How to Create and Train an AI Agent?

An AI agent answers visitor questions automatically using the knowledge you train it on and the behavior rules you set. You create and manage agents from the AI Agents tab in ProProfs Live Chat

Watch this video walkthrough to see it in action.

Or continue reading for step-by-step instructions.

What Is an AI Agent?

An AI Agent is an automated assistant in ProProfs Live Chat that responds to visitor questions on its own. Each agent is built in three sections: Knowledge defines what it learns from — websites, files, text notes, Q&A pairs, a ProProfs Knowledge Base, or past chat transcripts; Behavior shapes its persona, instructions, guardrails, actions, and the visitor details it collects; and Preview lets you test it before going live.

Why Use It?

An AI agent takes repetitive questions off your team’s plate and keeps visitors answered at all hours.

• Your chat queue keeps growing after business hours — the AI agent answers visitors instantly, around the clock, and can transfer to an operator when needed.
• Your answers already exist across your website, help docs, and files — train the agent on those sources once and it handles the questions they cover.
• Routine tasks like ticket creation and lead capture eat into chat time — the agent’s actions can create tickets, book meetings, and add leads to your CRM on its own.

Example Use Cases

• A support agent lets the AI agent resolve pricing and how-to questions, with Transfer to Operator switched on for escalations.
• A team manager trains an agent on the company website and knowledge base so product questions get instant, sourced answers.
• A small business owner runs an agent overnight that collects each visitor’s name and email and books meetings into the team calendar.

Step-by-Step Instructions

Step 1: Log in to ProProfs Live Chat, click AI Agents in the top navigation, and click + Create New Agent.

Step 2: In the Knowledge section, pick a source tab — Website URL, Upload file, Text, Q&A, ProProfs KB, or Transcripts — then enter your website address under Website URL and click Add & train.

Step 3: When the source status turns Ready, click it to review or edit the learned query–response pairs.

Step 4: Open the Behavior section and, in Basics, write a first-person Persona or pick a template, then choose a Tone and an Answer length.

Step 5: In Instructions and Guardrails, toggle the rules you want, click + Add instruction or + Add guardrail to write your own, or use Generate from knowledge to draft rules from your trained sources.

Step 6: In Actions, switch on what the agent can do on your behalf — Transfer to Operator, Create Support Ticket, Schedule a Meeting, Send Notification, Add Lead to CRM, Update Order Status, or Trigger External API.

Step 7: In Visitor Info, click the fields the agent should collect — Name, Email, Phone Number — or click + Add Field for a custom one.

Step 9: Open Preview, send a sample question, and review the reply along with the Agent Trace panel, which lists the rules the agent applied.