🎓 Build Your Own AI-Powered Voice Agent

A hands-on, LOW-CODE workshop using Azure AI Foundry & Container Apps

🤔 Why Azure AI Foundry?

Copilot Studio and Power Platform are excellent low-code tools for building agents quickly. But sometimes you need more flexibility, customization, and control over your AI solutions.

🤖 Copilot Studio

Best for:

  • Quick Copilot deployment
  • Standard conversation flows
  • Pre-built connectors
  • Citizen developers

🚀 Azure AI Foundry

Best for:

  • Custom AI workflows
  • Advanced model fine-tuning
  • Full code control
  • Enterprise-grade solutions

🎯 Why Studio Couldn't Build CORA

CORA needs specific capabilities that go beyond traditional Copilots. Here's what we require and why Studio can't deliver:

🎤 Real-Time Voice Streaming

  • CORA needs: Interactive voice conversations using Web Speech API with real-time speech recognition and synthesis
  • Studio limitation: Test pane only supports dictation (record → send recording), not streaming voice interaction
  • M365 Copilot limitation: Similar dictation-only approach, no real-time speech SDK integration
  • Why it matters: Customer service training requires natural, flowing conversations - not "speak, wait, submit"

🎨 Highly Customized UX

  • CORA needs: Custom interface with mood selection (6 personality types), voice selection (multiple voices), quality controls (rate, pitch, volume), and real-time analytics dashboard
  • Studio limitation: Standard chat interface with limited customization options
  • Why it matters: Training simulators need controls that don't exist in typical chatbots - learners must experiment with customer personalities and voice characteristics

💾 Flexible Data Storage & Integration

  • CORA needs: Lightweight Azure Table Storage for conversation analytics, easy Managed Identity integration, simple schema (conversation ID, scores, timestamps)
  • Studio approach: Pushes toward SharePoint Lists or Dataverse (more complex, enterprise-focused)
  • Why Azure Tables works better: Simple key-value storage, no schema overhead, native SDK support, perfect for analytics aggregation
  • Managed Identity: CORA uses Azure managed identity for secure, passwordless authentication to Storage, OpenAI, and Application Insights - simpler than Studio's connection management

📊 Advanced Observability & Telemetry

  • CORA needs: OpenTelemetry integration with custom spans tracking mood, tokens, scores, duration, and cost per conversation
  • Studio limitation: Pre-built analytics dashboard, limited custom telemetry
  • Why it matters: We track business metrics (which moods use most tokens?) alongside technical metrics (API latency)

🧪 Multi-Criteria Scoring System

  • CORA needs: Custom 5-criteria evaluation (professionalism, clarity, problem resolution, empathy, efficiency) with 25-point scale
  • Studio limitation: Would require complex Power Automate flows to parse AI responses and calculate scores
  • Why code is better: Python parses JSON responses directly, calculates scores inline, stores in Azure Tables - all in <50 lines

🚀 Scalable Containerized Deployment

  • CORA needs: Docker container on Azure Container Apps, scale-to-zero for cost efficiency (~$10-20/month), git-based deployments
  • Studio approach: Managed service (great for most cases!), but less control over hosting, scaling, and costs
  • Why containers: Version control, CI/CD pipelines, infrastructure-as-code, portable to other environments

🎯 The Bottom Line: Studio is fantastic for standard chatbots and conversational flows. But CORA is a specialized training simulator with voice streaming, custom UX, detailed analytics, and production-grade observability. For this level of customization, Foundry + code-first is the right tool for the job!

This workshop teaches you: When to use Copilot Studio vs Azure AI Foundry, how to leverage the power of Azure AI services, and how to build production-ready AI agents with full customization and control.

🎉 See What You’ll Build!

CORA Meet CORA: The AI That Makes Humans More... Human

Look, we get it. Autonomous AI agents are everywhere these days - ordering your pizza, booking your flights, probably planning world domination. But here's the thing: humans still like talking to humans. 🗣️

Enter CORA - your secret weapon for making human customer service agents sound like they've been doing this for years! Think of it as a real-time performance coach that whispers "you're doing great!" (or occasionally "maybe try empathy?") while analyzing every conversation.

🎯 The Mission: Use AI to help real humans deliver better customer experiences. It's like having a tiny AI cheerleader in your pocket, but way more useful and less noisy.

Cora Voice Agent Simulator Demo

See CORA in Action: A fully functional AI-powered customer service training platform with real-time conversation analysis, performance tracking, and intelligent feedback!

🎬 Watch the live demo above - click to enlarge and see a real customer service conversation being scored in real-time!

