For a while, the vending machines at the University of Waterloo dispensed Twix bars, M&Ms, and packs of Juicy Fruit, while surreptitiously using facial-recognition software to detect the age and gender of consumers. And they would have kept doing it if a malfunctioning machine hadn’t displayed the error message: “Invenda.Vending.FacialRecognition.App.exe.”
Students were shocked and upset that the snack machines were monitoring and profiling them without disclosure or consent.
It’s disturbing and it’s wrong — and not at all surprising. The use of artificial intelligence is expanding in the food-service industry. And we should expect more incidents like this until they become so common that we concede the issue.
McDonald’s has been testing out AI since 2019, using licence plates to identify repeat customers (with consent) and customize ordering suggestions at the drive-thru. More recently, Point Jupiter, a company that conducted a trial of face-recognition software within McDonald’s ordering kiosks, was able to scan users’ faces, recognizing their “gender, estimated age, and sentiment” in order to make meal recommendations. Retailers and arenas use facial recognition to identify repeat shoplifters or people on ban lists (Madison Square Garden has done this to flag lawyers from firms involved in litigation against the venue). At the speed that this technology is improving, this is just the beachhead.
Chipotle is now using AI-powered voice bots to take phone orders. Taco Bell is testing conversational AI at the drive-thru. DoorDash is offering voice AI for order-taking. And Opentable is using ChatGPT to help restaurateurs respond to reviews.
On a recent earnings call, the CEO of Wendy’s discussed plans for new digital menu boards to test dynamic pricing in addition to “AI-enabled menu changes and suggestive selling.” The announcement was widely reported as the introduction of "surge pricing" and sparked a public backlash. Wendy’s issued a statement distancing itself from surge pricing without denying its intentions to use AI-based dynamic pricing (the theoretical difference: while “surge” pricing increases in times of high demand, “dynamic” prices also go below the set price during periods of low demand).
I spoke with the founders of Dynpricing, a Canadian company creating restaurant AI software for dynamic pricing. They say that the price adjustments are in cents, not dollars. But as they test the technology to gather data on price elasticity through online food-delivery platforms, they are not informing consumers or asking consent.
These technologies are spreading fast, despite breaches of privacy and the biases baked into software. This is happening. Public outcry may adjust the speed and messaging of how these tools roll out. But we can’t stop them.
From chatbots to personalized menus, from Autocado (the avocado-slicing droid in use at Chipotle) to face-scanning candy-bar machines, these intelligent technologies are all deployed to increase revenue while reducing labour costs.
Last year, 14 per cent of restaurateurs in the United States said that they had invested in technology that replaced an employee in 2023. Much of the automation is in quick-service restaurants. In full-service restaurants, diners still expect to be greeted by humans. But in QSRs, efficiency rules. For these restaurants, revenue has grown over the past year, while for the rest of the dining industry it has shrunk. While all restaurants have benefitted from the automation of taking reservations, managing inventory, and processing orders, QSRs are able phase out whole job categories.
The purely physical versions of new restaurant tech — sweetly named restaurant robots that make tortilla chips (Chipotle’s Chippy) or French fries (White Castle’s Flippy) — are children’s toys compared to the AI-powered software in today’s order-taking kiosks. The technology inside these kiosks is able to adjust pricing, promote specials, and customize the menu’s presentation in real time. The goal is not merely to automate a previously human job, but also to improve performance by increasing the average bill size and ordering frequency.
So, back to Waterloo. Of course students were upset that vending machines were spying on them. I would be, too. Who, if asked by a fridge-sized metal rectangle filled with Snickers bars, would consent to have their face scanned or purchase history tracked? But most of those same students likely use app-based food ordering on their phones. And I expect those apps — from platforms like UberEats and DoorDash to individual brands like McDonald’s and Tim Hortons — will ultimately be successful trojan horses for gaining our consent. Very few people read the terms of service as they sign up for apps. It will be simple for the next software update or rewards program to add a single click that indicates consent to a new TOS, signing away our privacy.
In 2020, Cadillac Fairview, the real-estate company that operates some of Canada’s biggest malls, was criticized by privacy commissioners for using cameras inside information kiosks to secretly collect millions of images for biometric purposes.
A 2021 Swiss study found that consumers were more ready to share their personal data when they understood how it would be used by a company. But a few years later, that attitude feels increasingly quaint. If the restaurant app companies successfully frame the issue as a benefit to consumers, they’ll succeed where the vending machines got caught with their robot mandibles in the cookie jar. They’re not asking to invade our privacy. They’re offering to optimize our experience!
There won’t be a referendum on this issue. Our governments are always streets behind the speed of the tech industry. Within a year, once enough of us have clicked away our expectation of privacy in exchange for 15 per cent off on our next order, this conversation may be over.
Donate