What happens to a company when the work is done by machines — and which two jobs are left for the people. The argument in short; the full paper goes deeper.
Most of what you recognise as management exists for one reason: the people doing the work have interests of their own. Take that away and the structure you have been running is not cheaper. It is unnecessary.
Every layer in it is there to solve one problem. The people executing the work want things you don't want, and they know things you don't know. So you supervise. You set targets. You design incentives. You add a manager for every seven people, and a manager for every seven managers. Economists have a name for what all of that costs, and the name has stuck since 1976: agency costs.
A machine doing the work satisfies neither condition. It has no career to protect, nothing to gain by hiding a mistake, and no preference between the boring task and the interesting one. Every action it takes can be written down at almost no cost.
This is the part that gets misread as a matter of degree. It is not cheaper supervision. It is the removal of the thing supervision was for.
The first is saying what you want precisely enough that literal execution produces the right thing. Your instructions have always been defective — vague, incomplete, contradictory in places — and experienced people quietly repaired them on the way past without ever billing you for it. Withdraw that repair service and the quality of your instructions becomes the thing that limits you.
The second is being able to answer for what comes out. Not signing off on it — answering for it. Those are different, and the difference has four parts: you can actually stop the process, you can establish what it did and why, nobody can overrule you without taking on the accountability themselves, and you have enough time to think. A named person without all four is not accountable. They are somewhere to point after the fact.
Three things follow, and none of them are comfortable.
Your headcount stops being a measure of capacity. The limit on a manager was never how many people they could supervise; it was always which mechanism was doing the coordinating. Once that mechanism is the specification, one person can direct a great deal — but the number of outcomes a single person can genuinely answer for is small, and shaped differently. That number is your real constraint.
Your junior pipeline is where the damage shows first. People learned to direct by having been directed. Remove the bottom rung and you have not saved a salary, you have stopped producing the people who were going to run the place. Every firm gains by cutting it and every firm loses if they all do, which is why no single company will fix it.
And your durable advantage narrows to two things, neither of which can be bought. A record of how work in your domain actually goes wrong — which requires having got it wrong, in your domain, and keeping the account. And the standing to be believed when you say what happened. Both compound. Neither is available to a competitor with a bigger budget.
The full paper works through the coordinating mechanism, the anatomy of the firm, the competitive position, what the regulator has to do with it, and the conditions under which the whole argument turns out to be wrong.
Read the paper on SSRN →