Agents

Agents vs. AI Employees: An AI Employee is the assistant you configure and chat with inside the app. An Agent is a public-facing chatbot you embed on your website — each agent is powered by one of your AI employees but has its own system prompt, appearance, and lead-capture settings.

​

What Are Agents?

• Answer visitor questions automatically
• Use your knowledge base to give accurate responses
• Capture leads and contact information
• Provide instant support without human intervention
• Work around the clock

​

Channels

• General — name, system prompt, knowledge base access, contact capture, and response delay
• Integrations (Beta) — let the agent use connected tools while it talks
• Handoff — hand a conversation to a human when the agent can’t help
• Notifications — who gets told about new conversations and captured leads

​

Voice agents

​

Conversations

​

Creating Your First Agent

​

Step 1: Open Agents

1. Select your company from the sidebar
2. Click Agents (under Inbound)
3. Click Create Agent (or the + button)

​

Step 2: Basic Settings

• Name: What you’ll call this agent internally
• Avatar: Upload an image for your agent
• Employee: Select which AI employee powers this agent

​

Step 3: Choose a System Prompt Template

• Answers questions based on your training data
• Stays focused on helping customers
• Professional and helpful tone

• Qualifies prospects with questions
• Collects contact information
• Guides visitors toward next steps

• Write your own instructions from scratch
• Full control over behavior

​

Step 4: Configure Responses

• Fallback Response: What to say when the agent doesn’t know the answer
• Welcome Message: The first message visitors see

Hi! I'm here to help answer your questions.
What can I assist you with today?

​

Step 5: Save and Test

1. Click Save or Create
2. Use the Preview to test your agent
3. Refine settings as needed

​

Adding the Agent to Your Website

​

Getting the Embed Code

1. Go to your agent in the list
2. Click the three-dot menu (⋮)
3. Select Get Embed Code
4. Copy the code snippet

​

Installing on Your Website

<script>
  window.agentConfig = &#123;
    agentId: 'your-agent-id',
    position: 'bottom-right',
    theme: 'light'
  &#125;;
</script>
<script src="https://yourplatform.com/widget.js"></script>

​

Platform-Specific Instructions

1. Go to Appearance → Theme Editor
2. Edit footer.php
3. Paste the code before </body>

1. Go to Online Store → Themes
2. Click Actions → Edit code
3. Find theme.liquid
4. Paste before </body>

1. Go to Settings → Custom Code
2. Add new custom code
3. Paste the embed code
4. Set to load on all pages

1. Go to Settings → Advanced → Code Injection
2. Paste in the Footer section
3. Save

​

Customizing Your Agent’s Appearance

​

Widget Options

​

Updating Appearance

1. Edit your agent
2. Find the appearance settings
3. Customize colors, position, and messages
4. Save changes
5. Get new embed code if needed

​

Connecting to Your Knowledge Base

​

How It Works

1. A visitor asks a question
2. The AI searches your knowledge base
3. Finds relevant information
4. Generates an accurate response based on your content

​

Setting Up Knowledge Access

1. Make sure your AI employee has access to the knowledge base
2. Upload relevant documents to your knowledge base
3. The agent automatically uses this information

​

Tips for Better Responses

• Add FAQ documents to your knowledge base
• Include product information and pricing
• Upload support articles and guides
• Keep information up to date

​

Lead Capture

​

Setting Up Lead Capture

1. Edit your agent
2. Find Lead Capture settings
3. Enable lead capture
4. Configure required fields

​

Fields You Can Collect

• Name
• Email address
• Phone number
• Company name
• Custom fields

​

When to Ask for Information

• At the start of the conversation
• After answering initial questions
• When visitor asks for a callback
• Before providing detailed information

​

What Happens to Leads

• Appear in your conversations list
• Trigger workflows
• Sync to your CRM via integrations
• Generate notifications

​

Viewing Agent Conversations

​

Accessing Conversations

1. Go to Agents
2. Click the Conversations tab
3. Browse all visitor chats

​

What You Can See

• Full conversation history
• Visitor information (if collected)
• When conversations happened
• How the agent responded

​

Using Conversations to Improve

• Find questions the agent couldn’t answer
• Add missing information to your knowledge base
• Update your agent’s instructions
• Identify common visitor needs

​

System Prompt Best Practices

​

What to Include

You are a helpful customer support assistant for [Company Name].
Your job is to answer questions about our products and services.

