Skip to content

Never walk back to the table again.

Dummy Waiter is a wearable clip for waitstaff. It listens to the table, an AI turns the conversation into a structured order, and the waiter approves it before anything reaches the kitchen.

A concept project with working software behind it. The dashboard and the AI order parsing are built; the clip is a design concept. See what exists today.

#0142Table 7MayaElapsed 00:45

Heard at the table

Guest 1Margherita, please. Actually, make that two.

Guest 2Caesar salad, no anchovies, I’m allergic. And a tiramisu.

Parsed order

Awaiting approval
  • 2Margherita pizza$28.00
  • 1Caesar salad$12.00
    • No anchovies

    Allergy: anchovy

  • 1Tiramisu$9.00

Total$49.00

Sample data. This is the screen the waiter reviews before anything reaches the kitchen.

Every walk back is a table kept waiting.

Servers walk back to confirm what a guest said, then to the kitchen, then back to the floor. Each trip takes time off the floor and gives a detail, like a modifier or an allergy note, a chance to get lost between the table and the pass.

A dish made wrong because the order got garbled is wasted food and a disappointed guest. That is the one avoidable slice this project goes after.

Context from published sources

$25B+

a year is the full cost of the 11.4 million tons of food waste the US restaurant sector generates.

ReFED, Restaurant Food Waste Action Guide (opens in a new tab)
33M

people in the United States have at least one food allergy, which is why a missed note is a safety problem and not just a comped dish.

FARE, Food Allergy Facts and Statistics (opens in a new tab)

Industry figures from the linked sources. They describe the problem, not results from Dummy Waiter.

Four steps from conversation to kitchen.

The waiter stays in control the whole way. The AI does the transcribing and sorting, the person makes the call.

  1. 1

    Clip it on

    A small clip goes on the apron or collar. During early access, any device with a microphone does the same job.

  2. 2

    Take the order as usual

    Talk to guests the way you always do. The audio is transcribed and the AI picks out items, quantities, and modifiers. A note like “no anchovies” stays attached to the item it belongs to.

  3. 3

    Review what it heard

    Every item carries a confidence score. Allergy notes and uncertain items are flagged so you check them first, and fixing one takes seconds. Nothing is guessed at quietly.

  4. 4

    Approve and send

    One tap sends the approved order to the kitchen display. The AI proposes and the waiter decides, so nothing reaches the kitchen without an approval.

Approved orders land on the rail.

Orders arrive for review, move to the kitchen on approval, and show up on the floor plan and analytics. Open the demo to click through the working dashboard yourself.

Open the live demo

No sign-in. Sample data only.

  • #0143Table 9JonElapsed 00:12
    • 4Lunch special$52.00
    • 2Iced tea$7.00
    • 1Kids pasta$8.00
      • No cheese

    Total$67.00

    New
  • #0141Table 3PriyaElapsed 08:12
    • 1Carbonara$18.00
      • Extra pecorino
    • 1House red$11.00
    • 1Panna cotta$9.00

    Total$38.00

    Preparing
  • #0138Table 12MayaElapsed 21:40
    • 1Risotto$22.00

      Allergy: shellfish

    • 2Bruschetta$16.00
    • 3Sparkling water$12.00

    Total$50.00

    Late
  • #0136Table 5JonElapsed 14:05
    • 2Pizza diavola$30.00
      • Extra basil
    • 1Caprese$13.00
    • 2Limoncello$14.00

    Total$57.00

    Ready

Sample data. The demo runs the real dashboard on the same kind of data.

Concept render of the Dummy Waiter clip: a slim black rectangular badge with a small microphone opening and a spring clip on the back

A clip you forget you are wearing.

The concept is a slim clip that picks up the table conversation and relays it to the waiter’s phone. Small enough to ignore, quiet enough that guests do not notice it.

Design targets

Capture
MEMS microphone array with noise cancellation for loud dining rooms
Link
Bluetooth Low Energy to the waiter’s phone
Battery
20+ hours, enough for a double shift
Form
Closer to a name badge than a gadget, built to be forgotten

Concept render. The clip is not built yet, so during early access the software takes audio from any device with a microphone.

What exists today, and what does not.

Dummy Waiter is a concept product and a portfolio project. Here is an honest split between what runs, what is next, and what is only a design.

Built

Working software

  • Manager and waiter dashboard

    Order review and approval, kitchen display, floor plan, menu manager with import, analytics, staff and shifts.

  • AI order pipeline

    Speech to text with Whisper, then menu-aware order parsing with Claude and a rule-based fallback.

  • Data and auth

    Postgres with row-level security, realtime order updates, and device-based sign-in with staff PINs.

Planned

Next up

  • Phone companion app

    React Native app that relays the clip’s audio over Bluetooth Low Energy.

  • POS integrations

    Two tiers: a clean order summary beside the POS a restaurant already runs, then direct kitchen tickets for teams ready to switch.

Concept

Design only

  • The clip

    Enclosure render and a reference architecture around a MEMS microphone and an nRF52840 radio. No hardware has been built.

Named for the dumbwaiter, the small elevator that carries food so people do not have to make the trip.

The founder’s first restaurant job had one. The name stuck, and so did the idea: let a machine do the carrying so people do not have to make the trip.

Order mistakes cost restaurants money and goodwill, and the people who feel them first are the ones on the floor. Dummy Waiter is built from that point of view: help the waiter, never replace them.

Join the waitlist.

Leave your email to hear when there is something new to try. The role question helps show who this is for: owners, managers, or the people on the floor.

Prefer to look around first? Explore the live demo

WaitlistEarly access

Your email is used only for project updates. Read the privacy policy.