Ninety percent of the world’s largest quick service restaurants run on AWS. That stat alone makes Deborah Matteliano one of the most uniquely positioned people in the restaurant industry to talk about what AI actually looks like at scale. As Global Head of Restaurants and Food Tech at AWS, Deborah sits across from the executives at the world’s biggest QSR brands every week and sees firsthand what is in pilot, what is actually in production, and what is still mostly a press release. In this episode, she joins Matt Wampler to talk about the state of restaurant AI, the real reason most brands are struggling to build it, and where the industry is headed in the next two to three years.
What You’ll Hear in This Episode
Why the Restaurant Industry’s Data Problem Is Not What Most People Think The common assumption is that restaurants don’t have enough data. Deborah pushes back on that directly. The data is there. What’s missing is a connector. Since the pandemic, the average restaurant technology spend has gone up by 70%, leaving most brands with a growing pile of disconnected systems that can’t talk to each other. She uses a vivid image to describe it: a power strip where everything is plugged in but nothing is actually connecting. Until the industry commits to data standards, she argues, the brain of the restaurant will always be working with a broken nervous system.
The Two Stats Dividing Restaurant Boardrooms Right Now Deborah opens every executive briefing with two numbers. The first is that 71% of consumers now want agentic AI in their shopping or eating experience. The second is that 87% of restaurants have seen positive impact from AI implementation. She explains why these two stats reliably split a room, with half the table asking what exactly counts as AI and the other half questioning what consumer demand for agentic AI even means in a restaurant context. Both debates, she says, are worth having.
The GM Assistant: Why Every Restaurant Executive Wants One and Why It Is So Hard to Build The number one thing restaurant executives tell Deborah they want to implement is a digital general manager assistant. A system smart enough to synthesize all the fragmented data inside a restaurant and surface the right decision at the right moment. She breaks down why that goal is so difficult to achieve when most brands are running four different POS systems across their enterprise, and why Yum Brands’ decision to build their own tech stack from the ground up with Bite by Yum is the clearest example of what owning your own nervous system actually looks like in practice.
What Amazon Prime Teaches Restaurants About Supply Chain Intelligence Deborah walks through one of the more surprising analogies in the episode: the internal Amazon order fulfillment system, which calculates up to 64,000 different permutations to determine the best way to deliver a single package to a customer. Everything is built backward from one question, which is what did you promise the customer and how do we make sure that promise is kept. She explains how that same supply chain logic applies directly to restaurant inventory, ordering, and kitchen production.
Why Bad Sales Forecasting Is the Root Cause of Most Restaurant Problems One of the sharpest moments in the conversation comes when Deborah describes a restaurant brand that redesigned 300 locations with single lane drive-throughs, only to realize later they should have built double lanes. The question she asks them is whether the real problem is the drive-through design or the faulty sales forecast that caused the expansion decision in the first place. Her answer is that a surprising number of the operational and capital problems restaurants face trace back to forecasting that was never accurate enough to make good decisions from.
Agentic AI: What the Next Two to Three Years Actually Look Like When Matt asks Deborah what restaurant executives should be expecting from AI in the near future, she points to agentic AI as the most promising development. Not individual point solutions, but fully orchestrated chains of agents working together. A negotiation agent that reaches out to suppliers for better pricing. An inventory engine feeding into a GM assistant. A GM assistant feeding into a frontline employee bot that helps the team own the day. She ends with a simple framing: whoever builds the best brain will win.
Why Restaurants Should Be Experimenting Now and Not in Two Years Deborah is direct about the cost of waiting. The experiments happening today are generating the data and learnings that will shape strategy for years. She points to a partnership Amazon recently launched with TaskRabbit through Alexa Plus as an example of starting something without knowing exactly how it will go, and learning from it as it runs. Her advice to restaurant operators is the same: fail forward, because the brands that are not experimenting right now will be drinking from a fire hose when they finally start.
Key Topics Covered
- Why 90% of the world’s largest QSRs run on AWS
- The data connectivity problem behind restaurant AI adoption
- Agentic AI and what consumers actually want from it
- The digital GM assistant and why it is harder to build than it sounds
- Bite by Yum as a case study in owning your own tech stack
- Amazon Prime supply chain logic applied to restaurant operations
- Why bad sales forecasting is the hidden root cause of major operational problems
- Computer vision in restaurants: where the ROI is real and where it is not
- Load balancing and why it matters for multi-unit restaurant scale
- How to start small with AI and fail forward
Subscribe to the Restaurant AI Podcast on YouTube or Spotify
Follow the Restaurant AI Podcast on LinkedIn
Connect with Matt Wampler and ClearCOGS on LinkedIn
