In Fall 2021, I joined Mezli, a startup building the world's first fully autonomous restaurant with no humans on site. During my time there, I spent countless hours thinking about trust. While we never fully implemented the ideas I'll share below, they represent what I believe is an important framework for the future of food automation.

Healthy, Affordable, and Convenient

The restaurant industry faces a rigid constraint: food can be healthy, affordable, or convenient—but never all three at once. This isn't a business challenge mathematical certainty born from the economics of traditional restaurants.

Want to serve healthy food? You'll need high-quality ingredients and skilled preparation. That means significant labor costs and expensive real estate for a full kitchen. To make the numbers work, you'll have to charge premium prices or move to inconvenient locations with lower rent. Want to make it affordable and convenient instead? That's what fast food chains did, but they achieved it by compromising on ingredients and nutrition.

Mezli was different than most food robotics startups. It didn't try to replace cooks with robots—that's been tried before and misses the fundamental problem. Instead, the startup aimed to completely redesign the restaurant model around automation from first principles.

The key insight was that by using robots to separate ingredient preparation from final assembly, a new model could be unlocked. Professional chefs could prep high-quality ingredients in a central kitchen, achieving economies of scale. Robots could handle the final assembly in small automated pods, eliminating the need for expensive restaurant real estate. These pods could go anywhere—office buildings, street corners, college campuses—making getting great, healthy food more accessible than ever before.

Building Trust inside a Black Box

But this solution created an unexpected challenge: trust.

When you walk into a traditional restaurant, you're immersed in trust signals. You see the kitchen, smell the food cooking, watch the chef at work. Your server makes recommendations and adjusts dishes to your preferences. These human interactions build confidence in the experience.

How do you get people to have the same trust when they are eating chicken from a black box they can't see inside?

The Three Models of Automation Trust

Through studying how people interact with other automated systems, I categorized trust into three main categories:

1. Empirical Trust

Think of CNC machines and 3D printers. Users build trust through direct measurement and verification. A machinist knows exactly what a good part looks like and can immediately spot issues. Trust accumulates through data points rather than abstract confidence.

2. Supervised Trust

Supervised trust describes systems like self-driving cars and robotic surgery assistants. Humans maintain oversight while delegating specific tasks. Trust develops through partnership. Users can see what the system is planning, override decisions, and gradually expand the autonomy they grant. The key is that humans remain actively engaged in monitoring and control.

3. Institutional Trust

Institutional trust covers medical devices, aircraft systems, and industrial equipment. Users may not understand the mechanisms, but trust transfers from rigorous testing protocols and regulatory oversight. You trust your pacemaker not because you understand how it works, but because you trust the institutions that validated it.

Why Food Automation is Different

Food automation breaks all these models in fascinating ways. Consider how empirical trust usually works: a machined part is either within tolerance or it isn't. The measurements are binary and objective. But food quality exists on multiple axes simultaneously—technical measurements like temperature and portion weight matter, but so do subjective elements like taste, texture, and presentation. You can nail every quantifiable metric and still end up with food that people don't want to eat. Nail the presentation, and people might think home-made pasta is Michellin star worthy.

Supervised trust models don't fit either. With self-driving cars, users can gradually build confidence by watching the system work, ready to take control if needed. But nobody wants to supervise their lunch being made—the whole point of a restaurant is convenience. There's no meaningful way to "override" automated food preparation. Once a bowl starts being assembled, it either comes out right or it doesn't.

Even institutional trust proves insufficient. Health inspections and safety certifications matter, of course. But passing inspection doesn't make people want to eat at your restaurant. Traditional restaurants build trust through human elements—the chef's reputation, the server's attention, the visible care in preparation. We had to find new ways to create these trust signals without relying on human interaction.

The Vending Machine Problem

This challenge was particularly complex because we were fighting deeply ingrained preconceptions about automated food service. People's primary reference point was vending machines, those sad boxes sitting in forgotten corners, filled with pre-packaged sandwiches of questionable age. Even high-end vending machines in Japan carry the stigma of being a last resort rather than a destination.

We needed to position ourselves clearly as a restaurant that happened to be automated, not a fancy vending machine. That meant competing on the metrics that matter for restaurants: food quality and consistency. Similar to self-driving cars, being just as good as humans wouldn't be enough. In order to succeed, we would need to use automation's capabilities to exceed traditional restaurants in other ways.

Radical Transparency as a Solution

My hypothesis was that radical transparency could bridge the trust gap. Most restaurants operate behind a curtain of opacity. You never know what portion of your bill goes to ingredients versus overhead. The true cost of meals is buried under tipping, service fees, and delivery markups.

Automation lets us show exactly what you're paying for. At traditional restaurants, only 25-30% of your money goes to ingredients. We spent 40% on premium ingredients because automation slashes overhead. Every price would show its breakdown: ingredients, labor, pod operations, delivery, and our margin.

First, this transparency would demonstrate how automation enables better food at competitive prices. Customers could see exactly how eliminating traditional restaurant overhead lets us invest more in quality ingredients. When supply chain costs decrease, they'd see those savings in their meals. When ingredient prices rise due to seasonal changes, we'd show the data behind price adjustments.

Second, it would transform transparency into a core brand value, similar to how Everlane revolutionized retail. By sharing costs, they made it so every product told a story about ethical manufacturing and fair pricing. We could do the same for restaurants, using automation to reveal every detail about food costs and quality that traditional restaurants keep hidden.

Traditional restaurants can't match this level of transparency. Their costs fluctuate daily, their operations are spread across locations, and their margins are too thin to expose. Automation gives us both the precision to track these metrics and the economics to share them openly. While other restaurants hide behind "market price" labels and vague cost increases, we could show customers exactly why each meal costs what it does.

The same technology that powers our kitchen would power our brand. Real-time displays would track ingredient costs and food safety metrics. Digital menus would adjust prices instantly based on supply chain changes. Every decision would be explained, every cost justified. In an industry built on opacity, radical transparency would become our defining advantage.

Looking Forward

Many of these ideas remained theoretical during my time at Mezli. The challenge of implementing radical transparency in the midst of engineering challenges and an impending launch proved more difficult than anticipated.

However, studying automation across industries revealed a clear pattern: transparency breeds trust. I believe that the next breakthrough in food automation won't come from incrementally better robots. Instead, it will come from reimagining the relationship between restaurants and customers.