How it works
From the first ring
to the kitchen ticket.
What Ocrila does between your phone ringing and the order landing in your POS. Six steps, all of it inside the first few seconds.
The phone rings. Ocrila picks up.
Under a second. No hold music, no “press 1 for hours.” A warm hello in an ElevenLabs voice you pick with us.
Use a new number we set up, or port your existing one. Regulars dial the same digits they always have.
It listens through the noise.
Kitchen clatter. A table of six all shouting their preferences in the background. The caller who changes their mind halfway through. Ocrila is built on Deepgram for speech and Claude for reasoning, so these stop being problems your staff has to absorb.
Nobody presses 1 for anything.
It takes the order. Or books the table.
Your full menu is loaded into the assistant the moment the call starts. It knows your margherita from your marinara, what tonight’s special is, what’s 86’d, and which modifiers go with which dish. The questions a host usually fields (“does the carbonara have peas?”, “what’s gluten-free?”) get answered in the moment.
For reservations, it pulls live availability and confirms the table before the call ends.
It sends the order where it needs to go.
Square restaurants get direct integration today — orders push straight to the seller’s Square dashboard / KDS as an open ticket. Toastmenu sync is live; one-way order push is on the way once we’re a Toast integration partner.
For reservations, OpenTable and Resy bookings land in your Ocrila dashboard today; staff confirm them in those tools while we ship the direct booking API.
No POS yet? Tickets show up in the Ocrila dashboard, which restaurants typically pin to a kitchen tablet. Most Starter restaurants run it that way until they’re ready to integrate.
It transfers to a human when it should.
Complaints. Catering for fifty. The regular who only deals with Maria. You set the conditions and the number to transfer to. Ocrila stays on the line until your staff picks up.
Every call leaves a trail.
Full transcript and structured order data — caller number, items, totals — saved for every call. The dashboard shows you the call list with AI summaries today; the in-app transcript viewer and call search are rolling out next.
What’s under the hood
We don’t train our own speech or voice models. The teams who do it best are already shipping. We put the right pieces together and tune them to the way a restaurant call actually sounds.
- Speech
- Deepgram. The lowest word-error rate we’ve measured on real restaurant audio.
- Reasoning
- Claude. Holds the thread on ambiguous orders and mid-call changes.
- Voice
- ElevenLabs. We audition voices with you on the onboarding call and pick one before you go live.
Hear it answer a real call.
We'll set up a demo number with your menu loaded, before our meeting.
Book a demo