Microsoft Developer presents a demo session featuring Pablo Salvador Lopez, who illustrates the creation of real-time omnichannel voice AI agents on Azure, including architectural patterns and integration strategies.

Building Omnichannel Voice AI Agents with Azure

Presented by: Pablo Salvador Lopez (Microsoft AI GBB Team)

In this demo-driven video session, you’ll discover how Azure services enable the development of omnichannel voice AI agents that can handle real-world scenarios such as insurance claims. The session demonstrates:

  • Multilingual Conversations: Voice AI agents manage real-time conversations in multiple languages, breaking barriers for global operations.
  • Intent-Based Handoffs: The system detects user intent to facilitate automated agent switching or human escalation when needed.
  • Complex Workflow Execution: Orchestration of backend processes is handled automatically based on the conversation’s context and user’s intent.

Key Architecture Components

AI Layer

  • Multi-Agent Orchestration: Coordinates numerous AI agents for different tasks or specialties within a single conversation.
  • Speech-to-Text and Text-to-Speech: Utilizes Azure AI Services for accurate speech recognition and natural-language response generation.

Application Layer

  • Integrates real-time technologies like WebSocket and WebRTC for seamless voice communication and streaming.

Telephony Integration

  • Azure Communication Services provide virtually limitless telephony integration, allowing agents to connect with users on traditional phone networks as well as digital platforms.

Demo Scenario

The showcased case involves an insurance claim, simulating:

  • Automated voice interactions for claim initiation and status tracking
  • Handling multilingual requests and mapping user input to specific intents
  • Backend workflow executions, like scheduling appointments or verifying documents

Learn More

  • Watch the full episode: https://aka.ms/SandSEp23
  • Explore Azure Communication Services and Microsoft AI platform documentation for further implementation guidance.