Skip to main content

​
What is AI Agent?

If you receive repetitive questions from clients and would like to manage day-to-day conversations automatically, Periskope’s AI Agent can handle it for you. Train this AI to answer important questions on your behalf, escalate issues to the right team members, and create tickets for important tasks intelligently.

​
How to Enable the AI Agent

Everything starts on the Overview page — go to AI > Overview.
1

Finish the setup checklist

If required setup is missing, the Overview page shows a Setup checklist with the steps still open:
  • Set the agent’s role and instructions (required) — takes you to Personalization
  • Set its identity and personality (required) — takes you to Personalization
  • Add answers it can use (optional) — takes you to the Knowledge Base
The agent cannot be turned on until both required steps are done.
Overview - Enable AI Agent
2

Turn the AI Agent on

Toggle AI Agent at the top of the Overview page. When active, the AI Agent starts responding to messages based on your configured activation criteria.
Overview - Enable AI Agent
3

Configure the five areas

The Configuration cards below the status card are the five things that shape how the agent behaves. Select a card to configure it — see the table below.
Overview - Enable AI Agent

​
The Five Configuration Areas

When self-learned answers are waiting on you, the Knowledge Base card shows a need review badge with a Review shortcut.

​
Where Everything Lives

The AI section of the sidebar is grouped into three parts:

​
How the AI Decides When to Respond

The AI Agent activates in a conversation when a message meets the Activation Criteria configured on the Activation page. The system uses intelligent prompt-based activation rather than simple keyword matching.

​
Activation Process

  1. Base Activation Rules: The system evaluates if a message meets core criteria such as:
    • Contains a question, complaint, technical problem, or concern
    • Requests an action, assistance, or support
    • Appears to be something that requires business attention
  2. Custom Activation Rules: Additional rules you define to customize when the AI should or should not respond
  3. Internal Contact Filtering: Messages from contacts marked as “internal” are automatically excluded from AI responses
Once activated, the AI will respond after a configurable delay using your Knowledge Base, Personalization settings, and available tools. If it doesn’t have a suitable response, it will follow the escalation rules you’ve configured.

​
AI Agent States

The AI Agent operates in four distinct states:
  • INACTIVE: AI Agent is enabled and monitoring conversations, waiting for messages that meet activation criteria
  • ACTIVE: AI Agent is actively listening and will respond to the next external message
  • THINKING: AI Agent is actively processing a message and preparing to respond
  • SNOOZED: Temporarily inactive after human intervention, will reactivate after the snooze duration

​
How to Stop the AI from Responding

When the AI Agent is active or thinking, you will see the AI button in the chat box along with a status indicator. Taking Over from AI Click the AI button to take over the conversation:
  • If AI is in THINKING state: AI immediately enters SNOOZED state for the configured snooze duration. It will not respond during this period and then returns to INACTIVE.
  • If AI is in ACTIVE state: AI immediately enters INACTIVE state. It will resume listening for messages that meet the activation criteria.
  • If an internal team member replies: Always triggers SNOOZED state, using the same snooze duration setting.
The AI also automatically enters SNOOZED state when an internal team member responds to the conversation.

​
AI Agent Decision Workflow

For every incoming message, the AI Agent follows this decision process:
  1. Message arrives in a conversation
  2. Internal Contact Check: If the sender is an internal team member → skip AI processing entirely
  3. AI Agent Status Check: If AI is disabled for this chat → no response
  4. Activation Criteria Analysis: AI evaluates the message against the activation prompt
  5. Response Delay: Wait for the configured delay period (giving human agents time to respond first)
  6. Knowledge Base Search: AI searches FAQs, Documents, and Self-Learned context
  7. Tool Calls: AI calls any relevant built-in or custom tools to gather information
  8. Response Generation: AI crafts a response based on all available context and your Personalization settings
  9. Action: AI sends a direct reply, creates a private note, creates a ticket, or escalates to a team member
  10. State Management: AI returns to ACTIVE state and continues monitoring the conversation

​
Next Steps

  • Activation - Configure activation rules, response delays, and operating hours
  • Personalization - Define the AI’s name, role, personality, and behavior
  • Knowledge Base - Train the AI with FAQs, documents, and self-learned context
  • Built-in Tools - Configure what actions the AI can take
  • Custom Tools - Connect the AI to your own APIs