Drive-Thru Automation: Voice AI, Gesture, or Both?

Drive-thru automation has two halves, and most of the conversation only covers one. Voice AI takes the order at the speaker post. Gesture control handles the screens the order then touches: the kitchen display where crew bump and recall orders with gloved hands, and the confirmation board the customer reads from the car. Voice replaces a headset conversation. AirTouch replaces a greasy bump bar and a touchscreen in the weather, using a standard RGB camera and the touch and key input those displays already understand. The best lane uses both, each where it is strong.

What does voice AI do well at the drive-thru?

Taking the order. A voice agent greets, upsells, handles the menu, and hands a structured order to the point of sale without a crew member on the headset. It frees the person who used to split attention between the speaker and the window, and it is consistent through a rush. Several large chains have run voice AI pilots and rollouts, and the technology has improved quickly.

It also has well-documented limits. Noise from idling engines, music, and kids in the back seat; accents and menu slang; complex customizations; and the moment a customer changes their mind mid-sentence. Operators handle these with a human fallback, which is sensible and also why voice alone rarely automates the whole lane.

Where voice wins: the customer conversation at the speaker post, menu knowledge, upselling, and consistency across shifts.

What does gesture control do that voice cannot?

Control a screen in a place where talking does not work. Two places, specifically.

• The kitchen display. Crew bump, recall, park, and advance orders dozens of times an hour with hands that are gloved, wet, or full. Bump bars collect grease and touchscreens get tapped with a knuckle. The kitchen is also the loudest room in the building, so voice is a poor input for the crew even where it is excellent for the customer. AirTouch maps a raised hand and a few gestures to the same key and touch events the kitchen display system already accepts, from the station, without contact. The full setup is in gesture control for drive-thru and kitchen displays.

• The order confirmation board. A customer confirms or corrects an order from the car window without touching a screen in the rain. The board sits behind glass, which the camera tracks through, within reason, and patented intent detection keeps a passenger's wave from confirming anything in most setups.

Where gesture wins: discrete, repetitive actions on screens that are loud, dirty, out of reach, or behind glass.

Where gesture loses: it does not take an order. Nobody wants to pinch through a menu from a car.

Why is the combination the right answer?

Because the two technologies fail in different places. Voice fails where it is noisy and the action is a precise tap. Gesture fails where the task is a conversation. Put together, the lane looks like this: the voice agent takes the order at the speaker post and sends it to the point of sale; the order lands on the kitchen display, where the crew bumps it by gesture without taking off a glove; the customer confirms on the lane board with a raised hand; the window handoff stays human. No part of the existing point of sale or kitchen display software changes, because AirTouch outputs standard input and the voice agent outputs a standard order.

It also sequences well. Voice is the larger project with the larger risk, since a misheard order costs a customer. The crew-side display is small, cheap to pilot, and invisible to the customer, which makes it the sensible first step whether or not voice is on the roadmap.

Which should you choose?

What does a gesture station need?

A camera per station with a clear view of where the crew member stands, delivering 20 frames per second or more, and the Windows, Linux, or Android computer already driving the display. Video is processed on that computer and does not leave it. Food service gloves track well, and saturated colors use the Support Color Glove setting. Steam, heat shimmer, and grease on a lens degrade tracking, so placement and a wipeable housing matter, and some stations may call for a specialized camera. Through our partnership with Orbbec, AirTouch also takes advantage of 3D depth sensing and IR cameras when a deployment calls for them. See the Orbbec page.

Where to start

Start with the bump bar everyone complains about. Contact us with a photo of the line and the kitchen display system you run, and we will suggest camera placement and a gesture map. Integrators can start a free trial on a PC with a webcam and drive the display software today.

Related: gesture control for drive-thru and kitchen displays, the touchless kiosk, gesture control for Android displays, touchless screens for food plants and warehouses, and gesture control and hand tracking software.