AI Fraud Detection & Human Review

AI → Human

Real-time transaction scoring with human review for edge cases.

5 nodes · 4 edgesfinance

agenthumanevent

Visual

Transaction Receivedevent

Payment or transfer initiated.

↓sequential→ Fraud Risk Scoring

Fraud Risk Scoringagent

ML model assigns 0-100 risk score.

↓conditional→ Auto-Approve

↓conditional→ Auto-Block

↓conditional→ Fraud Analyst Review

Auto-Approvesystem

Score < 30: approve immediately.

Auto-Blocksystem

Score > 90: block and notify.

Fraud Analyst Reviewhuman

Score 30-90: human reviews transaction details.

Open in Editor

uc-fraud-detection.osop.yaml

osop_version: "1.0"
id: "fraud-detection"
name: "AI Fraud Detection & Human Review"
description: "Real-time transaction scoring with human review for edge cases."

nodes:
  - id: "transaction"
    type: "event"
    name: "Transaction Received"
    description: "Payment or transfer initiated."

  - id: "score"
    type: "agent"
    subtype: "llm"
    name: "Fraud Risk Scoring"
    description: "ML model assigns 0-100 risk score."
    timeout_sec: 2

  - id: "auto_approve"
    type: "system"
    name: "Auto-Approve"
    description: "Score < 30: approve immediately."

  - id: "auto_block"
    type: "system"
    name: "Auto-Block"
    description: "Score > 90: block and notify."

  - id: "human_review"
    type: "human"
    subtype: "review"
    name: "Fraud Analyst Review"
    description: "Score 30-90: human reviews transaction details."
    security:
      approval_gate: true

edges:
  - from: "transaction"
    to: "score"
    mode: "sequential"
  - from: "score"
    to: "auto_approve"
    mode: "conditional"
    when: "score.value < 30"
  - from: "score"
    to: "auto_block"
    mode: "conditional"
    when: "score.value > 90"
  - from: "score"
    to: "human_review"
    mode: "conditional"
    when: "score.value >= 30 && score.value <= 90"

←

compliance-audit

AI Patient Risk Alert

→