What AI tools work with this customer-support prompt?

This prompt works with chatgpt, claude, gemini. Simply copy and paste the template into any of these AI assistants.

Tips for getting the best results?

Include real-world variations like typos, slang, and incomplete sentences to improve model robustness Create edge cases where intents might overlap to test your classifier's decision boundaries Balance the dataset across all intent categories to prevent model bias toward frequent intents Add context-dependent examples where the same phrase could have different intents Include negative examples and out-of-scope queries to train proper rejection handling

customer-support advanced

Create Chatbot Intent Classification Training Examples

Generate comprehensive intent classification examples to train your chatbot. Create labeled datasets for better customer support automation.

Works with: chatgptclaudegemini

Prompt Template

You are an expert in natural language processing and chatbot training. Create a comprehensive set of intent classification examples for a customer support chatbot in the [INDUSTRY] industry. Generate [NUMBER_OF_EXAMPLES] diverse training examples across the following intent categories: [INTENT_CATEGORIES] For each example, provide: 1. User utterance (the actual customer message) 2. Intent label 3. Confidence reasoning (why this utterance maps to this intent) 4. Key entities or parameters extracted (if applicable) Requirements: - Include variations in phrasing, formality levels, and sentence structures - Cover edge cases and ambiguous scenarios - Include typos, abbreviations, and colloquial language that real customers use - Ensure balanced distribution across all intent categories - Add examples with multiple potential intents to test classification robustness - Include both positive and negative sentiment variations for each intent Format each example as: **Example [NUMBER]:** User Input: "[actual customer message]" Intent: [intent_label] Confidence Reasoning: [why this classification makes sense] Extracted Entities: [relevant parameters, if any] Additional Context: [BUSINESS_CONTEXT] Target Audience: [CUSTOMER_DEMOGRAPHICS] Special Considerations: [SPECIAL_REQUIREMENTS]

Variables to Customize

[INDUSTRY]

The industry or business sector for the chatbot

Example: e-commerce retail

[NUMBER_OF_EXAMPLES]

How many training examples to generate

Example: 25

[INTENT_CATEGORIES]

List of intent categories to classify

Example: order_status, return_request, product_inquiry, technical_support, billing_question, shipping_info, account_management

[BUSINESS_CONTEXT]

Specific business context or constraints

Example: Online clothing store with 30-day return policy and international shipping

[CUSTOMER_DEMOGRAPHICS]

Primary customer base characteristics

Example: Millennials and Gen Z, tech-savvy, value-conscious shoppers

[SPECIAL_REQUIREMENTS]

Any special considerations or requirements

Example: Must handle multiple languages and seasonal product variations

Example Output

**Example 1:** User Input: "hey where's my order? been waiting forever 😤" Intent: order_status Confidence Reasoning: Customer is inquiring about order delivery status with frustrated tone Extracted Entities: emotion=frustrated, urgency=high **Example 2:** User Input: "Do you have this sweater in medium? The blue one from the homepage" Intent: product_inquiry Confidence Reasoning: Customer asking about product availability with specific attributes Extracted Entities: size=medium, color=blue, product_type=sweater **Example 3:** User Input: "I want to return these jeans but I threw away the receipt" Intent: return_request Confidence Reasoning: Clear return intent with complicating factor (missing receipt) Extracted Entities: product_type=jeans, issue=missing_receipt **Example 4:** User Input: "my card got charged twice last month wtf" Intent: billing_question Confidence Reasoning: Customer reporting billing discrepancy with emotional language Extracted Entities: issue=duplicate_charge, timeframe=last_month, emotion=angry

Pro Tips for Best Results

Include real-world variations like typos, slang, and incomplete sentences to improve model robustness
Create edge cases where intents might overlap to test your classifier's decision boundaries
Balance the dataset across all intent categories to prevent model bias toward frequent intents
Add context-dependent examples where the same phrase could have different intents
Include negative examples and out-of-scope queries to train proper rejection handling