Skip to main content

Chat Agent

The Chat Agent is an AI-powered assistant that handles text-based conversations with your website visitors. When someone opens the chat widget and starts typing, the Chat Agent responds in real time — answering questions, providing information, and guiding visitors based on what you have trained it to know.

What the Chat Agent Does​

The Chat Agent takes over the live chat channel and interacts with visitors on your behalf. It can:

  • Answer common questions about your business, products, and services
  • Provide helpful information pulled from your Knowledge Base
  • Greet visitors with a custom welcome message
  • Guide visitors toward next steps (like booking a call or filling out a form)
  • Hand off to a human team member when the conversation requires it

The Chat Agent works around the clock, so visitors always get a response — even outside your working hours.

Configuration​

To set up your Chat Agent, go to Settings > AI Agents > Chat Agent. You will find four configuration tabs: AI Engine, Behavior, Chat Interface, and Routing.

AI Engine​

The AI Engine tab controls the underlying model and response style your Chat Agent uses.

  • Model selection — Choose the AI model that powers your Chat Agent. Different models offer different balances of speed, accuracy, and capability.
  • Response style — Control how your Chat Agent formulates responses. You can adjust the creativity and precision of the AI's answers to match your preferences.
tip

If you are unsure which model to choose, start with the default. You can always switch later and compare how the Chat Agent performs with different models.

Behavior​

The Behavior tab is where you shape the Chat Agent's personality and give it instructions.

  • Personality — Define how the Chat Agent comes across in conversation. You can make it friendly and casual, professional and formal, or anything in between.
  • Instructions — Give the Chat Agent specific directions about how to handle conversations. For example, you can instruct it to always ask for the visitor's name, to recommend specific products, or to avoid discussing competitor pricing.
  • Tone — Fine-tune the overall tone of the Chat Agent's responses. This works alongside the personality setting to create a consistent voice.
note

The more specific your instructions are, the better your Chat Agent will perform. Vague instructions lead to generic responses. Be clear about what you want the AI to say, how you want it to say it, and what topics it should focus on or avoid.

Chat Interface​

The Chat Interface tab controls how the Chat Agent's chat window looks and feels within the widget.

  • Widget appearance — Customize visual elements of the chat window, including colors and layout, to match your brand.
  • Greeting message — Set the first message visitors see when they open the chat. This message appears before the visitor types anything, so it sets the tone for the entire conversation.

Good greeting messages are welcoming and action-oriented. For example:

  • "Hi there! How can I help you today?"
  • "Welcome! I can answer questions about our products, pricing, or services."
  • "Hey! Looking for something specific? Just ask."

Routing​

The Routing tab controls when and how the Chat Agent escalates conversations to a human team member.

  • Escalation triggers — Define the conditions under which the Chat Agent should stop handling a conversation and pass it to a human. This might include when a visitor explicitly asks to speak to a person, when the AI cannot find a good answer, or when the conversation involves sensitive topics.
  • Routing rules — Specify which team members or teams should receive escalated conversations.
warning

Without routing rules, the Chat Agent will try to handle every conversation on its own. Make sure you configure escalation triggers so that complex or sensitive conversations reach your human team members when needed.

Testing Your Chat Agent​

Before going live, you should test your Chat Agent to make sure it responds the way you expect:

  1. After configuring your settings, use the preview or test option in your Chat Agent settings
  2. Type in sample questions that your visitors are likely to ask
  3. Check that the responses are accurate, helpful, and match the tone you configured
  4. Try edge cases — ask something the AI might not know and confirm it escalates or handles it gracefully
  5. Adjust your Behavior instructions or Knowledge Base content based on what you find
tip

Test your Chat Agent regularly, especially after updating your Knowledge Base. New content may change how the AI responds to certain questions.

Best Practices​

  • Train your Knowledge Base first — The Chat Agent is only as good as the information it has. Before enabling it, make sure your Knowledge Base is comprehensive and up to date.
  • Set clear instructions — Tell the Chat Agent exactly what you want it to do and what to avoid. Think of the instructions as a briefing document for a new team member.
  • Review conversations — Periodically check the chat history in your dashboard to see how the Chat Agent is performing. Look for conversations where it struggled and update your Knowledge Base or instructions accordingly.
  • Keep the greeting simple — A short, friendly greeting message works better than a long paragraph. Visitors want to ask their question, not read an essay.

Next Steps​

If you want to offer AI-powered voice conversations as well, set up the Call Agent.