Clari5

Machine learning models packaged by channel and product score transactions in real time, and Genie layers generative and agentic AI over investigations. Rules, scenarios and case management would still run without them, so the models power the detection rather than being the whole product.

The thresholds that trigger a hold are not published. In core banking, suspicious transactions are stopped or held before they advance and a case is created for the fraud investigation team, watch list hits can block a transaction while an alert is raised, and regulatory reports can be viewed and modified in a principal officer workbench before submission. One path runs without people: the cybercrime complaint solution identifies accounts and marks liens from portal complaints with no manual intervention.

Validation reports, accuracy figures and a mapping to model risk guidance are not published. Genie explains in plain language why an alert was triggered, and real time decisions are described as explainable and traceable, with layered event enrichment to cut false positives.

Khan Bank, XacBank, Punjab National Bank, Mashreq Bank, Ujjivan Small Finance Bank, dfcu Bank and Fino Payments Bank are named deployments. Figures stay at the platform or product level and attach to no named bank: more than 10 billion transactions and a billion accounts, investigations up to 70 percent faster with Genie, and manual case closure time cut 80 percent with Maestro.

Pattern intelligence is drawn from cross institutional analysis, which means one bank's experience shapes what another bank's models see, with no stated boundary on what is shared or how it is anonymized. Training and retention terms for Genie's inputs go unstated.

Data handling terms for the transaction and customer records banks run through the platform are not set out in its product material. The privacy statement attached to its contact forms covers enquiries rather than bank data.

Security certifications, audit reports and a trust center for the platform are not published. Maestro redacts personal data before analysis, encrypts with 256 bit keys and keeps all call data inside the bank's infrastructure.

Clari5 sits inside regulatory filing paths: its goAML module generates suspicious transaction and currency transaction reports in the standard of the United Nations Office on Drugs and Crime, tracks submissions and uploads acknowledgements from financial intelligence units, and dfcu Bank in Uganda runs its goAML reporting on it.

Fairness or bias testing of the scoring models and of Genie is not published. Decisions are described as explainable and traceable, a transparency commitment rather than a measure of who gets flagged, on a product whose behavioral biometrics and customer scoring touch every account holder.

A guarantee, indemnity or correction route behind a decision is absent. What exists is traceability: decisions are described as explainable and traceable, and held transactions become cases a bank's investigators can review and release.

The language models behind Genie, Maestro's speech to text and any managed service host go unnamed. Detection models are Clari5's own, packaged by channel and product, Genie is built natively on the Clari5 platform, and Maestro runs an eight stage AI pipeline inside the bank.

Specific core banking products are not named. Monitoring runs inside core banking, employee fraud tools integrate core banking, customer relationship management and human resources systems, regulatory reporting follows the goAML format used by financial intelligence units, and the cybercrime complaint solution connects to India's National Cybercrime Reporting Portal. The platform is offered on IBM LinuxONE and Power10 and on Red Hat Enterprise Linux and JBoss.

Residency terms for managed services are not stated, which matters for banks across 30 countries with their own rules. Maestro runs on a bank's premises with no data leaving its infrastructure, or as a managed service, and the main platform runs on bank infrastructure such as IBM LinuxONE, Linux on IBM zSystems, IBM Power10 and Red Hat Enterprise Linux.

Engagement runs through a demo request. Price, tiers and the unit of charge for the fraud, anti money laundering and Genie products are unpublished, so cost becomes known only in a sales conversation.

Banks are the core buyers, and insurers and payments businesses each get their own products: anti money laundering for insurance, and payment card fraud management for issuers, processors and acquirers. Named deployments run from Mongolia and India to Uganda, with market material written for Nigeria, Kenya, Indonesia and Egypt.