Skip to main content

What is an Agent?

An agent is an AI-powered conversational assistant that you deploy across your channels to interact with visitors and customers. Each agent is trained on your content and can answer questions, provide support, and engage users in natural conversations.
Think of an agent as a knowledgeable team member who’s always available to help, powered by AI and trained on your specific content.

Agent Types

Humanizing Agents supports two agent types, each designed for a different deployment channel:
TypeDescriptionDeployment
ChatbotA chat widget embedded on your website. Visitors click the chat bubble to ask questions, get support, and interact with your brand.Embed script on any website
ReceptionistA kiosk-based reception terminal for offices, hotels, and lobbies. Greets visitors, searches contacts, and manages check-ins with a multi-screen guided flow.Public URL on a tablet or touchscreen
When creating a new agent, you choose the type in the first step. Each type has its own onboarding wizard, navigation, and feature set.

Chatbot

The Chatbot is the original agent type. It powers an embeddable chat widget with:
  • RAG-powered responses from your knowledge base (web pages, text, Q&A pairs)
  • 3D avatar with text-to-speech
  • Customizable themes, colors, and widget positioning
  • Deploy with one line of embed code
See the Chatbot Onboarding guide for setup instructions.

Receptionist Agent

The Receptionist is a new agent type designed for physical locations. It provides:
  • Multi-screen kiosk flow (language selection, visit type, contact search, calling, confirmation)
  • Employee contact directory with CSV import
  • Multilingual visitor welcome in 45+ languages
  • Portrait and landscape kiosk orientations
  • Deploy via a simple public URL (/a/[publicKey]) on any touchscreen
See the Receptionist Overview for full documentation.

Core Concepts

Knowledge Base

Every agent has a knowledge base - a collection of information sources that the agent uses to answer questions. Your knowledge base can include:
Source TypeDescriptionBest For
Web SourcesPages from your website imported via URL or sitemapProduct pages, documentation, blog posts
Text SourcesCustom written content you provideCompany policies, internal knowledge
Q&A SourcesSpecific question-answer pairsFAQs, common support queries
The more relevant content you add to your knowledge base, the more helpful your agent becomes.

Avatar

Each agent has a 3D avatar that represents it visually. Avatars:
  • Animate and speak responses using text-to-speech
  • Create a more engaging, human-like experience
  • Can be customized to match your brand personality

System Prompt

The system prompt defines your agent’s personality, behavior, and response style. You can customize:
  • The agent’s name and role
  • How it greets users
  • Its communication style (formal, friendly, technical)
  • Specific instructions for handling certain topics

Widget

The widget is the chat interface that appears on your website. It’s how visitors interact with your agent. The widget includes:
  • A chat bubble that opens the conversation
  • A message input area
  • The avatar display
  • Language and voice options for visitors

Agent Lifecycle

1

Create

Create a new agent through the onboarding wizard by entering your website domain. The platform will automatically discover and suggest content to import.
2

Train

Add knowledge sources to teach your agent about your business. Import web pages, add custom text, and create Q&A pairs for common questions.
3

Customize

Configure your agent’s appearance, behavior, and widget styling. Choose an avatar, set colors, and write the perfect system prompt.
4

Deploy

Generate an embed script and add it to your website. Your agent is now live and ready to help visitors.
5

Monitor

Track conversations, view analytics, and continuously improve your agent based on user interactions.

Key Features

RAG-Powered Responses

Agents use Retrieval-Augmented Generation (RAG) to find relevant content from your knowledge base and generate accurate, contextual responses.

Multi-Language Support

Serve visitors in their preferred language with built-in localization support for the widget interface.

Conversation History

Review past conversations to understand what visitors are asking and how your agent is responding.

Analytics Dashboard

Track usage metrics, conversation volumes, and identify popular topics to optimize your agent.

Agent vs Workspace

Understanding the relationship between agents and workspaces:
ConceptDescription
WorkspaceA container for organizing agents and managing team access. Each workspace can have multiple agents and members.
AgentAn individual AI assistant within a workspace. Each agent has its own knowledge base, settings, and deployment.
You can create multiple agents within a single workspace - for example, one agent for sales inquiries and another for technical support, each trained on different content.

Best Practices

Focus on adding accurate, helpful content to your knowledge base. Well-written source material leads to better agent responses.
Use the Playground to test your agent with real questions before deploying to your website. Refine your system prompt based on responses.
Review conversations regularly to identify gaps in your knowledge base. Add new content to address common questions your agent couldn’t answer.
Write clear, specific system prompts that define exactly how your agent should behave. Avoid overly broad instructions.

Next Steps