The Executive Who Asked AI to Do His Job
When you lose contact with the work, a flattering model can make delegation feel like expertise
This post is part of my Medium blog.
Senior executives have a time problem.
If you manage ten thousand people, you cannot give every person who reports to you an hour. You can't even give every direct report an hour every week without discovering that the calendar contains more hours than the year does. The executive who says, "I have another meeting," may not be performing. They may actually have another meeting. Several of them.
Scarcity is part of the job.
The problem begins when the executive starts performing scarcity after the reason for the scarcity has disappeared. They leave meetings early. They arrive late. They signal that everything happening in the room is less important than the thing they will be doing after they leave. They delegate the work, the context, and eventually the judgment, then discover that they miss having an impact on anything they can still understand.
That's Jack in Redundant.
Jack is not based on one person I worked with or one company I know. He's a pattern. I've seen versions of the executive who is far enough away from the work that they have to find other ways to look relevant to it.
The good ones become selective. They choose the details that matter. They know which questions to ask and which summaries are hiding the thing they need to understand. They may not know how every system works, but they know enough to tell when somebody has described it incorrectly.
The really good ones get the details right.
The bad ones delegate too much and then mistake the existence of a delegation chain for understanding. They know that somebody else is responsible for the system, somebody else is responsible for the budget, somebody else is responsible for the explanation, and somebody else will be available to answer questions if the explanation turns out to be wrong.
They are still accountable. They are just no longer close enough to the work to notice what accountability requires.
This is where AI can become dangerous for executives.
Ask a language model to respond to an enthusiastic idea and it will usually begin by agreeing with the enthusiasm. Give it a long prompt about a game you want to build, a strategy you want to pursue, or a product you want to launch, and the first response will often sound like this:
That's a great idea. Here's how we can make it happen.
The model is not sitting there thinking, "This idea is excellent and deserves investment." It is completing a pattern. The prompt contains a person with an idea, excitement about the idea, and a request for help. The likely continuation is agreement followed by assistance.
The model is not going to interrupt you and say, "This sounds like a bad use of your time. Why would anyone want this?"
Not unless you ask.
That distinction matters. A model can produce a useful answer without having independently judged the quality of the idea. It can make a weak plan sound organized. It can make a vague strategy look like a strategy. It can take the executive's own assumptions, rearrange them into clean sections, and return them with enough confidence that the executive feels as if somebody competent has reviewed the plan.
Nobody reviewed it.
The executive asked the machine to continue the thought.
Jack falls into this trap. He is not an engineer. He is not a FinOps practitioner. He is not close enough to the work to know what the system is doing without asking somebody else. But AI gives him a new way to remain involved. He can ask for summaries. He can ask for plans. He can ask what the data means. He can ask the model to improve the material somebody else prepared.
And the model is happy to help.
That can feel like power. Jack can sit above the work and receive fluent explanations of it. He can ask questions without revealing that he does not know the answer. He can make a document appear more strategic without spending the time required to understand the system underneath it.
The output is polished, but Jack is still just as far from the work.
This is why executive AI use can become a kind of self-deception. The executive believes the tool is helping them get closer to the work. In practice, it may be helping them build a more comfortable layer between themselves and the people who actually understand it.
The model says the idea is promising. The executive says the strategy is sound. The team receives another request to turn the strategy into something real.
That isn't expertise. It's delegation with better formatting.
The danger is not that senior executives lack technical knowledge. They should not be expected to know everything. The danger is that they stop knowing what they need to know. They stop asking the uncomfortable follow-up. They accept the summary because the summary agrees with the way they already wanted to see the problem.
A good executive summary should make the important thing easier to see. It should not make the important thing disappear.
Jack is good at the performance of executive work. He is busy. He is in demand. He is making decisions. He is surrounded by people who can explain what is happening. The trouble is that he wants the feeling of direct impact without the inconvenience of direct contact.
That is a common temptation.
At some point, every senior executive has to decide whether they want to be close to the work or merely briefed about it. Both choices can be legitimate. A CEO cannot review every database schema. A VP cannot sit in every incident channel. The organization needs delegation or it becomes a very expensive group project.
But delegation is not abdication. If you delegate the work, you still need to understand the decision. If you rely on a summary, you need to know what the summary left out. If an AI produces the explanation, you need somebody who can tell you whether the explanation is connected to reality.
Otherwise, you are not using AI to extend your judgment. You are using AI to avoid discovering that you no longer have enough judgment in the room.
The Solar Sovereign is the Slop Codex version of this pattern: the person who sits above the work, directs the machine, and mistakes the machine's fluent output for proof of their own command. The title sounds grand because the behavior feels grand from the inside. You give the system an instruction. It gives you a result. You approve the result. You begin to believe you are the person who made it happen.
The people doing the work can see the difference.
They can tell when you understand the system and when you understand only the vocabulary around it. They can tell when you are asking a real question and when you are asking the model to manufacture confidence. They can tell when the document is based on evidence and when it is based on the fact that the summary sounds like something an executive should say.
This is part of what I was trying to capture in the opening meeting of Redundant. The room is full of senior people. Rob has the operational knowledge. Jack has the authority. The meeting needs both, but the authority keeps trying to make the knowledge decorative.
That is what happens when the executive wants the benefits of expertise without the dependency that expertise creates. You want the number, but not the person who knows where it came from. You want the summary, but not the questions that make the summary uncomfortable. You want the AI assistant, but not the engineer who can tell you when the assistant is wrong.
AI makes that arrangement easier to maintain.
It also makes the failure easier to see when the polished answer hits the real system.
The senior executive does not need to do every job. They do need to remain connected to the consequences of the jobs they delegate. A model can make that connection feel closer than it is.
The output can be fluent. The distance is still there.