🎤 Speech Technology: Uses browser Web Speech API (desktop browsers). Text chat works on all devices, voice requires desktop Chrome/Edge/Safari.

💡 Don’t Be Nervous - This is LOW-CODE!

✨ You don't need to be a coding expert!

📦

All Code Provided

Complete, working application code ready to deploy

🔧

Pre-configured SDKs

Python packages and Azure SDKs already set up

🏗️

Infrastructure Templates

Bicep templates deploy everything with one command

📚

Step-by-Step Guides

Detailed explanations for every command and concept

Your role: Follow the instructions, run the commands, and understand how the pieces fit together.
We provide the code - you provide the curiosity! 🚀

📚 Workshop Modules

Module 1 30 min Not Started

Solution Overview

Understand the architecture, use cases, and components of the voice agent simulator. Learn how Azure AI Foundry integrates with other Azure services.

  • What is Azure AI Foundry?
  • Architecture overview
  • Use cases & business value
Start Module 1 →
Module 2 45 min Not Started

Infrastructure Setup

Deploy Azure infrastructure using Azure Developer CLI (azd). Set up Container Apps, Storage Account, and authentication with a single command.

  • Azure Developer CLI (azd)
  • Container Apps environment
  • Storage Account setup
Start Module 2 →
Module 3 60 min Not Started

Application Deployment

Deploy the Flask application to Azure Container Apps. Understand the code structure, frontend design, and Docker containerization process.

  • Flask application walkthrough
  • Docker containerization
  • Deploy to Container Apps
Start Module 3 →
Module 4 45 min Not Started

Azure OpenAI & AI Foundry

Configure AI Foundry project, deploy GPT-4 model, and implement intelligent conversation analysis with scoring capabilities.

  • AI Foundry project setup
  • GPT-4 deployment
  • Conversation scoring
Start Module 4 →
Module 5 45 min Not Started

Analytics Dashboard

Implement performance tracking with Chart.js visualizations and Azure Table Storage integration for historical data analysis.

  • Chart.js implementation
  • Table Storage integration
  • Performance metrics
Start Module 5 →
Module 6 45 min Not Started Not Started

Advanced Topics Optional

Explore application monitoring, AI Foundry evaluation tools, and production optimization strategies for enterprise deployments.

  • Application Insights
  • AI evaluation tools
  • Production best practices
Start Module 6 →

⏱️ Workshop Format

Self-Paced

Complete modules at your own pace

4-6 hours total

Instructor-Led

Follow along with live guidance

6 1-hour sessions led by the CSA team

📋 Prerequisites

✅ What You'll Need Before Starting:

  • Azure demo/development tenant with active subscription
  • Contributor role on Azure subscription/resource group
  • VS Code (or preferred IDE)
  • Azure CLI & Azure Developer CLI (azd) - Required!
  • Python 3.9+ installed

🚀 What We'll Deploy Together in the Workshop:

  • Azure OpenAI with GPT-4/GPT-4o model
  • Azure Container Apps for hosting
  • Azure Storage Account for analytics
  • Azure AI Foundry project
  • Managed identities and authentication

💻 No coding experience required! This is a guided, low-code workshop where all code is provided. All Azure resources are deployed automatically with a single azd up command!

📋 View Detailed Prerequisites →

🚀 Getting Started

Choose Your Download Method:

Option 1: Git Clone

Recommended for Git users

git clone https://github.com/yourusername/cora-voice-agent-training.git
cd cora-voice-agent-training

Option 2: Download ZIP

No Git required! Perfect if you're not familiar with Git.

Download Workshop Files (ZIP)
  1. Download the ZIP file
  2. Extract to your local machine (e.g., C:\Workshops\Cora)
  3. Open the folder in VS Code

Once You Have the Files:

1

Review prerequisites - Make sure you have everything ready

2

Choose your path - Self-paced or instructor-led

3

Start Module 1 - Begin your AI learning journey!

💰 Cost Estimate

Running this workshop costs approximately $5-10 total:

  • Container Apps: ~$0.50-2.00/day
  • Azure OpenAI: ~$0.03-0.10 per conversation
  • Storage: <$0.10/day

⚠️ Remember to clean up resources after completion!

📖 Additional Resources

Azure AI Foundry Docs Container Apps Docs Azure OpenAI Docs GitHub Issues

🆘 Need Help?

GitHub Issues

Report bugs or ask questions on our repository

Instructor Support

Available during live training sessions

Community Forums

Connect with other learners in Azure community

Ready to Build Your AI-Powered Voice Agent?

Start your journey into AI-powered customer service solutions!

🚀 Start Module 1 Now!

Last Updated: December 2025 | Version 1.0