Restaurant: AI phone ordering, live on the line.
Terrace Cantina is an owner-operated restaurant in Oak Brook Terrace, IL. They take phone orders 24/7 with a conversational AI agent that talks like a host, reads the actual menu, and drops the order straight into the kitchen via Clover POS. The host stays at the door. The kitchen prints clean tickets. The phone stops being a tax on the dinner rush.
The problem
Phone orders during dinner rush are where restaurants leak money.
Not in big dramatic ways. In small, repeated, exhausting ones. A missed call here, a re-keyed ticket there, a host pulled off the floor for three minutes during the worst slot of the night. Multiply by every Friday and every Saturday and the number is real.
7:42 PM Friday. Three calls in five minutes.
Dinner rush is loud. The host has a four-top waiting on a table, a couple at the bar wanting menus, and the phone is ringing. Two of the three callers wanted to order. One hung up. The kitchen never knew either call existed.
After-hours calls go to voicemail. Voicemail goes nowhere.
Someone calls at 10:47 PM trying to place a pickup for tomorrow. The call rolls to a voicemail nobody checks until the next afternoon. By then the customer ordered from somewhere else.
When the host takes the order, the host is not at the door.
A real phone order is two to three minutes if the menu is read right and the modifications get written down. That is two to three minutes the host is not seating, not taking names, not running plates. The whole front of house slows.
Hand-written tickets to the kitchen. Mistakes happen.
Phone order on a sticky note, walked back to the line, re-typed into the POS. Allergies missed. Modifications dropped. The customer is upset and the kitchen takes the hit.
What we built
A voice agent that does the host's job on the phone, every single call.
Not an IVR menu. Not a chatbot pretending to be a person. A real conversational agent on a real phone number, wired through to the real POS.
A real conversational AI on the phone, not a phone tree.
Callers reach the restaurant at +1-949-541-4031 and talk to an ElevenLabs conversational agent that sounds like a host. It greets the caller, reads the actual menu, takes modifications, confirms the order back, and asks pickup-or-delivery.
Order writes straight through to Clover POS.
Once the caller confirms, the order is injected into Clover via the Clover API in real time. The kitchen printer fires the same way it would for an in-person ticket. No re-keying. No sticky notes. No host pulled off the floor.
SMS confirmation closes the loop.
The caller gets a text from +1-833-892-8740 confirming the order, the pickup window, and the total. If they want to change something, they reply to the text. If something went sideways, the owner sees it in the inbox before the customer arrives mad.
It runs 24/7.
Saturday at 11 PM, Sunday morning before doors open, Tuesday during the lunch crush. The agent does not get tired, does not lose its voice, and does not miss a call because the host is seating a four-top.
The call flow
One inbound call, end to end.
- 01
Call hits Twilio.
Caller dials +1-949-541-4031. Twilio routes inbound audio to a webhook on infrastructure the OneHub360 team operates.
- 02
ElevenLabs agent picks up.
Conversational agent greets the caller, asks how they can help, listens. Reads from the live menu the restaurant maintains.
- 03
Order is taken in conversation.
Caller speaks the order, modifications, name, pickup time. Agent reads it back and confirms. No "press 1 for".
- 04
Order writes to Clover.
On confirmation, the order is posted to the Clover API. Kitchen printer fires the ticket like any other order. Inventory and reporting stay clean.
- 05
SMS confirmation goes out.
Caller gets a text from +1-833-892-8740 with the order, total, and pickup window. Reply to change. The owner sees the same thread.
The stack
Four pieces. No glue layer in the middle.
Same infrastructure that runs OneHub360. Voice, telephony, POS write-through, and the operations layer that keeps it up.
ElevenLabs
Text-to-speech and conversational AI agent that talks to the caller and reads the menu.
Twilio
Telephony layer. Routes inbound calls to the agent and sends SMS confirmations.
Clover API
Real-time order injection into the restaurant POS. Same ticket the kitchen prints for walk-ins.
OneHub360 ops
Health monitoring, deploy pipeline, on-call for the voice stack. Auto-restart and SMS alerts if the agent or the POS bridge drops.
What this actually is
Custom build for Terrace Cantina by the OneHub360 team.
Terrace is not a OneHub360 SaaS customer. They are a tailored client. We built and we operate the voice agent, the Clover bridge, and the SMS layer for them, on infrastructure we run.
The reason it lives on this site: the voice + POS + automation depth here is the same depth OneHub360 customers tap into. If you are building on OneHub360 and you need a real phone number talking to a real AI agent talking to a real POS, this is what that looks like in production.
Want similar voice, POS, or automation work on top of OneHub360? Talk to us.
FAQ
Questions worth answering.
Is Terrace Cantina a OneHub360 customer?
No. Terrace Cantina is a custom-build client of the OneHub360 team. They run a tailored AI voice + Clover POS stack we built and operate for them. The same infrastructure (ElevenLabs, Twilio, Clover, the deploy and monitoring layer) is what powers OneHub360.
Does this only work with Clover?
Clover is what Terrace runs, so that is what the integration writes to. The same pattern works for Square, Toast, or any POS with an order-injection API. The voice and SMS layers do not change.