Essay/Management · AI systems

Augment, don't replace. The case for redesigning the manager before automating the task.

Every deployment plan starts with the task. The ones that work start with the person who will have to answer for the task once it runs itself.

Jeff Wald28 Aug 20269 min read

Essay28 Aug 2026Management · AI systems

Every deployment plan I have read this year starts in the same place: a list of tasks, ranked by how much time they take and how easily software could do them. It is a sensible list. It is also the reason most of those plans stall inside two quarters.

The task is the wrong unit because a task does not exist on its own. It sits inside a role, and the role sits inside a set of expectations about who decides what when the task goes wrong. Automate the task and you have not removed work. You have removed the routine part of it and left the person holding everything that does not fit the rule, with no change to their authority and often no change to their title.

Managers feel this first. In the audit we ran across 27 organizations, the deployments that stalled did not run out of engineers. They ran out of managers with the authority and the time to redesign roles around what the software now did. The talent shortage was real, but it was a shortage of management capacity, not technical capacity.

Automate the task and you have not removed work. You have removed the routine part and left the person holding everything that does not fit the rule.

So the case for augmentation is not sentimental. It is sequential. Redesign the manager first: what they decide, what they can change without asking, what they are accountable for once the routine is gone. Then automate the task, into a role that has already been rebuilt to receive it.

The organizations that do this look slower for a quarter. They are the ones still running in the third.

Moves the library. Cited in the next scheduled review of Thesis 003. Thesis 003, Management Is Becoming an Operating System.