The MBA Was an AI Manual

Half of what business school taught me just died. The other half became code. And the cruelest theory in the canon turned out to be the kind one. Part of the series.

I am a recovering corporate sellout. I have the business degree. I spent years in finance and corporate strategy, in the kind of rooms where people say "operationalize" without flinching. I optimized my entire career to avoid doing anything real — I delegated to smarter people, I managed instead of made, I got other people to do my CS homework and, later, my actual work. If there was a way to win by arranging things rather than building them, I found it.

So when I started running a company on machines, I assumed all of that was wasted. The strategy decks, the org-behavior seminars, the risk-management training I half-slept through at the bank. None of it taught me to code. None of it, I thought, could possibly matter here.

I was completely wrong, and the way I was wrong is the most interesting thing I've learned all year.

Here's the realization. Every management theory I ever studied was built to solve one problem: how do you get useful work out of a crowd of bounded, fallible, self-interested agents — people who can't hold the whole picture, who make mistakes, who might tire or shirk or steal? Taylor's stopwatch, Weber's bureaucracy, agency theory, the entire apparatus of corporate control — I always thought it was about people. It wasn't. It was about agents. And an AI is simply a new kind of agent. The canon I dismissed didn't die when I left the corporate world. It was waiting for me to meet its real subject.

Two things happened to that canon when I pointed it at machines, and both of them surprised me.

The first: the cruelest idea I ever learned became the kind one.

Frederick Taylor's scientific management — the stopwatch, the time-and-motion study, the breaking of human work into small, measured, repeatable tasks — was the most quietly horrifying thing they taught us. It is the philosophy that turns a person into a component. I didn't need the seminar to understand it; I'd lived it as a junior at the bank, functionalized into a thousand-line cog in a machine that held what I couldn't. Taylorism was cruel for one reason and one reason only: it was done to humans.

Do the exact same thing to an AI, and the cruelty inverts. You decompose the work into small bounded jobs, you measure them, you standardize them — and the worker you've just functionalized doesn't mind, doesn't break, doesn't lie awake feeling like a fraud. The machine takes the stopwatch roles. The human walks away from them. Taylorism was always waiting for a worker it wouldn't destroy, and we finally built one. The theory I hated most in school is the one that frees me now. That's not a footnote. That's the whole moral case for what I'm doing.

The second: half of everything I learned just went silent.

A huge part of the management canon exists for a single reason — humans don't want to work. McGregor's Theory X and Theory Y, the Hawthorne studies, motivation, morale, culture decks, incentive design, the whole sprawling human-relations school: all of it is the science of coaxing effort out of agents who get bored, get scared, get resentful, and need to believe in something. I sat through a lot of it.

The AI needs none of it. It doesn't get bored. It doesn't quit on a Friday. It will never need a mission statement or a one-on-one. So that entire half of my education — the motivational half — simply goes quiet when your workforce is machines.

What's left is the other half, and it's the half the bank actually ran on: control. Agency theory — the principal and the agent who might not act in your interest. Robert Simons' boundary systems — the discipline of telling an empowered actor what it must never do, so it can move fast everywhere else. Internal controls, segregation of duties, the entire grim apparatus of the risk department I used to find so tedious. Subtract the human-motivation problem from management, and control is all that remains. That is why, when I say control, not autonomy, I'm not offering an opinion. I'm reading off the residue. It's what's left after the subtraction.

One piece of the canon survives untouched — the one I paid the least attention to. Drucker split all work into two kinds: efficiency, doing things right, and effectiveness, doing the right things. Business school spent most of its hours on efficiency — the part the machine now owns completely, and does better than any human ever could. The little it said about effectiveness — judgment, taste, deciding what is even worth doing — turns out to be the only part that's still a job. Mine. The machine runs at ninety percent of the throughput and I keep the ten percent that was never reducible to a process.

There's a joke I'm still sitting with. I sold out to learn how big systems run, and I spent years certain I'd wasted the time. It turns out I was being handed the operating manual for a kind of company that didn't exist yet. The corporate machine I escaped is the one I'm now rebuilding — the rules, the controls, the segregation of duties, the whole bureaucratic skeleton Weber described a century ago — except this time the parts are machines instead of people, and the single human left inside it gets to do the one thing the bank would never let a junior do.

Decide.

Next: telling it what not to do — the boundaries you wrap around a worker more capable than you are.