Conversations

What are Conversations?

Conversations are the chat interactions between your users and AI agents. Each conversation is tracked per agent, showing the full message history, user information, and escalation status. This allows you to monitor how your agents are performing and step in when human support is needed.

Two Perspectives

The conversation system has two sides:

• Admin View — As a business owner, you see all conversations across all agents from your admin dashboard
• User View — Your WebApp users can access their escalated conversations through a dedicated Support page

Conversation Statuses

Each conversation has one of these statuses:

• Active — Ongoing conversation between user and AI agent
• Escalated — Transferred to human support, waiting for or receiving human assistance
• Resolved — Issue has been resolved by human support
• Ended — Conversation has ended

How Escalation Works

Conversations can be escalated to human support in two ways:

User-Initiated Escalation

Users can explicitly request escalation by saying things like "I want to speak to a human", "escalate this", or "transfer me to support". The agent will immediately escalate the conversation.

Agent-Suggested Escalation

The AI agent may suggest escalation at appropriate moments — for example, when a question is too complex or outside its knowledge. The agent will ask if the user would like to escalate, but will only proceed if the user confirms.

Direct Escalation

Users can also create a support conversation directly from the Support page, bypassing the AI agent entirely. This is useful when a user knows they need human help from the start. The conversation is created with an "Escalated" status immediately.

What Happens When Escalated

When a conversation is escalated:

• The conversation status changes to "Escalated"
• A system message is added indicating the escalation
• The AI agent stops responding to messages
• An escalation timestamp is recorded for tracking
• The conversation appears in the admin's needs-response queue

Admin: Viewing Conversations

From your agent's Overview tab, you can see conversation statistics and a list of recent conversations. You can:

• Filter — By status (e.g., Escalated), by users, or by sources used in responses
• Search — Find specific conversations by content
• Sort — Order by date, message count, or other criteria

Click any conversation to view the full details.

Conversation Detail View

The conversation detail page shows:

• User Information — Name, email, and user type (WebApp or embed)
• Conversation Metadata — Start time, message count, and current status
• Full Message History — All messages with timestamps, AI vs human distinction, and source references
• User Feedback — Thumbs up/down ratings on agent responses
• Pages Visited — Which pages the user was on during the conversation

Source References

For AI responses, you can see which sources and Q&As were used to generate each response. This helps you understand how your agent is using its knowledge base and identify areas for improvement.

Admin: Responding as Human

When a conversation is escalated, you can respond directly as a human agent:

1. Open the escalated conversation from your agent's Overview or the needs-response indicator
2. Type your response in the message input area
3. Optionally attach files (images, PDFs, documents up to 10MB)
4. Send the message — it will appear with a purple "Human Agent" badge

Human responses are clearly distinguished from AI responses so users know when they're talking to a real person.

Resolving Conversations

Once you've addressed the user's needs, click the "Resolve" button to mark the conversation as resolved. The user will see the conversation moved to their resolved section.

User: Support Page

WebApp users can access their escalated conversations through the Support link in the WebApp navigation. This page shows:

Active Issues

Escalated conversations that need attention, organized by priority:

• Conversations with unread messages — Highlighted with a blue background and badge
• Needs your response — Human agent has replied and is waiting for user
• Waiting for agent reply — User is waiting for human support to respond

Resolved Conversations

Previously resolved conversations are shown in a collapsible section. Resolved conversations from the last 30 days are visible here. If a resolved conversation has unread messages, it will appear in the Active Issues section instead.

Conversation Details Modal

Users can click the info icon on any conversation to see an activity timeline showing:

• When the conversation started with the AI bot
• When it was escalated to human support
• When a human agent engaged
• When it was resolved (if applicable)

User: Continuing Conversations

From the Support page, users can click any escalated conversation to open the full chat view and:

• See the complete message history
• Continue chatting
• Send file attachments (images and documents)
• Resolve the conversation themselves if satisfied

Dashboard Widget

Users also see an "Escalated Conversations" widget on their dashboard showing their most recent active issues. This provides quick access without navigating to the full Support page.

User Types

Conversations track where users came from:

• WebApp Users — Authenticated users who logged into the WebApp
• Embed Users — Users chatting through an embedded widget on your external website
• Email Users — People who interact with your agent via email. Email conversations display the sender's email address, name, and subject line

For embed users, if you pass user data through the widget integration, conversations will be tracked with their email. Anonymous embed users are identified by session ID.

Email conversations are automatically created when an email is sent to an address assigned to an agent with email integration enabled. Replies are threaded into the same conversation, and escalation works the same way — the sender receives an email notification when their conversation is escalated to human support.

File Attachments

Both admins and users can send file attachments in conversations:

• Images — JPEG, PNG, GIF, WebP
• Documents — PDF, Word, Excel, PowerPoint, text files
• Maximum — Up to 5 files per message (10MB per file)

Recipients can download attachments directly from the conversation view.

Best Practices

• Monitor escalations regularly — Check your needs-response count to ensure timely human support
• Use the Escalated filter — Filter conversations by "Escalated" status to prioritize those needing attention
• Respond promptly — Quick response times improve customer satisfaction
• Resolve promptly — Mark conversations as resolved once addressed so users know their issue is handled
• Review source references — Check which sources your AI uses to identify knowledge gaps
• Review feedback — Pay attention to thumbs down ratings to improve agent responses
• Analyze escalation patterns — If certain topics frequently escalate, add better sources or Q&As to address them
• Configure human agent settings — Set up your human agent name and avatar in Agent Settings so users know who they're talking to