This document highlights representative structured decisions from the repository so reviewers can understand the system behavior quickly.
Input:
Customer message: Where is my order?
Decision:
{
"override_required": false,
"reason_codes": ["NONE"],
"confidence": 0.90,
"recommended_next_step": "automated_order_lookup"
}Interpretation:
- routine operational request
- no human review required
- automated path remains active
Input:
Customer message: I want a refund and I will contact a lawyer if this is not resolved.
Decision:
{
"override_required": true,
"reason_codes": ["LEGAL_RISK"],
"confidence": 0.95,
"recommended_next_step": "escalate_to_legal_team"
}Interpretation:
- explicit legal-risk signal
- automated handling should stop
- case is routed to a specialized human path
Input:
Customer message: My invoice looks strange.
Decision:
{
"override_required": true,
"reason_codes": ["HIGH_AMBIGUITY"],
"confidence": 0.80,
"recommended_next_step": "request_clarification_and_human_review"
}Interpretation:
- not obviously legal, safety, or privacy-related
- still unsafe to auto-resolve because intent is unclear
- demonstrates that the gate is not limited to explicit risk keywords