Summary

•McDonald's multiyear AI drive-thru partnership with IBM ended in 2024 after viral failures. Taco Bell's voice AI at 500+ locations is now officially a "sometimes tool."

•Kitchen robots have a narrow, genuine win: high-volume, single-task stations like Sweetgreen's Infinite Kitchen and Miso's Flippy fry station show real ROI data.

•The places where tech has failed are the same: technology deployed to solve a broad problem it wasn't designed for.

•Just 9% of operators plan to invest in restaurant robots in 2025, per the National Restaurant Association.

•Every piece of automation is a piece of equipment. It breaks. It needs service. That part of the story rarely makes the press release.

One day, you might order a burger from a voice AI, watch a robotic arm serve fries, and have another robot deliver your meal to your table. Tech companies have promoted this vision for nearly a decade. Industry magazines reported on it, and restaurant chains released statements using terms like transformative and revolutionary.

The actual results have been considerably less transformative.

What’s happened with big tech’s push into foodservice kitchens is this: a few narrow wins in specific, controlled conditions, and a string of expensive, embarrassing stumbles elsewhere.¹

The Failure That Started Created Doubt

McDonald’s invested several years working with IBM to pilot voice AI technology at its drive-thrus. in July 2024, the company decided to discontinue this partnership.² The decision was prompted by a series of widely publicized incidents, like AI erroneously generating an order for 260 chicken McNuggets and attempting to add bacon to ice cream. These instances were recorded and shared by customers, ultimately leading McDonald’s to terminate the program.

The company has since moved to Google Cloud for its AI ambitions, but the IBM episode exposed the complexities of deploying it into the “real” world.³

Taco Bell arrived at roughly the same conclusion through a slightly longer route. The chain deployed voice AI across more than 500 drive-thru locations, confident that two years of testing had prepared it for scale.⁴ By fall 2025, the company was having second thoughts. One customer ordered 18,000 cups of water to test the system’s limits. Others documented the AI’s inability to understand a simple large Mountain Dew order. The chain’s own chief digital officer acknowledged he had mixed personal experiences with the technology as well. Taco Bell now advises franchise teams to treat voice AI as a “sometimes tool,” best paired with human staff during peak hours.⁵

The Pattern in the Failures

It’s not that technology doesn’t work. It would be more accurate to say it doesn’t work the way they’ve been trying to sell it to us.

The restaurant environment is very challenging for precision systems. According to Intouch Insight’s 2025 Emerging Experiences study, 22% of AI drive-thru orders still require human intervention.⁶ That’s roughly one in five customers needing a person to fix what the machine couldn’t.

The same pattern shows up in kitchen robotics. Kernel, the robot-forward restaurant concept launched by Chipotle founder Steve Ells, generated enormous press coverage when it opened in New York. A custom-built robot arm handled food movement and oven operations.

The restaurant closed within a year.

Chili’s tested robot servers at several locations. They eventually dropped the program. Sweetgreen sold its Spyce robotic division to Wonder for $186 million in 2023, a notable exit for a technology Sweetgreen itself had built.⁷

Where It Actually Works

To be fair to the techies: there have been real wins.

Sweetgreen’s Infinite Kitchen one of the success stories in the segment. The automated Infinite Kitchen system assembles up to 500 salads or grain bowls per hour, needs fewer staff, increases average check size, and improves employee retention compared to traditional locations.⁸ The CEO of Sweetgreen describes it as central to creating a more profitable and scalable model.

Miso Robotics’ Flippy fry station tells a similarly complicated story. The marketing claims are aggressive ($75,000 in annual new profits per location, 89% reduction in fry station labor costs), and they come from Miso’s own promotional materials.⁹ But White Castle’s regional operations director has publicly confirmed expanding Flippy to multiple St. Louis locations, and employee feedback at those sites is genuinely positive: 87% less walking, fewer burns, less physical exhaustion. The company’s financials are a different story entirely. Miso generated roughly $385,000 in net revenue in 2024, down from $493,000 the year prior, with only 14 active Flippy units as of late 2025, down from 17 two years earlier.¹⁰

Wendy’s FreshAI drive-thru system has delivered the most reliable financial results among voice AI deployments. Its AI-powered upselling helped improve margins by 80 basis points at company locations. Wendy’s is expanding the system from 100 to 500–600 sites in 2025, with a capital spend of $100–$110 million. That’s not a feasible investment for many operators.

Chipotle’s Autocado handles the tedious, dangerous work of cutting, coring, and peeling avocados in 26 seconds per fruit, roughly half the time it takes a human worker.¹² The chain uses 5.2 million cases of avocados annually. The math on labor savings is real. But Chipotle is testing it in two California locations. The jump from a test kitchen to 3,500 restaurants has yet to be proven.

The Diagnosis

All this technology can work when it solves one specific, repetitive, high-volume problem in a controlled environment. Frying consistent batches of the same item. Portioning identical ingredients into identical bowls. Upselling a customer on a digital menu.

Tech fails when it’s asked to replace human judgment in noisy, variable, customer-facing situations. Taking a drive-thru order from someone who changes their mind mid-sentence, has a regional accent, and is ordering for five people in a car with the radio on.

A major problem is that restaurants rolled out technology according to investor demands rather than their own operational readiness. Many chains publicized AI initiatives because they attracted media attention and pleased shareholders, but real performance data only came later—and received far less publicity. Only 9% of operators plan to invest in restaurant robots in 2025, according to the National Restaurant Association

Will Big Tech Take Over Foodservice?

So far, the technology has a mixed record. This means that if you have felt a lot of pressure to adopt AI or robotic solutions, you can relax. The old ways still work.

There’s one more dimension to this: every piece of automation is a piece of equipment. It has motors. It has sensors. It has software that needs updating and hardware that wears out. It breaks.

A fry robot that goes down during a Friday lunch rush is not a labor solution. A voice AI system that fails mid-service forces a human worker to step in and apologize to the customer it just confused. The uptime assumptions baked into most automation ROI projections are optimistic, and the service infrastructure for novel equipment is still catching up to the deployment curve.

Big Tech has been very good at announcing the future of foodservice. It has been considerably less good at maintaining it.

Sources

Big Tech Has Been Trying To Take Over Foodservice…. How’s That Going?

Summary

The Failure That Started Created Doubt

The Pattern in the Failures

Where It Actually Works

The Diagnosis

Will Big Tech Take Over Foodservice?

Recent Posts

Recent Comments

Archives

Categories

Meta