PiR2-IT · Agentic AML Hyperautomation

Agentic AML Hyperautomation: live demo

Interactive demo of PiR2-IT Agentic AML Hyperautomation: onboarding and screening, real-time monitoring, alert triage, conflicts of interest, organised-network detection, money tracing, procurement integrity, copilot and AI governance, with human sign-off. Synthetic data.

Synthetic data only. No real customers, officials, companies, transactions or institutions. A PiR2-IT product, demonstrator by Marius Mihail Russo-Got.

Inside the demo

Access control — Access depends on the role and on attributes of the user, the data and the context: jurisdiction, clearance level, case…
Insider threat — Every action in this demo is recorded per user and session and checked against behavioural rules. The alerts below come…
Compliance matrix — Each requirement is mapped to the screen where it can be seen working and to the audit events that evidence it in this…
Data sovereignty — Choose where the platform runs. The diagram and the table show where customer data, encryption keys and the AI models…
Connectors & APIs — Sources and destinations the operational platform connects to, and the public API. In this demo they are placeholders:…
Training (LMS) — Short courses built from anonymised cases, each with a quiz. Completion is recorded per user and feeds the mandatory…
Executive view — Key performance and risk indicators across monitoring, investigations, procurement integrity and governance, plus a…
Onboarding & screening — Digital identity verification (document OCR, liveness, face match), real-time screening against sanctions and PEP lists…
Monitoring & rules — ISO 20022 messages (pacs.008) are evaluated against the rule set as they arrive. Thresholds adapt to country, channel…
Alert queue — Alerts raised by the existing monitoring system. Open an alert to watch the agents work step by step, or run them on…
L2 cases — Alerts the agents could not close. Each arrives with enrichment, typology scores, a counterparty graph and a drafted…
SAR / STR — Agents draft the narrative, citing transaction IDs, rules triggered and typologies. Only the MLRO can approve and file;…
KYC · UBO resolver — The agent traverses (simulated) company registries and ownership documents, multiplies shareholdings along every path…
Conflicts of interest — Every award is cross-checked: officials and committee members of the buyer (or the bank employee approving credit) are…
Organised networks — Entities are linked through shared devices, phones, addresses, directorships, account ownership and money flows. Each…
Trace X → Y — Time-respecting path search over the transaction graph: every hop must happen after the previous one and within the…
Public procurement — Contracts in open format (OCDS) are checked for bid splitting under the direct-award threshold, unit prices above the…
Investigation copilot — Questions in plain language are answered only from documents the user is allowed to see (role- and attribute-based…
AI governance — Every model is registered, validated and approved before use. Each decision can be explained factor by factor, model…
Client RFI & FIU exchange — Requests for information go to the customer through a secure portal with deadlines and reminders; overdue items…
Security & resilience — Access control, insider-threat and evasion signals, and the resilience targets of the operational platform.
Architecture — Event-driven, tool-augmented agents on top of the institution's existing systems. Agents decide which data to query,…
Audit log — Every agent disposition and human decision is recorded with the previous entry's SHA-256 hash, so any later change…
Policy & thresholds — The auto-closure gate is a policy decision of the institution, not of the model. Changes are logged and apply to the…
38-module map — Where each module of the unified specification can be seen in this demo: interactive, described with illustrative data,…