Call Center AI — Microsoft AI Phone Agent

What it is

Call Center AI is a Microsoft open-source project that enables AI-powered phone calls. You can make outbound calls via API or receive inbound calls on a configured phone number. The system uses Azure Communication Services for telephony, Azure Cognitive Services for speech, and GPT-4.1 for conversation handling.

It is designed for teams building automated phone support for insurance claims, IT helpdesk, appointment scheduling, and customer service. The project is Apache-2.0 licensed.

How it saves time or tokens

Traditional IVR systems require rigid decision trees that frustrate callers. Call Center AI replaces those with a natural language agent that can understand context, ask follow-up questions, and resolve issues conversationally. The LLM handles intent recognition and response generation in a single pass.

Deployment uses Azure services, so there is no need to build telephony infrastructure from scratch. The project provides Docker Compose configuration for quick local testing.

How to use

1. Clone and configure:

git clone https://github.com/microsoft/call-center-ai.git
cd call-center-ai
cp .env.example .env
# Fill in Azure Communication Services credentials
# Fill in Azure Cognitive Services key
# Fill in OpenAI/Azure OpenAI API key

1. Deploy with Docker Compose:

docker compose up -d

1. Configure a phone number in Azure Communication Services and point the webhook to your deployment.

Example

# Example: Making an outbound AI call via the API
import requests

response = requests.post(
    'http://localhost:8080/call',
    json={
        'phone_number': '+1234567890',
        'scenario': 'insurance_claim',
        'context': {
            'claim_id': 'CLM-2026-001',
            'caller_name': 'Jane Smith'
        }
    }
)

print(response.json())
# {'call_id': 'abc-123', 'status': 'initiated'}

The AI agent follows the scenario script, asks clarifying questions, and logs the conversation transcript.

Related on TokRepo

• AI Tools for Automation -- Other AI-powered automation tools and workflows
• AI Tools for Voice -- Voice AI tools for speech synthesis and recognition

Common pitfalls

• Azure Communication Services requires a paid Azure subscription and a purchased phone number. There is no free tier for telephony.
• Speech recognition quality depends on the Azure Cognitive Services tier. The free tier has lower accuracy than the standard tier.
• GPT-4.1 usage costs can add up quickly for long phone calls. Set conversation length limits and monitor token usage via Azure OpenAI metrics.