Case study, status: prototype
Voice AI phone ordering agent
A phone agent that takes food orders by voice, built on telephony, an LLM with tool use and speech synthesis.
Overview
Problem
Restaurants lose orders when phones ring during rush hours, and staff spend time repeating the menu and taking details by hand.
Objective
Answer calls automatically, take a complete and valid order in natural conversation, and hand structured order data to the kitchen system.
Solution
- A telephony backend on Telnyx receives the call and streams audio to the conversation layer.
- Claude drives the conversation with a small set of tools: look up menu items, add to order, confirm order.
- A separate prototype explored the same flow with an ElevenLabs conversational agent to compare voice quality and latency.
Architecture
Caller
Telnyx telephony
Ordering backend
Claude (tool use)
Menu + order tools
Order output
AI workflow
- 1Greet the caller and ask for the order
- 2Match spoken items to the menu with a lookup tool
- 3Ask for missing details such as size or quantity
- 4Read back the full order and ask for confirmation
- 5Emit the structured order only after confirmation
Technology
- Telnyx
- Claude API
- ElevenLabs
Want something similar built?
Send a short brief. We'll reply personally with questions or a suggested next step — usually a technical discovery call.