Back to Scenarios

Support — Refund Outside Threshold

Preview Only
Support
Customer Support
Global
No HITL
Tool Call
Policy
Autonomous

What Happened

A customer support AI agent processed a refund request that exceeded the standard policy threshold. The action was executed via a tool call without human intervention. The timeline captures the tool invocation, the policy reference, and the outcome, highlighting the absence of human oversight in the decision path.

What This Demonstrates

  • Autonomous AI action without human-in-the-loop
  • Tool call execution with evidence hash
  • Policy threshold reference in decision context
  • Explicit marking of "no human oversight" in reconstruction

What This Does NOT Claim

  • Does not determine if the refund was appropriate
  • Does not assess policy compliance or violation
  • Does not recommend changes to threshold or process

This scenario displays static preview data. Only the FinTech scenario can run against the backend.

Jan 12, 2026, 16:45:10
KNOWN

Customer Support AI

Customer refund request received and evaluated. Amount: $450. Standard threshold: $300.

sha256:f2c8d...f2a5
trace-sup-001
Jan 12, 2026, 16:45:11
KNOWN

Customer Support AI

Policy reference: Premium tier customers eligible for refunds up to $500 without approval.

sha256:a3d9e...d6f8
Jan 12, 2026, 16:45:12
KNOWN

Customer Support AI

Tool call executed: process_refund(amount=450, customer_id=C-92847). No human oversight.

sha256:b4e0f...a7c9
trace-sup-002
Jan 12, 2026, 16:45:13
KNOWN

Payment System

Refund processed successfully. Transaction ID: TXN-45829.

sha256:c5f1a...a8b0