Skip to main content
The HoopAI platform supports AI-powered chat capabilities through the chat widget, allowing you to automate visitor interactions, qualify leads, and provide instant responses without requiring a live agent. AI chat operates continuously — 24 hours a day, 7 days a week — so no visitor goes unanswered, even outside business hours.

​
AI capabilities in the chat widget

There are two ways to add AI to your chat widget experience:

​
1. Voice AI widget type

The Voice AI widget type connects visitors to a Voice AI Agent configured in your account. When a visitor clicks the widget, a voice call session is initiated directly in the browser. The Voice AI Agent handles the conversation using natural language, responding to questions, collecting information, and routing based on the visitor’s intent. Voice AI is suited for:
  • Businesses that receive high volumes of repetitive inbound inquiries (pricing, hours, services)
  • Teams too small to staff live chat but that want to provide immediate responses
  • Pages where spoken communication is more natural than typing (service businesses, professional services, healthcare)
To use the Voice AI widget type, you must first configure a Voice AI Agent in the AI Agents section of your account.

​
2. AI-assisted live chat

For text-based chat, the conversational AI assistant can suggest responses to agents in real time. As an agent reads an incoming message, the AI generates a suggested reply based on your account’s knowledge base and conversation history. The agent can:
  • Send the suggested reply as-is with one click
  • Edit the suggestion before sending
  • Ignore the suggestion and type their own reply
This approach keeps a human in the loop for quality control while significantly speeding up response times for routine questions.

​
Setting up the Voice AI widget

1

Create a Voice AI Agent

Navigate to AI Agents in the left sidebar. Click + New Agent and select Voice AI. Configure the agent’s persona, greeting, knowledge base, and call handling logic. Save the agent.
2

Create a Voice AI chat widget

Go to Sites > Chat Widget and click + New. Select Voice AI as the widget type.
3

Configure the widget appearance

Set the widget’s chat prompt, icon, color, position, and welcome message in the Style tab. The welcome message should set the expectation that clicking will initiate a voice call (e.g., “Click to speak with our AI assistant.”).
4

Link the Voice AI Agent

In the widget settings, select the Voice AI Agent you configured. All calls initiated through this widget will be handled by that agent.
5

Install and test

Save the widget and install it on your page using the embed code or the funnel/website settings dropdown. Test the widget by clicking it and completing a sample conversation. Verify the agent responds correctly and that the conversation is logged to your Conversations inbox.

​
AI escalation to a live agent

Voice AI and AI-assisted chat can be configured to escalate to a live agent when:
  • The visitor explicitly requests to speak with a person
  • The AI reaches a question it cannot confidently answer
  • The conversation meets a qualifying threshold (e.g., visitor expresses intent to purchase)
When escalation is triggered, the conversation is transferred to the live agent queue and an internal notification is sent to available agents. The visitor is informed of the transition with a message such as “Let me connect you with a team member who can help.” Configure escalation rules within your Voice AI Agent settings under the Call Handling or Escalation section.

​
Training the AI on your content

AI agents use your account’s knowledge base to answer questions accurately. To improve response quality:
  • Add FAQs — document common visitor questions and their correct answers in the AI Agent’s knowledge base
  • Link product and service pages — provide the agent with URLs to key pages so it can reference accurate, up-to-date information
  • Define off-limits topics — specify subjects the agent should decline to answer and route to a human instead
  • Review conversation transcripts — regularly check AI conversation logs for incorrect or incomplete answers and update the knowledge base accordingly

​
Tracking AI chat performance

Conversations handled by AI agents appear in the Conversations inbox labeled with the agent name. Review transcripts to evaluate:
  • Deflection rate — percentage of conversations fully resolved by AI without requiring a live agent
  • Escalation rate — percentage of conversations escalated to a live agent
  • Lead capture rate — percentage of conversations where the visitor provided contact information
  • Common question themes — what visitors are asking most frequently, which informs knowledge base updates and site content gaps
AI chat responses are only as accurate as the knowledge base and configuration provided. Review the AI agent’s responses regularly, especially after product changes, pricing updates, or policy revisions. Outdated or incorrect AI responses can erode visitor trust quickly.
Start with a narrow scope for your AI agent — train it to answer only the 10–15 most common questions and escalate everything else. As you review transcripts and identify patterns, expand the scope gradually. This approach produces higher-quality responses than trying to configure an all-knowing agent from the start.
Last modified on March 5, 2026