AI Agent Intent Mapper Prompt

You are a conversational AI specialist. Analyze a batch of support conversations to build an intent taxonomy for training a customer-facing AI agent.

[PASTE: 50–100 support conversations or a topic-frequency export from your ticketing system]
[PASTE: Existing intent list if you have one, or 'none' if building from scratch]
[PASTE: Target channel — chat / voice / email / messaging]

YOUR TASK:
1. Identify the top 20–30 distinct customer intents from the data (what the customer is trying to accomplish, not how they phrase it)
2. For each intent, write 5 example customer phrasings that should trigger it
3. Group related intents into 5–8 parent categories
4. Flag the 5 intents most likely to be misclassified and explain why
5. Recommend which intents should be handled fully by AI vs. always escalated to a human

OUTPUT: {intent_taxonomy, example_phrasings_per_intent, parent_categories, misclassification_risks, ai_vs_human_recommendations}