Building Omnichannel Voice AI Agents with Azure: Multilingual Multi-Agent Architecture
Microsoft Developer hosts Pablo Salvador Lopez and Priyanka Vergadia in this episode, providing hands-on insights into building scalable omnichannel Voice AI agents with Azure’s Speech Services and multi-agent orchestration.
Building Omnichannel Voice AI Agents with Azure
Overview
This session from Sip and Sync with Azure, featuring Pablo Salvador Lopez and Priyanka Vergadia, demonstrates how organizations—like those in insurance—can create real-time, multilingual customer service bots using Azure Voice AI.
Key Topics
- Demo Walkthrough: Real-world insurance claim scenario
- Multilingual Support: Seamless switching between languages (English, Spanish)
- Intent-Based Workflow: Automatic agent handoffs for specific queries, e.g., claims or policy details
Architecture Breakdown
AI Layer
- Multi-agent orchestration enabling specialized agents for tasks
- Intent recognition and AI-driven conversation management
Speech Layer
- Integration of Azure Speech Services for speech-to-text and text-to-speech
- Language detection and real-time voice interactions
Application Layer
- Use of WebSocket and WebRTC for real-time voice pipelines
Telephony Integration
- Azure Communication Services connects voice AI to telephony systems
Tools and APIs
- Voice Live API accelerates voice application development
- Multi-Agent Accelerator solution provides reusable architecture patterns
Best Practices
- Architect for scalability and seamless agent handoffs
- Optimize multilingual pipelines for diverse user bases
- Leverage real-time processing through Azure’s Speech and Communication Services
Getting Started
- Source Code Repo
- Azure Speech Services
- Voice Live API Docs
- Multi-Agent Accelerator
- Try Azure Free
- Sip and Sync Playlist
Speakers
- Pablo Salvador Lopez - Principal Solution Engineer (LinkedIn)
- Priyanka Vergadia – Principal Cloud Advocate (LinkedIn, Twitter)