AI Agent
What The AI Will And Won't Do
Guardrails for commitments, uncertainty, and human escalation.
Ticket0 is optimized for safe support automation. It can answer routine questions, but it should escalate when risk or uncertainty is high.
What the AI will do
- Answer routine product and policy questions using retrieved context
- Match your team's configured tone and instructions
- Ask for escalation when it cannot answer confidently
- Pass a handoff summary to operators when escalation happens
What the AI won't do
- Promise discounts, refunds, credits, or compensation on its own
- Make legal or compliance commitments
- Invent policies, prices, order status, or account actions
- Pretend confidence when context is missing
Uncertainty behavior
When confidence is low, the AI should not guess. It escalates with a reason code (for example ai_uncertain, financial_dispute, or legal_compliance) so the ticket is routed with the right priority and tags.
Operator control
Workspace instructions can tighten policy further, for example:
- "Never offer compensation without manager approval"
- "Escalate all billing disputes"
- "Use this exact escalation phrase"
Continue with Confidence thresholds.