Iris is a deterministic signal-processing engine that detects harmful behavioral patterns—such as coercion, exploitation, and fraud—across messages, transactions, channels, and time. Unlike generative AI tools, it provides auditable, reproducible results with calibrated confidence scoring and a full evidence trail. The platform maps detections across six behavioral dimensions, including control intensity, manipulation sophistication, and escalation trajectory.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional conversation analytics, transaction monitoring, and case review tools examine events in isolation, missing harmful behavior that only becomes apparent when exchanges are viewed as cumulative patterns across messages, transactions, channels, and time. This structural blind spot allows coercion, exploitation, and fraud to go undetected until after significant harm has occurred.
Solution
Iris is a deterministic signal-processing engine with calibrated confidence scoring, purpose-built for institutions that need auditable answers rather than generative guesses. The platform represents behavioral patterns at signal level, enabling detection across exchanges rather than isolated messages. Every detection carries calibrated confidence scores, ensuring the same input produces the same reproducible result. Each output includes a complete evidence trail that can be traced back to source signals and the regulatory obligations they help evidence. Iris maps every detection to one or more of six behavioral dimensions—control intensity, manipulation sophistication, exploitation severity, escalation trajectory, network manipulation, and evasion sophistication—creating a consistent language for patterns across financial services, safety, abuse, and exploitation contexts.
Target Audience
Primary customers are institutions in financial services, safety, abuse, and exploitation contexts that require auditable, reproducible detection of harmful behavioral patterns across digital conversations and transactions.
Features
- Signal grammar that represents behavioral patterns at signal level for cross-exchange detection
- Calibrated confidence scoring layer ensuring reproducible, deterministic outputs
- Complete evidence trail tracing every detection back to source signals and regulatory obligations
- Six behavioral dimensions including control intensity, manipulation sophistication, exploitation severity, escalation trajectory, network manipulation, and evasion sophistication
- Changepoint detection and cycling pattern analysis for escalation trajectory monitoring
- Iris Authorities Selection Methodology (ASM) that identifies, assesses, and classifies reference authorities per vertical, starting from primary regulators and broadening by onward citation
- Detection capabilities spanning coded language, channel-switching, and evidence destruction patterns