AI Chatbot for Customer Support

Why Customer Support Is the Top Use Case for AI Chatbots

Most support teams spend the majority of their day answering the same 20 to 30 questions. What are your hours? How do I reset my password? What is your return policy? An AI chatbot trained on your knowledge base handles these instantly, with no queue and no wait time. Visitors get an answer in seconds instead of minutes or hours, and your human agents stop burning out on repetitive work.

The numbers make sense too. A chatbot response costs 5 to 15 credits (fractions of a cent) compared to the fully loaded cost of a human agent handling the same ticket. Even if the chatbot only resolves 60% of incoming questions on its own, that cuts your support workload by more than half.

What a Support Chatbot Can Handle

• FAQ answers: Product questions, pricing, policies, and how-to instructions drawn directly from your knowledge base
• Account lookups: Order status, appointment details, and billing questions when connected to your data
• Troubleshooting: Step-by-step guidance for common technical issues based on your support documentation
• Routing: Identifying the type of issue and directing the visitor to the right department or resource
• After-hours coverage: Answering questions at 2am the same way it would at 2pm, with no staffing cost

Setting Up a Support Chatbot

Create a chatbot in the AI Chatbot app and set its type to "support" (or "supportaudio" if you want voice features). The support type focuses the AI on accuracy and knowledge base retrieval rather than open-ended conversation. It is designed to find the right answer in your training data and deliver it clearly.

Load your support content into the knowledge base. This includes FAQ documents, help articles, product manuals, return policies, and anything else customers ask about. The more thorough your training data, the more questions the chatbot can handle without escalating. See how to train a chatbot on your documents for the step-by-step process.

Write a system prompt that sets the chatbot's tone and boundaries. Tell it your company name, what tone to use (friendly, professional, casual), and what topics it should not try to answer. A good support prompt also instructs the chatbot to say "I don't have information about that, let me connect you with our team" rather than guessing when it is unsure.

When to Hand Off to a Human Agent

No chatbot should pretend it can handle everything. Billing disputes, emotional complaints, and novel technical problems need a real person. The platform's chatbot to human handoff feature detects when a visitor is frustrated or asking for a person, and automatically transfers the conversation to your Live Operator inbox. The human agent sees the full conversation history so the visitor never has to repeat themselves.

Setting Up Handoff Rules

Enable the Live Operator toggle on your chatbot to create an inbox entry for every conversation. Your team can monitor conversations in real time and jump in whenever they see the chatbot struggling. The chatbot's built-in escalation logic also triggers automatically when visitors repeatedly request a human. See the handoff setup guide for complete instructions.

Measuring Support Chatbot Performance

Track two things: resolution rate and escalation rate. Resolution rate is the percentage of conversations where the visitor got their answer without needing a human. Escalation rate is how often the chatbot transfers to a live agent. A well-trained support chatbot should resolve 50% to 70% of conversations on its own within the first month, improving as you add more training data based on the questions it could not answer.

Review conversation logs weekly. Look for questions the chatbot answered incorrectly or could not answer at all. Add that information to your knowledge base and the chatbot immediately improves. This feedback loop is the most important part of running a support chatbot. See the chatbot analytics guide for more on measuring performance.

Getting Started

The fastest path to a working support chatbot: create the chatbot, upload your top 20 FAQ answers as training documents, write a simple system prompt, and embed it on your contact or help page. You can have a functional support chatbot running in under 30 minutes. Refine from there based on real conversations.