What are AI Agents?

LeadClose AI Agents are a team of trainable, automated virtual agents that work alongside your human team within your WhatsApp inbox. Unlike a single, generic chatbot, you can create multiple specialized agents, each with its own purpose and knowledge base.

This multi-agent architecture allows for more accurate and efficient handling of customer inquiries.

Key Components of the AI Agent System

1. Specialized Agents You can create a team of distinct AI agents, each with a specific role. For example:

• Support Bot: Trained on your FAQ documents to answer common customer questions.
• Sales Assistant: Trained on product specifications to handle initial sales inquiries.
• Policy Bot: Trained on your return and shipping policies.

For each agent, you define its name, a concise summary of its role, and its persona (e.g., "You are a friendly and helpful support agent").

2. RAG-Powered Knowledge Base Each agent is powered by a Retrieval-Augmented Generation (RAG) system.

• You provide knowledge by uploading simple text or markdown files.
• The AI agent will only use the information in these documents to answer questions.
• This "grounding" in your own data dramatically reduces the risk of the AI providing incorrect information or "hallucinating" answers.

3. The Intelligent Dispatcher When a new, unassigned conversation begins, it first goes to the Dispatcher, a master AI agent that acts as a router.

• The Dispatcher analyzes the customer's question.
• It compares the query to the summary of every available AI agent.
• It then routes the conversation to the best agent for the job.
• If no agent is a good fit, or if the user asks to speak to a person, the conversation is automatically routed to the general queue for a human agent.

4. Seamless Human Handoff AI Agents are designed to recognize their limits. If an agent cannot find a relevant answer in its knowledge base or if it detects user frustration, it will automatically un-assign itself from the conversation and leave an internal note explaining why, allowing a human agent to step in with full context.