Command Palette

Search for a command to run...

WhatsApp Flow

Understand how WhatsApp conversations move through Botrelic.

Overview

Botrelic processes WhatsApp conversations using Workspace routing, AI agents, sessions, and human handoff workflows.

Complete WhatsApp Message Flow

Customer sends WhatsApp message
Meta Cloud API

Receives message & forwards webhook

Botrelic receives conversation
Workspace routing occurs
Session created
AI Agent processes message
Response generated
Fallback detection
Human Handoff (if required)

Core Flow Mechanics

Workspace Routing

  • Every WhatsApp number belongs to a Workspace
  • Incoming conversations route into the correct Workspace
  • Workspaces manage: AI agents, human teams, and sessions

Workspace isolation ensures conversations remain separated between organizations, guaranteeing data privacy.

Session Lifecycle

When a customer sends a WhatsApp message:

  • Botrelic creates a Session
  • Session stores conversation history
  • AI context persists across messages
  • Human handoff status is tracked

Sessions remain active until they hit a timeout or are manually closed by an agent.

AI Agent Processing

During an active session, AI agents:

  • Understand customer messages
  • Search the knowledge base
  • Generate responses
  • Detect fallback situations

Fallback detection triggers on:

  • Unsupported questions
  • Low confidence responses
  • Specific escalation keywords

Human Handoff

If a fallback triggers during processing:

  • The conversation automatically routes to the human team
  • Human agents receive the active session in their inbox
  • The customer continues inside the same WhatsApp thread

Sandbox vs Production

Best Practices

Optimize Fallback Rules

Ensure proper routing triggers so customers are not stuck in loops when AI lacks confidence.

Maintain Knowledge Base

Consistently update your Workspace documents to improve AI response accuracy over time.

Assign Human Teams

Correctly staff and assign human agents to monitor escalations and prevent user drop-off.

Test Before Deployment

Always use Sandbox testing to simulate edge cases before launching to production.

Related Documentation