The Separation of Accountability from Execution

For most of the twentieth century, and for much of your career, accountability has been bundled with execution. If you produced the work, you owned the result. The person who wrote the code was responsible for the code. The person who wrote the report was responsible for the report. Execution and accountability came as a package.

This made sense then. It made the attribution clear. If you did the work, you owned it. If something went wrong, you had to face it. If something went well, you got credit.

But AI changes this. An AI system can execute the work. A human can be accountable for the decision to use the AI system to execute the work. These are two different things. And their separation is not a loophole — it is the definition of management.

Consider a radiologist who uses an AI system to help interpret X-rays. The radiologist does not build the model. They do not train it. They do not write the code. But they do own the decision to use the model, the decision about what it is allowed to diagnose, the decision about when to override it, the decision about when to stop and get a second human opinion. If the AI system misses a tumor and a patient is harmed, the radiologist is accountable — not because they failed to execute the interpretation, but because they failed to manage the tool appropriately.

This is not unfair. This is responsibility. The radiologist is the one with the medical license. The radiologist is the one with the stakes — their reputation, their career, their ability to continue practicing medicine. The radiologist is the one who can be sued and fired and held accountable in all the ways that matter. So the radiologist is the one who has to own the use of the AI system.