Two Posts. One Design Question.

Two recent posts on this site arrived at the same question from different directions.

The first was about Buc-ee’s. A cashier with a neuromuscular condition asked for a chair at his workstation. The company said no, escalated the disagreement, and eventually terminated him. The EEOC filed suit in May. The post argued that the failure was not a policy problem. It was a systems problem. The organization had become so committed to consistency that it lost the ability to exercise judgment. A supervisor who might have approved the accommodation apparently was not empowered to do so.

The second was about Amazon. A supervisor asked an employee intrusive questions about his personal life. Questions that in a different organization might have produced a discrimination verdict. Amazon won anyway. Not because the behavior was acceptable, but because the decision to terminate was made by centralized HR, not the supervisors with the conduct problem. The system routed the decision through people who didn’t have the same relationship to the facts. The outcome reflected the policy, not the supervisor’s mood.

Same legal domain. Opposite results. Different systems.

The question underneath both cases

When I put those two cases next to each other, the difference comes into focus quickly.

At Buc-ee’s, the decision that should have been standardized, whether an employee qualifies for an accommodation, appears to have been left to the supervisor. The local manager became the ADA policy. That is exactly backwards. Whether someone receives a reasonable accommodation cannot vary by location, by supervisor, or by how that supervisor happens to feel about standing.

At Amazon, the decision that required consistency, termination for documented attendance violations, was handled by centralized HR with a documented paper trail. The supervisor with the problematic conduct was downstream of the outcome. The system did not depend on that supervisor behaving well.

This is organizational design. And most organizations get it wrong in the same direction.

What should vary and what should not

There is a principle worth naming directly.

Some decisions gain value from consistency. Whether an employee receives an ADA accommodation. Whether a safety complaint triggers an investigation. Whether a pay decision is defensible. These decisions should produce the same outcome regardless of which supervisor happens to be involved. Organizations that leave them to local discretion are not empowering their managers. They are creating hundreds of independent policies, most of which were never reviewed, never documented, and never tested.

Other decisions gain value from local knowledge. A customer’s complaint. An equipment problem that only the person running the line would notice. A signal that something in the workflow has quietly broken. These decisions cannot be made well from headquarters. They require proximity, context, and the judgment of someone who was there.

The failure mode for most organizations is inverting these two. They centralize information, suppress what people on the ground actually know, and leave consequential individual decisions to whoever is closest. The result is that consistent standards never develop and local knowledge never travels.

What the chair case actually illustrates

The Buc-ee’s case is often read as a training problem. Managers need better ADA education. Supervisors need to know their obligations. That may be true, but it is not the deeper lesson.

The deeper lesson is that whether an employee can receive a workplace accommodation cannot be a judgment call at the store level. The moment it becomes a local decision, you have created as many accommodation policies as you have supervisors. Some will be generous. Some will be rigid. None of them will be defensible in the way a consistent, centralized process would be.

Buc-ee’s did not fail because a supervisor made a bad call. It failed because the organization appears to have had no infrastructure to take that decision out of the supervisor’s hands.

Amazon, whatever its other well-documented problems, had that infrastructure in place. The supervisor who asked the intrusive questions also recommended against termination. The person who made the decision was someone else entirely.

The design question

Healthy organizations ask this question deliberately: what should be consistent, and what should vary?

Standards should be consistent. Ethics processes. Accommodation procedures. Investigation protocols. Pay philosophy. These are not areas where local creativity adds value. They are areas where variability creates risk, inequity, and legal exposure.

Information should vary, or more precisely, it should flow freely from wherever it originates. Observations from the floor. Customer feedback. Early warning signals. Operational problems that only become visible close to where the work happens. Suppressing that information because it didn’t come through the right channel is how organizations lose the knowledge they need most.

The organizations that build this correctly are more resilient. Not because their people are better, but because they have designed for the reality that people are inconsistent. They route decisions requiring consistency through systems that produce consistent outcomes. And they create the conditions for information to travel from the edges to the center, where it can actually inform decisions.

The ones that get it wrong think they have done this. The policy is in the handbook. The training was conducted. The org chart shows the right reporting lines.

But a policy 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.

That gap is where risk lives.