Why In-N-Out Pays Managers $160k to Throw Food in the Trash

If you live on the West Coast, you know the sight.

It’s 11:00 AM in Burbank near where the Studios are. The drive-thru line is already 15 cars deep, snaking around the building and spilling out into the street to block the right lane of traffic.

You walk inside. The air hits you - toasted buns, grilled onions, and the frantic energy of the lunch rush.

To the customer, it looks like a synchronized dance. The crisp white uniforms. The paper hats. The “Animal Style” shouts. It feels nostalgic. It feels simple.

But step behind the counter, and the illusion of simplicity vanishes.

This is not a retro diner; it is more of a biological battlefield, as my AI research friend tells me. The deep fryers drift in temperature. The potatoes change chemical composition by the hour. And the entire system relies on “servers” who are essentially teenagers thinking about prom.

The Store Manager stands at the “Final Window” and spots a Double-Double. The lettuce is crisp, the bun is toasted, but the cheese fold is off by a fraction of an inch. It is edible. It is warm. It is 95% perfect.

The Manager doesn’t fix it. They trash it. They stop the line.

In 2018, it was revealed that the average In-N-Out store manager earned $160,000+ - triple the industry average. Why?

Because in a chaotic system, the person who refuses to ship “mostly okay” is the only asset that matters. The market pays a massive premium for the only person in the building who cares more about consistency than throughput.

And the reality I’ve been thinking about lately: We are all In-N-Out managers now.

The Shift

Why does this resonate so deeply? Because for the last twenty years, we didn’t have to worry about the “cheese fold.” We built with “steel.”

Traditional software engineering is Applied Physics. It is deterministic. You write a function. It executes. It does not get tired. It does not “hallucinate.” We grew accustomed to being Architects - designing structures that stood still.

But Generative AI has forced us into the kitchen. We have traded our steel for “biology.”

An LLM isn’t a calculator, it feels more like a biological system. Like that 19-year-old line cook, it has “moods” (temperature). It gets “confused” (drift). It “lies” to save face (hallucination).

The mistake I see a lot of best leaders making right now is trying to manage this biological chaos using an architect’s blueprint. We are trying to code our way out of a problem that can only be operated.

The friction comes from a simple mismatch: we are acting like Architects, but the job now demands a Shift Manager. We need to trade the safety of the blueprint for the heat of the kitchen and learn the sweaty, unglamorous art of Operational Rigor.

Here are three learnings from In-N-Out managers that we can apply directly to our AI strategy.

1. The “No-Avocado” Rule

In-N-Out managers face constant pressure to add items. Customers beg for avocado. They want bacon. They want chicken. It would be profitable. Yet, the answer is always No.

Why? Because avocado oxidizes. It turns brown in minutes. By introducing one new ingredient, you introduce a variable that decays faster than the meat. You create a “weak link” in the supply chain that threatens the consistency of the entire system.

In the AI world, we are currently making the opposite mistake. We are obsessed with volume, not value. We launch 100 different “menu items” - a Legal Bot, a Marketing Agent, a Coding Assistant - hoping one of them tastes good.

The result isn’t business impact; it’s a menu of shiny AI powered apps no one uses. We create a bloated portfolio of tools that look impressive on a roadmap but deliver zero sustainable business impact. We are scaling complexity before we have established value.

Earn the Expansion. Think twice about impact before you expanding the portfolio. Focus on the “Cheeseburger” - that single, boring use case that solves a real business problem with 100% reliability. You don’t earn the right to add the second item until the first one is actually feeding the business.

2. The “Sugar” Test

The potato is a liar.

It looks consistent on the outside, but its chemistry changes with the weather. A potato that looks perfect might have high sugar content. If you fry it, it burns.

So, the Manager performs a “Fry Test” every morning. They don’t trust the supplier’s rating. They slice a batch, fry it, and see how it reacts in their oil, at their temperature. They verify the ingredient works for their specific standard before serving it.

We are drowning in options. We have proprietary models, open-source weights, and home-grown fine-tunes. We look at public leaderboards and benchmarks to decide which one to use.

But public leaderboards are like supplier ratings - they tell you how the potato performs on average, not how it performs in your kitchen. No one knows your use cases better than you. A model that aces a generic benchmark might completely fail on your proprietary data.

Run Your Own Fry Test. Build your own internal evaluation set based on your specific use cases. Pre-test every model - whether it’s GPT-5 or a local Llama - against your data. Does it burn? Is it consistent? You can’t pick the right “potato” for the job if you haven’t tasted it first.

3. The Expensive Trash Can

The Manager and the Cook are natural adversaries. The line Cook’s incentive is speed. The Manager’s incentive is perfection.

The most critical tool in the store is the Trash Can.

To an outsider, the trash can looks like failure. It looks like lost money. Why throw away a burger that is 95% perfect? But to the Manager, the trash can is an investment. The cost of the wasted ingredients is the “tuition” they pay to ensure the customer never sees a mistake.

Now when building with AI, we are allergic to the trash can. We treat “discarded compute” as inefficiency. We want every token we pay for to reach the user. This is wrong. In a probabilistic system, “waste” is the only way to guarantee quality. You might need to generate three draft emails to find the one that is safe to send.

Budget for the Trash Can. Forget zero shot a perfect app. Build systems that generate, critique, and discard in the background. If you aren’t paying for “wasted” tokens - if your trash can is empty - your quality standards are just a hallucination.

The Apron Over the Hoodie

We spend so much time looking for the “10x Engineer” or the “10x Researcher”—the geniuses who invent the recipes. But the In-N-Out story teaches us that the real value isn’t in the recipe. It’s in the discipline.

The next era of AI won’t be defined by who has the smartest model. It will be defined by who has the most disciplined kitchen.

So, here is my question to you:

Look at your current AI projects. Are you acting like a Chef trying to invent a new dish, or a Manager trying to stop a bad burger from leaving the window?

I’d love to hear your thoughts.

#Leadership #AI #OperationalExcellence #Management #InNOut #FutureOfWork #Strategy