Building a restaurant telephony AI host with Amazon Connect
aws ·
This article describes how to build a voice-based ordering system for restaurants using Amazon Connect, designed to handle phone orders that typically burden staff. The system answers an inbound call, greets the caller, answers menu questions, finds a nearby pickup location, and confirms the order aloud, all without an app, website, or login. It uses Amazon Connect for telephony, Amazon Lex V2 with Amazon Connect Agentic Voice for real-time speech recognition and synthesis, and an Amazon Connect AI agent powered by Anthropic Claude Haiku 4.5 to orchestrate the conversation. Backend services for menus, carts, orders, and locations are exposed as MCP tools through AgentCore Gateway, allowing the AI agent to call them by name. The solution is deployed using AWS CDK and includes components such as AWS Lambda, Amazon DynamoDB, Amazon API Gateway, and Amazon Location Service, with monitoring and encryption provided by CloudWatch and KMS. The article also outlines the inbound call flow, prerequisites like model access and phone number quota, and notes that all telephony, speech, and AI agent components are managed within a single Amazon Connect deployment.
aws