of day, customer tier and risk, device and IP signals with AND/OR groups. Sentinel describes every rule in plain English, and you can backtest it against past transactions before it goes live.
- Backtest on real history to see what a rule would have caught, and how many alerts it would add.
- Audit trail of who changed which rule, and when.
Backtest, last 90 days: 41 alerts, 9 already confirmed as fraud in your cases.
Know who you're dealing with, and keep knowing.
Customers are screened at onboarding and again whenever a list changes. Fuzzy matching catches transliterations and spelling variants, while date-of-birth and country checks keep false positives down.
- OFAC SDN and UN Consolidated lists imported automatically.
- Nigeria Sanctions List and Nigerian PEPs, plus your own internal watchlists.
- Adverse media from global news sources over the last 12 months, queued for analyst review.
- Weekly rescreening of your whole customer base, plus on every list update.
OFAC SDN
US Treasury sanctions. Updated automatically.
UN Consolidated
UN Security Council list. Updated automatically.
Nigeria Sanctions List
Nigeria Sanctions Committee designations.
Nigerian PEPs
Politically exposed persons and associates.
Find the patterns nobody wrote a rule for.
An anomaly model learns normal behaviour for each customer from your own history. Once you've confirmed enough cases, a supervised model learns from your analysts' decisions too. Every score shows the factors behind it.
- Your model, your data. Each institution has its own model. Nothing is pooled across customers.
- Validated before promotion. New versions are tested against your confirmed cases before they replace the live one.
- Explanations on every alert that an analyst, auditor or examiner can follow.
- Precision at top 1%
- Recall on confirmed fraud
- Alerts vs. rules alone
Illustrative. Actual figures depend on your data and are shown in your dashboard.
Catch the fraudster before they open the account.
Verify BVN, NIN, CAC registration, bank accounts and phone numbers with name and date-of-birth matching. Device SDKs fingerprint each session and flag emulators, automation, rooted phones and anonymising networks.
- Pay per check from a prepaid wallet. Re-checking the same identifier within 30 days is free.
- Identity reuse detection when one BVN, NIN, phone or device appears across several accounts.
- Consent recorded with every verification, as the NDPA requires.
- Identifier
- 222••••••42
- Name match
- 0.96 · Match
- Date of birth
- Match
- Linked accounts
- 3 other wallets
- Device
- Emulator
Investigations with a paper trail an examiner will accept.
Group related alerts into a case, gather the transactions and screening hits, write the narrative and decide. Closing a case takes a second officer's approval, so no one signs off their own work.
- Maker–checker closure and role-based access for analysts, compliance officers and auditors.
- Complete audit log of notes, decisions, exports and configuration changes.
- One click to STR once a case is confirmed suspicious.
- Opened from 6 linked alertsMon 09:12
- Analyst added 14 transactions and a narrativeMon 11:40
- Marked suspicious · STR draftedTue 10:05
- Sent to compliance officer for approvalTue 10:06
File to the NFIU in the format it expects.
Sentinel generates Suspicious Transaction Reports from cases and Currency Transaction Reports for any date range as goAML XML, ready to upload. Board and regulator reporting comes from the same data.
- STR and CTR in goAML XML with your reporting-entity details filled in.
- CBN evidence pack: a PDF that maps your live configuration and activity to each capability in the CBN Baseline Standards for automated AML solutions.
- Daily compliance report emailed to your team every morning.
A fraudster caught by one member is flagged for all.
Members can share confirmed fraud signals (an account number, phone, device or BVN), and Sentinel warns every other member when the same identifier appears on their platform.
- Opt-in, governed by written terms. Only confirmed fraud is shared, and members can dispute an entry.
- No raw personal data changes hands. Identifiers are shared as keyed one-way hashes, so members only learn about a match.
- Used as a signal, not a verdict. A network match raises an alert for your team to review; it never blocks on its own unless you choose that.
See Sentinel on your own transaction patterns.
Book a 30-minute walkthrough with our team, or start in the free sandbox today and send your first event in minutes.