Shufti launches AI-driven transaction trust monitoring for fraud and AML teams

Shufti has launched Transaction Trust Monitoring, an agentic FRAML engine that ties transaction checks to verified identity and drafts regulator-ready reports in one workflow. The product aims to help banks, fintechs, payment firms, and crypto businesses spot risk faster, route alerts by severity, and cut manual work for compliance teams.

Why it matters: - Shufti’s Transaction Trust Monitoring is built to reduce manual work in fraud and anti-money laundering operations by linking transaction checks to verified identity and moving a case from alert to report in one workflow. - The system is designed to help compliance teams handle higher transaction volumes and more jurisdictions without adding headcount at the same pace. - The launch matters for banks, fintechs, payment service providers, remittance firms, virtual asset service providers, neobanks, and iGaming operators that need faster monitoring and more defensible audit trails.

What happened: - Shufti announced Transaction Trust Monitoring, an agentic FRAML engine that scores transactions, routes investigations, and drafts regulator-ready reports. - The platform is available now over RESTful API and SDK, with sandbox access and on-premise deployment. - Shufti also released an AI rule builder that turns plain-English risk descriptions into backtested monitoring rules. - The launch is featured in the Shufti Innovation Drop, alongside a live product demonstration. - A video demo is available here.

The details: - Transaction Trust Monitoring scores each transaction against the identity Shufti verified at onboarding. - The engine integrates with existing KYC and case management systems rather than replacing them. - FRAML combines fraud and anti-money laundering controls into a single discipline, and TTM scores both signal types in the same pass. - The system owns the full stack end to end, with no acquired modules and no third-party stitching. - Each transaction is checked against a biometrically verified customer, with deepfake and synthetic identity defences at onboarding and behavioural biometrics plus device fingerprinting after login. - The platform screens against more than 4,000 watchlists and more than 6 million PEPs, with sanctions coverage across OFAC, OFSI, EU CFSP, DFAT and the UN Consolidated List, plus multilingual contextual adverse media. - Shufti says TTM evaluates more than 1,600 data points per transaction in under 500 milliseconds. - The system uses location, device, banking and threshold data, plus more than 15 velocity checks running from one minute to 180 days. - Every transaction receives one risk score from 0 to 100. - Every alert includes a case summary, tier classification, false-positive assessment and recommended next action. - The AI rule builder calibrates rules by industry, regulation and risk appetite, then backtests them against 90 days of history to estimate alert volume, precision and fraud caught by value. - Scores of 75 and above go to the MLRO, 50 to 74 go to L2 investigation, and 25 to 49 go to L1 triage. - Flagged transactions can be placed on a soft hold and sent to an analyst queue instead of being automatically declined. - The AI triage co-pilot drafts narratives and recommends actions, but does not file reports itself. - Twenty-four ABSOLUTE rules always trigger and cannot be suppressed. - Every decision is held on a five-year immutable audit trail. - Unscoreable transactions return a NOT ASSESSABLE verdict. - Behaviour is measured against each customer’s own pattern using a rolling 90-day baseline across 16 dimensions. - Linked devices, IP addresses and contact details are used to expose multi-accounting and money mule networks. - Four AI agents are live across the lifecycle, with three more in development for rule recommendation, data readiness and regulatory change monitoring. - MCP integration allows external AI tools to drive a firm’s monitoring setup. - Suspicious activity and suspicious transaction reports are generated in the regulator’s native XML, including FinCEN BSA XML 2.0 and goAML. - The MLRO reviews, edits and authorises each report before submission, with no manual XML assembly or re-keying. - Each filing is logged with case data, analyst notes and score explanations. - Threshold reports are obligation-based rather than suspicion-based, so they auto-batch daily under a deadline timer, with MLRO oversight and the ability to pause any batch. - TTM also supports SWIFT and SEPA flows, cumulative player history, corridor and agent-level monitoring, wallet screening, on-chain risk scoring and Travel Rule support across 240 countries and territories. - The product can be configured as a standalone solution or as part of Shufti’s broader compliance lifecycle platform. - Shufti also published A Guide to Risk-Aligned Transaction Monitoring.

Between the lines: - The product positions Shufti closer to an end-to-end compliance operating system than a point solution, with onboarding, monitoring and reporting tied together on one engine. - The AI rule builder and backtesting tools suggest Shufti is targeting teams that want faster rule creation without giving up governance or auditability. - The role-based routing model and MLRO sign-off keep humans in control at the highest-risk stage, which may help the product fit more conservative compliance programs. - The emphasis on immutable logs, native XML reporting and score explanations signals a focus on regulator-facing defensibility as much as detection speed. - “Compliance teams are covering more jurisdictions and more volume without proportionate headcount,” said Frayam Asif, chief technology officer at Shufti.

What’s next: - Shufti said three additional AI agents are in development for rule recommendation, data readiness and regulatory change monitoring. - The company said external AI tools can connect through MCP integration to drive monitoring setup. - Firms can adopt TTM alone or deploy it alongside Shufti’s full compliance lifecycle platform. - Shufti is directing users to its guide and demo materials for more detail on risk-aligned transaction monitoring.

The bottom line: - Shufti is betting that compliance teams want one AI-assisted workflow for monitoring, investigation and reporting, with enough controls to satisfy regulators and enough automation to keep pace with transaction growth.