Building a restaurant telephony AI host with Amazon Connect

This article demonstrates how to build a voice-based restaurant ordering system using Amazon Connect, AI agents, and Agentic Voice that handles phone calls end-to-end without requiring apps or logins.

• Callers dial a phone number and an AI host greets them, answers menu questions, finds nearby pickup locations, and confirms orders entirely by voice
• Amazon Connect Agentic Voice provides real-time speech recognition and expressive text-to-speech with confidence-based end-of-turn detection
• Amazon Connect AI agents powered by Claude Haiku 4.5 orchestrate conversations and call backend tools through Model Context Protocol (MCP)
• AI Guardrails keep conversations safe and on-topic by filtering harmful content, blocking denied topics, and filtering profanity
• System identifies callers by phone number without login, stores orders in DynamoDB, and uses Amazon Location Service for pickup recommendations
• Solution deploys via AWS CDK with no Docker or Python required; costs approximately $35/month for 1,000 five-minute voice orders

The architecture separates telephony, AI reasoning, and backend services, allowing the same backend to serve multiple channels while keeping conversation logic independent from implementation details.