In the book, I discuss how Amazon’s internal systems around work-life balance went sideways. Here, is an example of how another Amazon system built a clean decision infrastructure around HR authority, and how that system can work exactly as intended.
The headlines focused on what the managers said. They asked an employee whether he was dating a co-worker. They asked about his pronouns. Twice.
Those are the kinds of questions that generate lawsuits, and yet Amazon won.
Not because the behavior was acceptable, but because the system around it was designed correctly.
Here is what actually happened.
The employee was an Onsite Medical Representative at Amazon’s facility in Windsor, Connecticut. He had a negative balance under Amazon’s attendance policy. HR sent written warnings. The balance stayed negative. Centralized HR and safety leadership made the termination decision, not the supervisors who asked the awkward questions.
One of those supervisors actually recommended against termination.
The court looked at the record and found something straightforward: the people who made the decision were not the people with the problem. There was no evidence connecting the supervisor’s conduct to the outcome. The attendance policy was documented. The warnings were documented. The decision process was documented.
Most leaders, when they read a story like this, focus on the behavior.
Train your managers. Watch what they say. Sensitivity matters.
All of that is true, and make no mistake, the fact there was a court case is a loss, but it misses an important lesson. The behavior created the risk. The system controlled it.
Think about what Amazon’s infrastructure actually did in this case.
The attendance system tracked policy violations automatically, independent of any manager’s judgment or mood. The decision-making authority was centralized, which meant the person closest to the problem was not the person who decided the outcome. The documentation existed before any conflict arose, not constructed afterward to justify a decision.
None of that is sophisticated technology. It is organizational design.
It is the kind of infrastructure most companies think they have but have not actually built.
There is a pattern that appears repeatedly in organizations that end up in legal, financial, or reputational trouble.
The people at the top believed the stated policy was the operating policy.
It was not.
A policy written in a handbook and a policy embedded in a system are two different things. One describes what should happen. The other determines what actually happens.
Amazon’s attendance policy did not just exist on paper. It was tracked automatically, applied consistently, and generated its own documentation. The system did not rely on managers to remember, follow through, or behave well.
That is the difference. There is also something worth noting about decision routing.
When a manager has a conflict with an employee, even an unconscious one, even a barely-perceptible one, that manager should not be the person who decides what happens next.
Not because they are necessarily biased, but because organizations should not depend on the absence of bias to produce fair outcomes.
Centralized HR authority is not just a compliance mechanism. It is a circuit breaker. It routes decisions through people who do not have the same relationship to the facts, which means the outcome is less likely to reflect the wrong things.
In Amazon’s case, that design worked exactly as intended. The supervisor with the questionable conduct was downstream of the decision.
The uncomfortable question for most organizations is this: If something similar happened here tomorrow, what would the record show?
Not what policy says. Not what you believe happened. What the actual record would show: the timestamps, the warnings, the decision trail, who decided and when.
Most leaders do not know the answer with confidence.
That gap is where risk lives.
While it is a legal story, it is at main, a systems story.
Organizations that build clean, consistent infrastructure, where policy is embedded in process, decisions are routed correctly, and documentation is generated automatically, are more resilient than organizations that depend on people to do the right thing at the right moment.
People do not always do the right thing. Systems can be designed to produce the right outcome anyway.
The Inside Advantage 
Leave a comment