AI Conversation Flow Builder Prompt
Prompt
You are a conversational designer. Build a structured conversation flow for an AI agent handling a specific customer support scenario end-to-end.
[PASTE: The specific scenario — e.g., password reset, subscription cancellation, order status inquiry]
[PASTE: Data the AI agent can access — customer account info, order history, product catalog]
[PASTE: Business rules — what can be resolved without human approval vs. what requires escalation]
[PASTE: Tone guidelines — formal/informal, empathy level, brand voice]
YOUR TASK:
1. Map the flow from entry (customer opens the conversation) to resolution (confirmed solved or escalated)
2. Write all agent messages for each node in the flow
3. Define the decision branches at each step (customer response options)
4. Specify the data lookup or action required at each step
5. Write the escalation handoff message and agent context summary for the human takeover
OUTPUT: {flow_diagram_description, all_agent_messages, decision_branches, data_and_action_requirements, escalation_handoff}Why it works
Pre-writing all messages ensures tone consistency regardless of which branch the customer takes. Explicit action requirements expose integration dependencies before build.
Watch out for
Rigid trees break on unexpected customer responses. Build a universal fallback path and test with edge case inputs before go-live.
Used by
Customer Success ManagersIT & Ops Teams