1. The transaction arrives

    TXN-Monitor receives the transaction as it is posted: from Apache Fineract business events, from an event stream your systems publish to, or through the REST API. Each one carries its reference, account, amount, type, channel and counterparty.

    A transaction with a reference already seen is recognised and not screened again.

  2. Context is gathered

    Before any rule runs, the system pulls together what it knows: the customer’s KYC profile and tier, their recent transaction velocity, the beneficiary, and sanctions list matches for both sides.

    Customer and KYC data can be synced from your existing customer database.

  3. Rules and the model evaluate it

    All active rules run at once across AML, CFT, fraud, KYC and NFIU reporting. Separately, the anomaly model scores how unusual the transaction is for this customer.

    The model is retrained every night and switched in without downtime.

  4. A risk score and band are assigned

    Rule results set the score; the model can add a limited uplift but can’t on its own push a transaction above Medium. The result is a score from 0 to 100 and a band.

    Critical 75–100, High 50–74, Medium 30–49, Low 0–29.

  5. Alerts are raised

    Transactions that need attention become alerts, prioritised by band and assigned to analysts. The right people are notified by email, SMS or Slack.

    Unattended alerts escalate automatically.

  6. The case is investigated

    Analysts investigate, record findings and a disposition, and send the case to the MLRO for approval. Every step is written to the audit trail.

    Cases have owners, approvers and due dates.

  7. Reports are prepared

    Where a reporting rule fires, a draft STR or CTR is prepared as PDF and XML for review, with reminders before the filing deadline.

    Results are also published as event streams for your other systems.