AI in the restaurant industry is brimming with potential, but unlocking it can be difficult.
“When we think about AI today, in many ways it’s a bit like the yellow brick road,” said Carl Orsbourn, SVP of food, hospitality and retail at Invisible Technologies. “We want to see the wizard, right? And there’s this promised land of generative and agentic AI that we are all super excited about, that we need in our businesses…but how?”
“Many businesses out there are struggling with what we call the chasm,” he said while moderating a panel at the eighth annual Food On Demand conference. Orsbourn, who also co-authored Delivering the Digital Restaurant and The Path to Digital Maturity, tracked the growth of off-premises dining over recent years. “In many ways now, we’re at another precipice, the world of AI and how that is impacting restaurants.”
“Eighty-five percent of the projects around generative AI today that are put into production never make it out of there. They never succeed. The challenge has been that there are so many point solutions.” With thousands of AI startups vying for attention, “the tech stack today looks very much like this rather convoluted tech Lego block, made from all these different pieces and models,” he continued.
This can result in disjointed systems and bad data. “When you have bad data meet good AI in a fight, the data always wins,” Orsbourn said.
Orsbourn was joined on stage by Deborah Matteliano, global head of restaurants and food tech at Amazon Web Services (AWS), Shiv Adhiappan, VP of global reliability & engineering operations at Yum! Brands, and Rich Faltot, sr. client partner and restaurant segment leader at Point B.
Carl Orsbourn, SVP of food, hospitality and retail at Invisible Technologies, moderated the panel at FODC.
Maturity in the market
AWS works with over 10,000 of the largest restaurant brands in the world. Matteliano noted that many are still in proof-of-concept stages with AI. “It’s all about the level of maturity. We are in excitement versus execution, and we are looking to bridge that gap,” she said.
“If you imagine a maturity model from level 1 to 5,” added Faltot, “with 1 being basic awareness and 5 being full transformative AI, most of the industry sits between a 2 and a 3.”
The question heard most is: Where do we start with AI? “You may not have a team of data scientists and developers, so then it becomes, maybe there’s tools that are already out there that we can bolt onto our existing stack to achieve those outcomes,” Faltot continued.
The data hurdle
“We walk into situations and see data that’s oftentimes a mess… we work through the process of helping a brand clean up and organize their data,” said Faltot. He also said to be aware of biases in the data. “Spend the largest part of your time preparing the data first, to make sure that the outcomes from AI models are going to help you in the way you want them to.”
Matteliano agreed. “Many don’t realize that in order for AI to be smart, you have to collect data from disparate sources into one unified infrastructure… I like to think about it as sort of a brain. What’s complicated and tricky about restaurants is there are so many different inputs: POS data, menu data, drive-thru AI data, loyalty program data, app data—the list goes on.”
“We’re data rich, insight poor. The problem isn’t, do we need more data? It’s how we extract value from the data we have, by making it talk to each other,” she said.
Deborah Matteliano, global head of restaurants and food tech at AWS, shared how she helps restaurant brands adopt AI.
Orsbourn asked how restaurants are using decision intelligence and agentic flows.
“Productivity is just one piece of it,” Matteliano said. “AI can help with inventory, supply chain, personalization, there are many use cases. But if you only look at labor or POS data, you miss other layers, like weather patterns affecting demand. That’s how you drive real efficiency.”
“AI works best when systems come together—like Byte’s AI GM assistant at Taco Bell. It’s like having a smart brain acting as the restaurant’s COO,” Matteliano said.
AI in action at Yum
Yum! Brands, the parent company of Taco Bell, KFC, and Pizza Hut, has built its own proprietary suite of AI tools under the banner Byte by Yum. This helps manage everything from mobile ordering and payments to kitchen operations, delivery logistics, inventory, and labor.
One standout tool is Yum’s restaurant coach app. “We’ve unlocked Gen AI capabilities so team members can access operational knowledge right from their phones,” Adhiappan said. “It’s like having a smart Einstein in your pocket.”
Given the number of locations under Yum’s umbrella, Adhiappan said it’s important to have tech solutions that can scale across different franchises. Features introduced for one brand can be easily deployed to others.
The long-term benefit? “If your data is powerful enough, you may not need to spend millions on traditional marketing. You can drive customer engagement through simple push notifications or personalized emails,” Adhiappan said.
Don’t chase buzzwords
Throughout the session, panelists reiterated that before adopting any AI solution, make sure you clearly understand the problem you’re trying to solve—don’t just chase the latest buzz.
“We had a brand that really wanted to create a Gen AI chatbot for employees,” Matteliano shared. “The biggest thing to solve was a lot of quality problems with their limited-time offers. It was meant to be there so employees could ask things like, ‘Hey, how much pepperoni goes on this item?’”
“Well, that’s not without a lot of investment,” she continued. “At the end of the day, we counsel this customer on, is your real problem solved by a chatbot? Or is it solved by a better training manual or perhaps LTOs that don’t involve 19 decision-making processes to put it into the wrapper and out the door? Can we actually use AI in a predictive way to help you identify the next viral item, like the Chili’s mozzarella stick, that’s operationally easy to execute, easy to produce at scale… So at the end of the day, what does the customer want?”
Real-world ROI
“Shaving off one second per drive-thru transaction can save a million dollars,” Matteliano said. “Taco Bell’s voice AI saved three seconds.”
She also pointed to another win: “The AI-powered contact center we built for DoorDash helps drivers get quick answers like ‘I picked up the wrong item’ or ‘I’m lost.’ Instead of waiting for a human agent, the AI responds instantly.”
Faltot shared, “We’re working with a burger chain that has over 1,000 locations that came to us and said, hey, we need a way to get our inventory from our back of house in our system faster.” With AWS, Point B built a tool that lets operators quickly inventory stock. What used to take hours now takes minutes.
Orsbourn encouraged operators to think practically. “The question you should ask is what if you had 100 extra marketeers, analysts, engineers, what would you put them to work on?” he said.
The Food On Demand Conference wrapped up May 7 at the Bellagio in Las Vegas.
