In the nearly three years since McDonald's announced it was partnering with IBM to develop drive-thru order-taking machines equipped with artificial intelligence, videos have appeared on social media of confused and frustrated customers trying to correct hilariously incorrect meal decisions.
“Stop! Stop! Stop!” two friends yelled with humorous anguish in a TikTok video after the AI drive-thru machine misunderstood their order and counted 240, then 250, then 260 Chicken McNuggets.
In other videos, the AI checks out a customer who ordered nine iced teas instead of one, can't explain why one customer couldn't order a Mountain Dew, and can't explain why another customer wanted to add bacon to their ice cream.
So it probably didn't come as a shock to customers when McDonald's announced in an internal email dated June 13 and obtained by industry publication Restaurant Business that it was ending its partnership with IBM and halting AI testing at more than 100 of its U.S. drive-thru restaurants.
The decision to abandon the IBM deal comes as many other companies, including competitors, are investing in AI, illustrating some of the challenges companies face as they race to unlock the potential of revolutionary technology.
Other fast-food companies have had success with AI ordering: Last year, Wendy's partnered with Google Cloud to build an AI drive-thru system, Carl's Jr. and Taco John's signed on with restaurant voice AI company Presto, and Panda Express has partnered with voice AI company SoundHound AI to install about 30 automated ordering kiosks at its windows.
White Castle, another SoundHound partner, has AI assistants taking orders at 15 drive-thrus and plans to roll out to 100 more, spokespeople for both companies said. White Castle vice president Jamie Richardson said the technology processes nearly 90% of orders without human intervention, working more efficiently with staff and reducing customer wait times during rush hours.
“It's great for our customers and it's great for our team members,” he told The New York Times. “I can't speculate why other companies don't invest in similar technology, but we're really happy with it.”
Keyvan Mohajer, CEO and co-founder of SoundHound, believes McDonald's withdrawal is simply another example of a partnership gone wrong.
“They're clearly abandoning IBM, but they're not abandoning voice AI,” he said. “They're going after other vendors very quickly.”
McDonald's confirmed its intention to eventually return to the technology, writing in an internal email that a “voice ordering solution” is in the chain's future.
IBM said in a statement that it looks forward to continuing to work with McDonald's, adding that it is in “discussions and pilots” with several restaurants interested in building the automated ordering technology. McDonald's confirmed to The Times that it is shutting down its AI drive-thru, but neither company responded to more specific questions.
Some industry researchers and experts see McDonald's withdrawal as an example of how new technology has yet to live up to expectations, and they question whether the company will return to testing AI ordering at its drive-thrus anytime soon.
“AI systems often have very large upfront costs,” said Neil Thompson, director of FutureTech, a research project at MIT's Computer Science and Artificial Intelligence Laboratory. (FutureTech is partnering with IBM, but Thompson said he didn't have inside information about the McDonald's deal.)
Thompson said that currently voice AI is often inaccurate, requiring some level of human oversight, reducing the cost savings, and McDonald's has a strong, higher-margin alternative: its mobile app.
“The app reduces the labor required to take an order by 100 percent in a way that an AI system, at least right now, can't,” Thompson says, “so the economics of using the app are much more attractive than using AI.”
McDonald's isn't abandoning its AI investments altogether: The company announced a partnership with Google Cloud in December 2023. A spokesperson for the tech giant said the AI would be applied to “business use cases” but declined to provide further details.
Alex Imus, a professor of behavioral science and economics at the University of Chicago, predicted McDonald's will likely sit on the sidelines as competitors explore the technology.
McDonald's business model isn't based on saving on the labor costs of a few drive-thru employees, Imus said. “I think they're going to wait until this machine is ready for commercial use.”
He expects to see McDonald's use AI in other ways, following the example of Target, which recently announced it's using the technology to help employees.
Presto's interim CEO, G Lefebvre, acknowledged that the technology is very new, saying that “fewer than 0.5% of U.S. drive-thru locations” are testing the use of AI to take voice orders.
But he also noted that many of the early attempts have been successful.
Wendy's said in an email to The Times that its AI drive-thru processes 86% of orders without human assistance, while Presto processes about 90% of orders without human assistance for the majority of its customers, Lefebvre said.
He believes McDonald's struggled because it was using the wrong type of AI.
“IBM's model was still based on natural language understanding,” Lefebvre said, explaining that the model works like a tree: When the AI hears a customer order, it has a limited number of branches that dictate its responses and actions.
Lefebvre said this approach works great when everything is going well, but in drive-thrus where indecisive customers frequently change their orders, he said the chain would be better off using large-scale language models to power chatbots like ChatGPT.
As companies continue to test AI drive-thru technology, we can expect to see more videos of people trading in food for bacon ice cream, condiments, or enough nuggets to feed a sports team.
But ask Mohajer about the future of voice AI and you'll understand why SoundHound has partnered with car companies like Kia and Jeep.
Imagine this:
On your way home from work, your car suddenly asks you, “Are you hungry?”
After a few minutes of conversation with your car, you decide to order a burger, fries, and a shake. The car finds the nearest fast food restaurant, places your order, and enters directions. Three minutes later, you pull up and find your dinner waiting patiently in the pickup lane.
