AI Robot Bartender


Interactive, LED, Mechanical, Product Design
Case Study

Here at Smooth we take fun very seriously: our annual holiday party is no joke. But often the fate of a good host is being too busy to enjoy the party, so in 2019 we built our first robotic bartender (Arthur) to help out with party-hosting duties. Our barbot’s wacky features are upgraded yearly and too complex to name, but usually include some form of user identification, a panoply of drinkable offerings, and a truckload of personality.

The first iteration of Arthur was dressed like a slovenly game show host and required our guests to shout their drink order into a microphone over the party’s din. Arthur used a camera to map their facial features so it could identify them when they returned for another drink. Upon their return Arthur would greet them by name, entertain them with a witty anecdote, remind them of their previous drink, and suggest their next.

As the party got louder these interactions got more strained and performative. Arthur grew weary and the drinks grew inconsistent—drinkable, but only just.

In 2021 we followed up Arthur with “The Full Stack,” a barbot that promised to make faster, more consistent drinks using a conveyor belt system with udder-like pumps. The conveyor belt would allow for multiple concurrent drinks to be made, and the medical-grade pumps would serve ingredients in a more dimensionally accurate way. The bot used RFID tags in branded glasses to keep track of users’ previous drinks and display the data on a large LED “LITRE-board.”

Performance-wise, the robot was a success. It poured almost 500 perfect drinks in one evening (the only baddies being due to poor recipe choices). The LITRE-board functioned well but encouraged people to compete in a somewhat irresponsible way (the high score for the evening will not be disclosed here).

But of course: a host’s work is never done. We’re already thinking about future iterations where the bot asks patrons to rank their previous drinks and adjusts the base recipe according to the group’s average response and… as always, we’re committed to taking a far-out idea even further.