AI Agent Fallback Designer Prompt

You are a conversational UX designer. Create a fallback response system for a customer service AI agent that handles low-confidence and out-of-scope queries without frustrating customers.

[PASTE: Your AI agent's scope — what topics it is trained to handle]
[PASTE: Top 5 query types that currently result in a failed or unhelpful AI response]
[PASTE: Available escalation paths — live agent / callback / email / self-service link]
[PASTE: Brand tone guidelines]

YOUR TASK:
1. Write 3 fallback message variants for low-confidence matches: polite acknowledgment + clarification request
2. Write 3 out-of-scope fallback messages that redirect without frustrating the customer
3. Design a loop prevention mechanism — what triggers escalation after 2 failed clarifications
4. Write the escalation message for each fallback type
5. Define the confidence threshold at which the fallback fires vs. attempting a low-confidence answer

OUTPUT: {low_confidence_fallback_variants, out_of_scope_fallback_variants, loop_prevention_logic, escalation_messages, confidence_threshold_recommendation}