In 2002, the U.S. military ran a war game called Millennium Challenge.
Cost: $250 million. The most expensive simulation in military history at the time.
The enemy commander, a retired Marine general named Paul Van Riper, refused to play by the script. He sent orders by motorcycle courier instead of radio. He signaled his aircraft with WWII-era light systems instead of anything electronic.
He sank the entire American carrier battle group.
Here’s the part that matters more than the loss itself. The military didn’t absorb the lesson. It reset the simulation, refloated the ships, rewrote the rules, and forced a scripted outcome so the exercise could still validate what it had already decided to prove going in.
That’s not a failure to predict complexity. That’s a system so committed to its own model that it rewrote reality instead of the model.
I’ve watched smaller, quieter versions of the same thing. A customer acquisition playbook, built for landing new clients, applied instead to mature accounts that needed something else entirely. A small team’s motivation problem, met with stricter processes instead of a real look at the individuals struggling. People were fired for “lack of motivation” when the actual gap was a mismatch between what new leadership expected and what the existing team had ever been told to expect.
In every case, the response to visible failure was more process. Tighter rules. Another approval layer.
The situation got worse each time, not better — and the fix, when it finally came, wasn’t a better process at all. It was one question, asked differently: not “how can we improve this process,” but “is this process actually helping you do your job?”
The answer, both times I asked it, was immediate and clear. The system had failed them. So had the people meant to be running it.
You can’t replace someone’s need to be heard with a five-step workflow requiring three rounds of approval. Come on.
Complicated systems respond to better engineering. Complex ones — the ones actually made of people, incentives, and shifting expectations — don’t get fixed by adding another gear to a machine that was never a machine in the first place.