AI Agent Overview

The AI Agent is your 24/7 support assistant that answers customer questions using your knowledge base content. It provides instant responses, only escalating to your human team when it can't help.

What the AI Agent Does

The AI Agent:

Answers questions using your published knowledge base articles and internal knowledge
Provides source citations so customers can verify information
Recognizes its limits and escalates when it can't help
Works around the clock providing instant responses
Learns from your content as you add and update articles
Improves over time by learning from resolved ticket conversations

Where the AI Agent Appears

Knowledge Base Chat

On your public knowledge base pages, customers see a chat interface. They can:

Ask questions about your product
Get instant answers from your documentation
Click through to full articles for more detail
Escalate to human support if needed

When configured for AI chat, your embeddable widget provides:

AI-powered conversation
Same knowledge base-powered answers
Seamless escalation to ticket creation
Embedded experience on your website

Email Auto-Response

When enabled on your email integration, the AI automatically responds to incoming support emails:

Instant automated responses to customer emails
Answers drawn from your knowledge base
Inline images from relevant articles
Feedback buttons let customers indicate if the answer helped
Automatic ticket resolution when customers confirm the answer worked

See Email Integration for setup details.

How It Works

RAG Architecture

The AI uses RAG (Retrieval-Augmented Generation):

Customer asks a question
System searches your knowledge base for relevant content
AI generates a response based on found content
AI selects relevant images and videos to include
Response includes citations to source articles

This means the AI only answers using your approved documentation—it doesn't make things up or use outside knowledge.

Intelligent Media Selection

When your knowledge base articles contain images or videos (YouTube, Loom), the AI analyzes them to decide which media will help answer the customer's question. The AI:

Reviews all media in relevant articles
Reads captions to understand what each image or video shows
Selects only media that directly illustrates the answer
May choose to show no media if none is relevant

Example: If a customer asks "How do I reset my password?" and your article has screenshots of the password reset flow plus an unrelated product tour video, the AI will include the screenshots but skip the video.

To help the AI make better selections, add descriptive captions to your media. See Creating Articles for details on adding captions.

The AI's Knowledge

The AI knows:

✓ Everything in your published knowledge base articles
✓ Content from your internal knowledge sources (documents, URLs, text)
✓ Insights from resolved tickets you've exported to internal knowledge
✓ How to cite sources
✓ When to escalate

The AI doesn't know:

✗ Customer-specific account details
✗ Information not in your documentation or internal knowledge
✗ Real-time system status

AI Agent vs Traditional Chatbots

Feature	AI Agent	Traditional Chatbot
Response generation	Dynamic, natural language	Pre-scripted flows
Knowledge source	Your KB articles	Manual programming
Setup time	Automatic (from KB)	Hours of configuration
Maintenance	Update articles	Reprogram flows
Conversation style	Natural	Structured menus

Availability by Plan

Plan	KB Chat	Widget AI Chat
Free	✗	✗
Starter	✓	✓
Standard	✓	✓
Pro	✓	✓
Enterprise	✓	✓

AI features require AI credits. View your credit usage in Settings > Billing.

Getting Started with AI

Step 1: Publish Knowledge Base Content

The AI needs content to work with. Start by publishing articles covering:

Frequently asked questions
How-to guides
Product documentation
Troubleshooting steps

Tip: The more comprehensive your KB, the better the AI performs.

Step 2: Enable AI Chat

AI chat is enabled by default on paid plans. Check your settings:

Go to Knowledge Base > Settings
Ensure AI Chat is enabled
Configure any preferences

Step 3: Test the AI

Before launching to customers:

Visit your knowledge base
Ask questions the AI should be able to answer
Ask questions it shouldn't be able to answer (verify escalation works)
Check that citations are accurate

Step 4: Monitor and Improve

After launch:

Monitor escalation reasons
Add content for common unanswered questions
Update articles that lead to incorrect answers
Export resolved ticket conversations to internal knowledge to improve future responses

Step 5: Build Internal Knowledge

Enhance the AI beyond your public knowledge base:

Go to Knowledge Base > Internal Knowledge
Add sources: upload documents, crawl URLs, or add text directly
When resolving tickets, click Add to Internal Knowledge to capture successful resolutions
The AI will use this private knowledge to answer questions more effectively

Tip: Internal knowledge is great for edge cases, company policies, or technical details that don't belong in public documentation.

AI Behavior

When the AI Can Help

The AI will answer when:

Your knowledge base contains relevant information
The question is clearly stated
The answer can be sourced from documentation

When the AI Escalates

The AI will offer escalation when:

No relevant content exists in your KB
The question requires account-specific information
The customer explicitly asks for human help
The AI isn't confident in its answer

When the AI cannot answer a question, it offers to create a support ticket with the conversation context included.

Limitations