IRIS 7
Platform
The Governance and Decision Architecture Behind Iris 7
Iris 7 enforces governance through:
The Iris 7 Platform Model
Agent Runtime & Orchestration
AI Agents execute multi-step investigative workflows across high volumes, with managed sequencing, tool and API calls, retry logic, and workload balancing to ensure reliable performance.
Unified Context Layer
Consolidated decision context is created by bringing together customer data, transaction history, network relationships, sanctions and AML sources, and relevant external intelligence.
Workflow & Operational Integration
Decisions and supporting artifacts are delivered directly into case management systems, screening platforms, monitoring environments, and downstream operational processes.
Governance & Assurance Framework
Policy boundaries, QA sampling, drift monitoring, version control, and complete audit traceability are applied across all AI Agent activity to maintain institutional oversight.
Context Layer
Each decision begins with a unified, case-specific information foundation. Relevant internal systems and external intelligence are consolidated into a coherent decision context, reducing fragmentation and variability.
Reasoning Engine
Institutional policies, investigative logic, and risk standards are applied in a structured and repeatable manner. Decisions reflect policy intent in context, without the inconsistency inherent in manual interpretation.
Decision Framework
Outcomes are produced within institution-defined approval limits, with each including supporting evidence and a complete reasoning path to ensure transparency for compliance, audit, and model risk.
Learning-Driven Reasoning
Iris 7 does not rely on rigid rules or predefined schemas that require rebuilding when policy or risk changes. Investigative reasoning adapts as typologies evolve and regulatory interpretations shift, without requiring rule rewrites, schema rebuilds, or ontology reconfiguration.
Continuous Learning Under Governance
Iris 7 evolves alongside policies, typologies, and investigative standards. Improvements are introduced through governed change controls, ensuring the platform adapts to evolving risk while remaining stable, inspectable, and regulator-ready.
Decision-Specific Context
Decisions are formed using dynamic, decision-focussed context rather than monolithic enterprise models. Only the signals, relationships, and risk indicators relevant to the decision are assembled, mirroring experienced investigators.
Traditional
FCC Systems
Starting Point
Requires rule-writing, model tuning, and analyst training before reliable outputs emerge
Decision Output
Produces alerts requiring analyst investigation and interpretation
Consistency
Outcomes vary
by analyst skill, workload, and region
Context Handling
Data gathered across fragmented systems during investigation
Governance
Oversight applied after decisions through manual quality assurance reviews
Adaptability
Policy updates require rule changes or model rework
Operational Model
Human-led investigation supported by tooling
Silent Eight
Iris 7
Starting Point
Deploys with embedded financial crime reasoning shaped by live production use
Decision Output
Produces policy-bound, evidence-backed
decisions within defined authority thresholds
Policy Application
Policy logic applied consistently
within agent reasonin
Consistency
Outcomes follow structured reasoning and defined policy standards
Context Handling
Unified, decision-specific context assembled before reasoning
Explainability
Complete reasoning path, applied policy logic, and retained decision record
Governance
Policy boundaries, thresholds, escalation rules, and traceability embedded into decisioning
Scalability
Decision capacity scales through parallel agent execution within governance controls
Adaptability
Reasoning adapts through governed updates aligned to policy and investigative practice
Operational Model
AI Agent-led decisioning with human oversight and accountability
