Deploy a multimodal WhatsApp ordering assistant with Amazon Bedrock AgentCore
Machine Learning Blog
This article demonstrates how to deploy a multimodal WhatsApp ordering assistant built with Amazon Bedrock AgentCore and Amazon Nova 2, enabling customers to order via text, voice notes, or voice calls on a single business number.
- Single WhatsApp Business number hosts the assistant with text, voice note, and voice call channels sharing one backend and cross-channel memory
- Amazon Nova 2 Lite handles text through Converse API; Amazon Nova 2 Sonic handles real-time speech for voice notes and calls
- AgentCore Gateway exposes backend REST API as MCP tools (GetMenu, AddToCart, PlaceOrder) that agents call by name
- Asynchronous webhook architecture acknowledges requests fast via Lambda ingest, queues work in SQS, and processes with worker Lambda
- Shared AgentCore memory keyed by hashed customer ID enables cross-channel continuity—customers recognized across text, voice note, and call channels
- Amazon DynamoDB stores customer profiles, orders, menu items, carts, and locations; Amazon Location Service handles geocoding and nearest-location lookups
- Deploy entire system with AWS CDK; voice-call runtime uses Amazon KVS managed TURN relay for WebRTC media over VPC
- Customer identity based on hashed E.164 phone number with secret pepper; no login required
The solution separates WhatsApp layer, agent runtimes, and backend ordering logic for independent deployability while maintaining unified customer experience across channels.
The AWS News Feed is currently looking for gold sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.
The AWS News Feed is currently looking for silver sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.