Conversation History

Conversation History in Agent Studio gives administrators a comprehensive view of chat conversations in your PrimePilot deployment. PrimePilot Express shows recent conversations on Activity. This page allows you to browse, search, filter, and analyze chat sessions to gain insights into user interactions and system performance.

Understanding Conversations

Conversations are chat sessions between users and your deployed squads. Each conversation represents a complete interaction thread where users ask questions, receive responses from AI agents, and may be handed off to human support when needed. All conversations are associated with a squad, which determines which agents are available to handle the conversation.

The conversation history shows key information about each chat including the user who initiated it, when it was created, the number of messages exchanged, and the current handoff status. Administrators can view full conversation threads, export conversation data, and manage human handoffs directly from this page.

Human Handoff Flow

PrimePilot supports seamless handoff between AI agents and human support users. This allows conversations to transition from automated AI handling to human intervention when needed, with full context preservation.

Normal AI Handling

By default, conversations are handled by AI agents through squads. The squad's agents process user messages, switch between agents as needed based on intent detection, and provide automated responses. During this phase, the conversation status shows no handoff status, indicating normal AI operation.

Automatic Handoff Detection

The system can automatically detect when a conversation requires human intervention. During agent switching, the AI may determine that a user's request is too complex, requires human judgment, or needs personalized assistance. When this happens, the conversation status is set to "pending" and appears in the handoff queue for support users to review.

Support User Intervention

Support users can view pending handoffs and take control of conversations that need human attention. To take over a conversation, support users simply:

  1. View the conversation by clicking the "View Chat" button in the conversation history table
  2. Read the conversation to understand the context and user's needs
  3. Reply to the conversation - the system automatically takes over the chat when a support user sends their first message

When a support user sends a message in a conversation they haven't taken over yet, the system automatically:

  • Updates the conversation status to "active"
  • Assigns the conversation to that support user
  • Creates the message as a human assistant message (not an AI response)

This automatic takeover happens seamlessly - support users don't need to click a separate "takeover" button. Simply viewing and replying to a conversation is enough to take control.

Support users can proactively take control of any conversation, not just those flagged as pending. This allows support teams to monitor and intervene in conversations as needed, even if the AI hasn't flagged them for handoff.

Returning to AI

After handling a conversation, support users can release it back to AI control. When released, the conversation returns to the squad's AI agents, which continue handling the conversation with full context of both the original user messages and the human support interaction. The handoff status is cleared, and the conversation resumes normal AI operation.

This handoff flow ensures that users always receive appropriate assistance, whether from AI agents or human support, with seamless transitions between the two and complete conversation context preserved throughout.