保險理賠詐欺偵測

AI → Human

AI 將收到的理賠申請與已知詐欺指標進行模式比對。

5 個節點 · 5 條連接insurance

eventagenthumansystem

視覺化

理賠申請接收串流event

透過門戶、業務員或理賠員提交的新申請案件。

↓sequential→ AI 詐欺模式比對

AI 詐欺模式比對agent

將理賠資料與歷史詐欺模式及紅旗指標進行交叉比對。

↓sequential→ 詐欺機率評分

詐欺機率評分system

依模式比對信心度計算 0-100 分。

↓conditional→ 理賠處置決定

↓conditional→ 理賠調查員審查

理賠調查員審查human

特殊調查小組審查被標記的案件並要求補件。

↓sequential→ 理賠處置決定

理賠處置決定system

核准、拒絕或移送執法機關。

在編輯器開啟

uc-claim-fraud-flag.osop.yaml

osop_version: "1.0"
id: "claim-fraud-flag"
name:"保險理賠詐欺偵測"
description:"AI 將收到的理賠申請與已知詐欺指標進行模式比對。"

nodes:
  - id: "claims_stream"
    type: "event"
    name: "理賠申請接收串流"
    description: "透過門戶、業務員或理賠員提交的新申請案件。"

  - id: "pattern_match"
    type: "agent"
    subtype: "llm"
    name: "AI 詐欺模式比對"
    description: "將理賠資料與歷史詐欺模式及紅旗指標進行交叉比對。"
    security:
      risk_level: "medium"

  - id: "fraud_score"
    type: "system"
    name: "詐欺機率評分"
    description: "依模式比對信心度計算 0-100 分。"

  - id: "investigator_review"
    type: "human"
    subtype: "review"
    name: "理賠調查員審查"
    description: "特殊調查小組審查被標記的案件並要求補件。"
    security:
      approval_gate: true

  - id: "disposition"
    type: "system"
    name: "理賠處置決定"
    description: "核准、拒絕或移送執法機關。"

edges:
  - from: "claims_stream"
    to: "pattern_match"
    mode: "sequential"
  - from: "pattern_match"
    to: "fraud_score"
    mode: "sequential"
  - from: "fraud_score"
    to: "disposition"
    mode: "conditional"
    when: "fraud_score.value < 20"
    label: "Auto-approve low-risk claims"
  - from: "fraud_score"
    to: "investigator_review"
    mode: "conditional"
    when: "fraud_score.value >= 20"
  - from: "investigator_review"
    to: "disposition"
    mode: "sequential"