A Robot’s Nightmare Is a Burrito Full of Guac
Welcome to the future: A robot can now prepare your favorite Chipotle order. Just as long as you don’t want a burrito, taco, or quesadilla. The robot cannot handle those. Your order must be a burrito bowl or a salad, and it must be placed online. Then and only then—and once the robot makes it out of testing at the Chipotle Cultivate Center, in Irvine, California—your queso-covered barbacoa bowl might soon be assembled by the chain’s new “automated digital makeline.”
Announced on Tuesday, the result of a collaboration between Chipotle and the automation company Hyphen looks like a standard stainless-steel Chipotle counter, burrito components arrayed on top. But inside, just above knee level, is a robotic assembly line that can prepare to-go bowls and salads from start to finish. A video from Chipotle shows a bowl pivoting through the machine, positioning itself below specified ingredients. White rice tumbles in, some grains scattering about. Later, a cascade of corn. At the end, a bowl ascends from the machine, complete, as an employee folds a burrito and wraps it in foil. Perfect synergy. She smiles widely.
Her smile might come from the knowledge that, at least for now, a robot is not going to put her out of a job. Fast-food work, often seen as simple and repetitive, has long seemed especially at risk of automation. A viral 2013 study from researchers at the University of Oxford put the probability of robots disrupting restaurant work at 92 percent. But even as chatbots are everywhere and self-driving cars prowl American streets, robots are still not yet capable of replacing a fast-food employee who serves as a cashier, preps ingredients, assembles orders, and closes up shop. Right now they don’t even fold burritos.
Chipotle’s bowl-bot isn’t the company’s first somewhat mundane food-prepping invention. Last year, it unveiled Chippy, “an autonomous kitchen assistant that integrates culinary tradition with artificial intelligence to”—wait for it—“make tortilla chips.” In July, the company announced the Autocado, an “avocado processing cobotic prototype” that cuts, cores, and peels avocados, eventually reducing guacamole prep time by 50 percent. The gigantic bowl of machine-prepared avocado flesh then gets mashed by a human.
The rest of the restaurant industry is in virtually the same uninspiring place. Flippy 2, a device from the start-up Miso Robotics that dunks frozen french fries in hot oil and dumps them on a tray, is now in 100 White Castle locations. McDonald’s much-hyped “fully automated” Texas location still has workers making food—just not taking orders and handing them to customers. Perhaps the most fully automated restaurant is Sweetgreen’s “Infinite Kitchen,” which just drops salad ingredients into a bowl, adds dressing, and then shakes it up. This is even as restaurant chains become some of the most technologically advanced businesses in the country. Sweetgreen is testing drone delivery. Chipotle is piloting a system that monitors ingredient levels and tells employees when to prep and cook more food, and how much. An intrepid Domino’s delivery person will find me, pizza in hand, wherever I drop a pin in the Domino’s app.
[Read: Silicon Valley is beating Domino’s at its own game]
That cooking robots are usually sufficient at a task or two is part of what makes them so tough for restaurants, David Henkes, a food-industry analyst at Technomic, told me. “It’s hard to justify rolling out a huge piece of technology [just] to solve the fry problem,” he said, referring to White Castle’s Flippy. Even if a device succeeds at a fairly simple chore, the location must adapt their production around it. That is, Henkes said, if it can even fit in the existing kitchen. Though Chipotle’s new bowl-assembly station is designed to mirror the existing low-tech countertop, devices like the Autocado might be harder to accommodate in a cramped setting. Like the automated production line, it has not yet graduated from the Chipotle Cultivate Center; a spokesperson for the company did not specify when the two robots might make their restaurant debut.
Chipotle is investing so much in these robots for a reason, of course. Even if machines don’t fully take over the chain’s kitchens, automating more parts of the process would save the company lots of money and produce a burrito bowl that is more consistent from store to store. Preparing bowls and salads “can be a repetitive task with minimum human interaction,” Curt Garner, Chipotle’s chief customer and technology officer, explained in an email. “It doesn’t have the same art as guac prep or rolling a burrito.”
But devotion to the art of manual burrito folding is likely not the sole reason Chipotle has kept its wrapping process. Though Garner said the machine is technically capable of folding a burrito, it’s unable to perform the final flourish: wrapping it in foil. Still, it’s unlikely that its burrito folding is ready for the barbarians who request double chicken, double rice, and double beans. The problem is foundational: Making custom burritos is actually “very, very difficult for a robot,” Dmitry Berenson, an associate professor of robotics at the University of Michigan, told me. Tactile sensing for robots is still “in its infancy,” he said, which makes it hard to avoid overstretching tortillas or crushing what’s inside. Robots also lack sufficient algorithms to predict how objects will deform when manipulated, Berenson pointed out. For even the most high-tech robots, globs of guacamole are overwhelming. “Overcoming these barriers,” he said, “is going to require a lot more fundamental research.”
[Read: AI is running circles around robotics]
Change will likely come slowly. Chatbots can learn from huge amounts of online text to improve their own responses, says the Stanford computer-science professor Chelsea Finn, but “we don’t have analogous data for motor control on the internet.” There’s no Wikipedia page that describes, in granular physical detail, how to mash an avocado. And in a restaurant, these seemingly minor tactile predicaments abound. A machine specially designed to flip hamburger patties can’t unpack them, place them on a grill, add cheese, and put them on a bun. If a robot takes over the fry machine, any malfunction—without a human present—could lead to a huge loss in sales. Robotics are good at “garnering headlines,” McDonald’s CEO said last year, but “it’s not practical in the vast majority of restaurants. The economics don’t pencil out.”
That isn’t going to stop fast-food restaurants from trying. An industry that has already made technology an inextricable part of its business model is determined to make kitchen robots happen. Virtually every fast-food company is now testing AI-powered drive-through; this week, Domino’s and Microsoft announced a partnership to bring more AI into the pizza-ordering process. But for now, the future of fast food looks similar to that of so many other industries: Instead of replacing workers with robots, the machines work alongside them. “It’s still very much a labor-intensive, people-oriented business,” Henkes said. “And that’s changing at the margins, but it’s not going to change dramatically anytime soon.” A Chipotle robot can provide me with automated, real-time updates on the status of my tacos. When it comes to making them, though, that’s still up to a human.