What does the AI host do during a call?

It greets callers, answers menu questions, finds a nearby pickup location, and confirms the order out loud, from greeting to confirmation.

What prerequisites do I need to deploy this?

You need an AWS account, Amazon Bedrock model access for Anthropic Claude Haiku 4.5, a phone number quota, Node.js 18.x or later, AWS CLI 2.x, git, and AWS CDK bootstrapped.

Where can I get the code?

The full solution is available in a sample repository on GitHub, which you clone and deploy with AWS CDK.

Large Language ModelsAudio & SpeechAI Business & IndustryAmazon AI Blog1h agoAug 25, 2026, 04:00 JST2 min read

AI host takes restaurant phone orders via Amazon Connect

Send on LINE

Key takeaway

AWS shows how to build an AI voice host for restaurants. It answers phone orders, handles menu questions, and confirms pickup.
The system integrates Amazon Connect with AI agents and backend tools.
It aims to reduce staff interruption from phone calls.

3 Key Points

What happened
AWS published a walkthrough for building a voice ordering system that answers a restaurant's phone line, greets callers, answers menu questions, finds a pickup location, and confirms the order aloud. It uses Amazon Connect, Amazon Lex V2 with Amazon Connect Agentic Voice, and Amazon Connect AI agents, powered by Anthropic Claude Haiku 4.5.
Why it matters
Phone orders often interrupt staff who are already helping customers in-store, causing hold times and handwritten mistakes. This system automates the call with no app, website, or sign-in—callers just dial a number—so staff can focus on in-person service.
What to watch
The solution is fully deployed with AWS CDK in a single script, and the sample repository is on GitHub. Prerequisites include an AWS account, Bedrock model access for Claude Haiku 4.5, and a phone number quota; US East (N. Virginia) is a suggested starting region.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

The post addresses a common pain point in restaurants: phone orders competing with in-person service. By automating the call with an AI agent, the solution lets staff stay at the counter while the AI handles the phone. The design separates the telephony channel from the ordering logic, so the backend can change without affecting the agent.

The architecture leverages Amazon Connect's native tools for telephony, speech, and orchestration. Amazon Connect Agentic Voice handles real-time speech recognition and synthesis, while an Amazon Connect AI agent powered by Claude Haiku 4.5 drives the conversation. Backend tools are exposed through MCP via AgentCore Gateway, making the system modular.

For business readers, the key takeaway is that AWS provides a turnkey path to automate phone ordering without building custom speech infrastructure. The walkthrough includes deployment details and prerequisites, but the underlying goal is to reduce operational strain during busy periods. Whether this becomes a common pattern depends on how easily restaurants can integrate it with their existing systems, but the modular design suggests AWS is positioning this as a repeatable solution.

FAQ

What does the AI host do during a call?: It greets callers, answers menu questions, finds a nearby pickup location, and confirms the order out loud, from greeting to confirmation.
What prerequisites do I need to deploy this?: You need an AWS account, Amazon Bedrock model access for Anthropic Claude Haiku 4.5, a phone number quota, Node.js 18.x or later, AWS CLI 2.x, git, and AWS CDK bootstrapped.
Where can I get the code?: The full solution is available in a sample repository on GitHub, which you clone and deploy with AWS CDK.

Amazon AI BlogRead Original Article