Customer Support Chatbot Design Template

Use buttons for the initial intent selection (first 1-2 steps) and free text for detail collection (email addresses, descriptions). Buttons reduce friction and prevent misrouted conversations. Free text is necessary when the bot needs specific information from the user. Mixing both is the right approach for most support bots.

Start with 5-8 flows covering your highest-volume ticket categories. These should handle 60-80% of inbound requests. Adding more flows before validating the first set creates maintenance burden without proven value. Expand based on fallback data showing which unhandled topics are most requested.

Detect negative sentiment (keywords like "terrible," "cancel," "angry," "ridiculous") and route to a human agent immediately with the full conversation context. Do not attempt to resolve emotional escalations with automated responses. The bot should acknowledge the frustration briefly and hand off quickly. See the [customer escalation template](/templates/customer-escalation-template) for the agent-side process.

Industry benchmarks range from 25-40% for general support bots and 50-70% for bots focused on narrow, well-defined tasks (password resets, order tracking). If your containment rate is below 20%, your flows are either too broad or your intent recognition is poor. Focus on improving the top 3 failing flows before adding new ones.

Review flows monthly for the first 3 months after launch, then quarterly. Prioritize updates when a flow's fallback rate exceeds 30% or when you ship a product change that affects a covered topic. Every product release should trigger a review of whether existing flows need updating. ---