How-to Guides
Write Effective Chatbot Descriptions
Learn how to write clear, actionable chatbot descriptions that improve persona and scenario generation quality.
Your chatbot description is the foundation for all Snowglobe simulations. It directly impacts the quality of persona generation and scenario creation, making it one of the most important inputs for accurate testing.
Questions? Join our developer community or contact support for help crafting your chatbot description.
What is an Chatbot Description?
A chatbot description tells Snowglobe what your chatbot is, who it’s for, and what it does. This information grounds both persona generation and scenario creation, ensuring your simulations reflect realistic user interactions.This is not the place for historical data or specific conversation examples. Use the historical data upload feature in the web UI for that. Chatbot descriptions should focus on defining your chatbot’s purpose and capabilities.
What Makes a Good Description?
The best chatbot descriptions naturally answer these key questions. You don’t need to structure your description around these - they’re just thinking prompts to help you include the right information: What does your chatbot do? The core functionality and value it provides to users. Who is it for? The types of people who interact with it and their typical goals. What can users ask? The kinds of questions, problems, or tasks it handles. What are its boundaries? What it doesn’t do, when it escalates, or topics it avoids. Write naturally about your chatbot - don’t worry about following a specific format or template.Writing Guidelines
- Start simple: 2-3 sentences covering what your chatbot does, who uses it, and what they ask about is enough to get started.
- Be specific but not exhaustive: “Customer support for billing issues” is better than “helps users with stuff” but you don’t need to list every possible question type.
- Focus on user value, not technical details: Describe what users experience, not how your chatbot is built.
- Don’t include conversation examples: Use the historical data upload feature for specific examples.
- You can always add more later: Start basic and expand your description as you learn what works for your simulations.
Examples
Good Examples
Customer Support Assistant (Basic)
Customer Support Assistant (Basic)
“Helps customers troubleshoot technical issues with our software platform. Customers range from beginners to advanced users who are experiencing problems and need help. Can diagnose common issues like software bugs, login problems, and billing questions, and provide step-by-step solutions. Escalates complex cases to human support when needed.”
Educational Tutor (More Detail)
Educational Tutor (More Detail)
“This chatbot acts as a virtual tutor for high school and college students working on homework assignments, particularly in STEM subjects. It’s designed for students who want to learn rather than cheat - it guides them through problems step-by-step instead of providing direct answers.Students typically ask for help with math problems (algebra through calculus), physics concepts, chemistry equations, and study strategies. The chatbot uses the Socratic method, asking guiding questions to help students discover answers themselves. It’s patient and encouraging but refuses to solve homework problems outright.The chatbot won’t provide direct answers to homework questions, won’t help with non-academic topics, and escalates to human tutors for advanced graduate-level content.”
Sales Qualifier (Comprehensive)
Sales Qualifier (Comprehensive)
“This chatbot pre-qualifies sales leads by engaging B2B prospects in natural conversations. Our prospects are typically IT managers, operations directors, or business owners at companies with 50-500 employees who have shown initial interest in our enterprise software solution.The chatbot’s job is to understand their needs, budget, timeline, and decision-making process - it focuses on collecting qualification data rather than closing deals. Prospects usually ask about their current pain points and challenges, existing tools and workflows, budget and timeline questions, technical requirements, and who’s involved in making decisions.The conversation style is professional but conversational, consultative rather than pushy. The chatbot asks thoughtful discovery questions and listens actively. A typical flow starts with rapport building, moves to needs assessment, explores current solutions, identifies gaps, discusses timeline and budget, and schedules next steps.A qualified lead has confirmed need, budget authority, and timeline within 6 months.”
Poor Examples
Too Vague
Too Vague
Bad: “This is an AI assistant that helps users with various tasks and questions.”Why it’s bad: No specific purpose, audience, or capabilities defined. This will generate generic, unfocused personas and scenarios.Better: “Helps small business owners manage their inventory by tracking stock levels, generating reorder alerts, and providing sales analytics.”
Too Technical
Too Technical
Bad: “REST API endpoint utilizing GPT-4 with RAG architecture, vector embeddings from Pinecone, and LangChain orchestration for multi-modal document processing workflows.”Why it’s bad: Focuses on technical implementation rather than user value. Personas won’t understand what this actually does for them.Better: “Analyzes business documents to answer questions about contracts, policies, and procedures. Helps employees quickly find information without reading through lengthy documents.”
Includes Irrelevant Information
Includes Irrelevant Information
Bad: “Customer support chatbot built by our team last year using Python and deployed on AWS. Here are some example conversations: ‘User: My password isn’t working’ Chatbot: ‘Let me help you reset that…’”Why it’s bad: Implementation details and conversation examples don’t belong in chatbot descriptions. Use historical data upload for examples.Better: “Helps customers resolve account access issues, billing questions, and basic troubleshooting for our SaaS platform.”
Lists Features Instead of Purpose
Lists Features Instead of Purpose
Bad: “Features include: natural language processing, sentiment analysis, multi-language support, integration with Slack, email notifications, and dashboard reporting.”Why it’s bad: Lists technical features without explaining what the chatbot actually does for users.Better: “Monitors customer feedback across multiple channels, identifies urgent issues requiring immediate attention, and routes them to appropriate support teams.”
Resources
Join the Community
Connect with other developers, ask questions, and share feedback.
See Examples
See examples and showcase of various chatbots simulated with Snowglobe.
Connect Your Chatbot
Ready to connect? Follow our step-by-step chatbot connection guide.
Define Simulation Intent
Learn how to define what you want to test with your simulations.
Questions? Join our developer community or contact support for help crafting your chatbot description.