Be friendly and professional. Use simple language.
If you don't know something, say so honestly.

Only answer questions about our company and products.
Don't discuss competitors or unrelated topics.
Never share confidential information.

If you can't answer a question, offer to connect the
visitor with a human representative.

​

Example Support Agent Prompt

### Role
You are a customer support agent for [Company Name].
Your job is to help visitors with questions about
our products and services.

### Guidelines
- Be friendly, helpful, and professional
- Answer questions based on the knowledge provided
- Keep responses concise and clear
- If you don't know something, say so

### Boundaries
- Only discuss our company and products
- Don't make promises about pricing or availability
- Offer to connect with humans for complex issues

### Collecting Information
When appropriate, ask for the visitor's name and
email so we can follow up.

​

Testing Your Agent

​

Preview Mode

1. Go to your agent settings
2. Click Preview or Test
3. Chat with your agent as a visitor would

​

What to Test

• Does the welcome message appear correctly?
• Are responses accurate and helpful?
• Does it handle unknown questions gracefully?
• Is lead capture working?
• Does the design look right?

​

Testing on Your Website

1. Visit your website in a new browser/incognito window
2. Click the chat button
3. Have a conversation
4. Check that everything works as expected

​

Managing Multiple Agents

​

Use Cases for Multiple Agents

• Support Agent: Answer product questions
• Sales Agent: Qualify leads and book demos
• FAQ Agent: Handle common questions
• Different Languages: Separate agents for each language
• Different Products: Specialized agents per product line

​

Managing Your Agents

1. Go to Agents
2. See all your agents in the list
3. Each shows:
   • Name and avatar
   • Linked employee
   • Status (active/inactive)
   • Conversation count

​

Duplicating an Agent

1. Find the agent to copy
2. Click the three-dot menu (⋮)
3. Select Duplicate
4. Edit the copy as needed

​

Troubleshooting

​

”Agent not appearing on website”

• Verify the embed code is installed correctly
• Check for JavaScript errors in browser console (F12)
• Make sure the code is before </body>
• Clear your website’s cache
• Check if ad blockers are interfering

​

”Agent looks wrong”

• Check appearance settings in your agent config
• Look for CSS conflicts with your website
• Try adjusting z-index if elements overlap
• Verify the embed code is current

​

”Responses are incorrect”

• Review and update your knowledge base
• Adjust the AI employee’s instructions
• Add more relevant documents
• Refine the system prompt

​

”Lead capture not working”

• Check lead capture settings are enabled
• Verify required fields are configured
• Test in preview mode first
• Check for form validation errors

​

”Agent is slow to respond”

• This is normal - AI takes a moment to generate responses
• Longer responses take more time
• Complex knowledge base searches may be slower
• Response times vary with server load

​

Best Practices

​

1. Write a Clear Welcome Message

• Be friendly and brief
• Tell visitors what the agent can help with
• Set expectations

​

2. Keep Knowledge Base Updated

• Add new information regularly
• Remove outdated content
• Include FAQs and common questions
• Update when products or pricing change

​

3. Monitor and Improve

• Note questions the agent couldn’t answer
• Find patterns in what visitors ask
• Update based on real visitor needs
• Adjust the system prompt as needed

​

4. Set Realistic Expectations

• The agent can’t do everything
• Have a way for visitors to reach humans
• Be clear about limitations in your welcome message

​

5. Test Regularly

• Test after any changes
• Try different types of questions
• Check on different devices
• Verify embed code still works after website updates

​

Frequently Asked Questions

​

Next Steps

• Create AI Employees
• Build Your Knowledge Base
• Set Up Integrations
• Manage Your Inbox