Multi-Channel Customer Support In this example, a Manager Agent orchestrates a customer support workflow across chat, email, and SMS channels, dynamically routing tasks to specialized agents based on user preferences and message content.

​
Workflow Overview

  1. User Inquiry Intake
    • The Manager Agent receives an incoming support request via the primary channel (chat, email, or SMS).
  2. Channel Classification
    • A sentiment analysis worker agent evaluates the message tone and urgency.
    • If negative sentiment or high urgency is detected, prioritize chat or SMS for real-time engagement.
  3. Intent Detection & Routing
    • An NLU agent parses the request to identify intent (e.g., billing, technical issue, feature request).
    • The Manager Agent maps the intent to the appropriate domain specialist agent (Billing Agent, Tech Support Agent, Product Feedback Agent).
  4. Contextual Response Generation
    • The chosen worker agent generates a draft response, leveraging the Knowledge Base for relevant documentation or past resolutions.
  5. Approval & Escalation
    • For sensitive cases (e.g., account closures), the Manager Agent routes the draft to a human supervisor agent for review.
    • Based on the supervisor’s decision, the final response is sent via the user’s preferred channel.
  6. Follow-up & Logging
    • After resolution, a follow-up task is scheduled to check user satisfaction.
    • All interactions, metrics, and context are logged in Long-Term Memory for future personalization.

​
Key Benefits Demonstrated

  • Adaptability: Dynamically switches channels for optimal user experience.
  • Scalability: Distributes workload across multiple worker agents to handle high volumes.
  • Resilience: Implements fallback to human review for edge cases.
  • Personalization: Leverages memory and KB context for tailored responses.