As a freshly minted second lieutenant, my first assignment after graduating from West Point was in South Korea. I was stationed at Camp Casey, which sat eleven miles south of the demilitarized zone that separates North and South Korea. The work was demanding, and for long stretches monotonous. We rehearsed everything we might have to do until we could do it in our sleep, updating the secure frequency on our radios, replacing a tank track, working through hundreds of map exercises, then running the same maneuver again and again while someone picked apart every decision and made us reflect out loud on how we might do it better next time.
While the days were hard, they were also filled with a sense of purpose, with clear goals and associated standards. I felt challenged professionally and fulfilled personally. Everything we did was deliberate, rehearsed, and critiqued against clear expectations, and it really worked. The units I served with in the Army rank at the top of organizations I have ever been associated with over my 30+ year career in and out of the military.
I bring up the military because of a worry now spreading through companies. AI is taking over the entry-level work that junior employees have always learned on. If that work disappears, the argument goes, so does the way we train the next generation.
Most modern companies train their junior people by throwing them into the deep end and handing them low-end, monotonous tasks — grunt work — and we have built in an assumption that this is the only way to learn. My experience in the military suggests there is another way to teach people. The corporate approach was simply the default, a method that developed organically rather than one anyone designed. Watching senior people, catching feedback in the hallway, and grinding through low-value tasks taught some of the job, but it mostly revealed who could already figure things out on their own. We hired capable people, placed them in entry-level seats, set ill-defined expectations, and evaluated them against a subjective standard. The ones who figured it out advanced and the ones who did not eventually left. We called it training. Mostly it was sorting.
The entry-level work was never the curriculum. It was the cheapest classroom we had, and we mistook the byproduct for the method.
A System Built on Purpose
The military cannot rely on osmosis. It assumes capability must be built, deliberately and repeatedly, long before it is needed. Training is the work, not a tax on the work. Repetition runs to a standard, not to a clock. Before the live-fire assessment that actually counted, we had run the range perhaps thirty times. We did not stop when we got it right once. We stopped when it held every time, under fatigue, in poor weather, with equipment that did not cooperate. Standards were set high and stated early, so everyone knew what the objective was and how their role contributed to the outcome.
Training is the work, not a tax on the work.
Then there is the after action review, the most transferable practice the military has and the one corporations most conspicuously lack. After every engagement the unit takes its own decisions apart in detail. The commander probes each choice, including his own, finds the holes, and presses on how to do it better next time. The standard is honesty about what went wrong, because the point of the room is to improve and not to assign blame. All of it escalates from the classroom to the tabletop to the field to live fire, each stage harder than the last, until the unit has rehearsed the decision so many times it can still make it when the plan falls apart. That is what allows centralized planning with decentralized execution: the intent is set at the top, and the execution is trusted to the people on the ground.
The Scarcer Resource
This points to the deeper shift, the one the entry-level conversation tends to miss. As AI takes over the doing, the human contribution moves toward the directing. Fewer people execute the task. More people define it, judge it, and decide what to do when the machine produces something plausible and wrong. That is leadership, distributed far more widely than most companies have ever required it, and it demands judgment under ambiguity, the rarest thing in any organization.
We are about to need many more leaders, and we barely know how to make them.
Most companies treat leadership as something that surfaces on its own. The capable rise, the title follows, and development amounts to a few courses and the hope that experience handles the rest. Leaders are expected to emerge. They are rarely developed.
The military develops them. It treats leadership as the primary goal that everything else comes from. That leadership is built through years of progressive responsibility, honest feedback, and exposure to conditions designed to test judgment before the stakes are real. A young officer can run a platoon through live fire in miserable weather on no sleep because the system spent years building him to do exactly that. Decentralized execution is possible only because leaders were deliberately constructed at every level, down to the lowest one.
Build It on Purpose
The predictable objection is that the military can train this way because it has time and budget no business will grant. The opposite is true: it is brutally accountable, the cost of its failures measured in lives, and refuses to leave readiness to chance. Enterprises already spend the money, on pilots that go nowhere. The deeper objection is that knowledge work cannot be simulated the way a tank maneuver can. That is what the Army believed too, until it built the National Training Center in the Mojave, where a permanent opposing force defeats most units that rotate through. The simulation became harder than the real thing.
Organizations should consider taking some of the capacity AI creates by replacing low-value tasks and pointing that time toward teaching people the things that are harder to learn: leadership, judgment, and influence. We can build deliberate training for analysts, associates, and engineers, and the leadership development beneath it, the way the military has for a century. Before cutting headcount, spend the capacity you can find here.
Replicating this inside a company will be hard, and the conditions that make it work in the military do not transfer cleanly. I am not suggesting wholesale adoption of the military approach. The transferable idea is narrower: building explicit capacity to teach people the necessary skills, and having them demonstrate mastery against well-defined standards.
The question was never whether a generation can be trained without entry-level work to absorb them. The military has answered that for decades. The question is whether we will build the system on purpose, or keep pretending that osmosis was ever a system at all.
I welcome you to join me in this conversation.
Build the system on purpose.