Orchestrate omnichannel AI agents with Foundry and Twilio | DEMSP393
Rachel Baskin demonstrates patterns for building omnichannel AI agents that keep conversation context across channels by combining Azure AI Foundry with Twilio.
Overview
Customer conversations often lose context when users switch channels (for example, from voice to SMS). This session shows how to close that gap using:
- Azure AI Foundry on Azure to establish models, voice, and agents
- Twilio for cross-channel memory, real-time orchestration, and conversation intelligence
- Agent Connect SDK to wire the pieces together
Session chapters (from the video)
Enterprise readiness and moving to hosted agent deployments
The session starts with enterprise readiness considerations and transitioning to hosted agent deployment.
Deployment options: hosted agents vs Azure Container Apps
A comparison of deployment options is covered, including hosted agents and Azure Container Apps.
Voice Live Connector: streaming inference and fast response
The presenter discusses Voice Live Connector, focusing on streaming inference and achieving fast responses.
Twilio setup for voice and SMS
The demo walks through configuring Twilio to connect channels into the agent workflow:
- Configuring a Twilio phone number to point to a webhook
- Setting up a webhook for an SMS connection
Omnichannel continuity demo (SMS continuation)
A demonstration shows continuing the same conversation over SMS, preserving context across channels.
Customization via GitHub repository
Customization options are discussed in the GitHub repo, including:
- Agent personality
- Agent knowledge
- Agent tools
Q&A: Conversation Relay and direct OpenAI connections
The session includes a question about Conversation Relay integration and whether a direct OpenAI connection is possible.
Provisioning for ISVs and multi-tenant setups
The presenter explains resource provisioning considerations for ISV use cases, including multi-tenant setups.