In January 1914, a crowd of job seekers gathered outside Ford’s Highland Park plant in Michigan. Ford had just announced that it would pay five dollars for an eight-hour day. The going rate at the plant had been $2.34 for nine hours.
Ford had more than doubled the wage. Part of the reason was the very thing that had made his factory so fast.
Twenty-nine operations
Around April 1, 1913, Ford’s engineers tried something new on the flywheel magneto, a part of the Model T’s ignition. Until then, one man assembled each magneto from start to finish, in about twenty minutes.
They broke his job into 29 operations, each done by a different man along a line. Assembly time fell to 13 minutes and 10 seconds. With more tinkering it came down to five.
Then they did the same to the whole car. Assembling a chassis on a stationary stand had taken, at best, 12 hours and 28 minutes. On the moving line, by early 1914, it took one hour and 33 minutes.
The men on the new lines disliked the work. By late 1913, labor turnover at Ford was 380%. To grow the workforce by 100 men, the company had to hire 963. The five-dollar day, as The Henry Ford museum tells it, persuaded many of them to put up with the pace.
Ford, for his part, saw no problem. “I have not been able to discover that repetitive labour injures a man in any way,” he wrote in 1922.
The man who first made the case for splitting work had seen it differently, a century and a half earlier.
Eighteen steps to a pin
In 1776, in the first chapter of The Wealth of Nations, Adam Smith described a small workshop making pins. Making a pin, he wrote, was divided into about eighteen distinct operations. “One man draws out the wire; another straights it; a third cuts it; a fourth points it.”
He had seen a shop where ten men did all of it. Between them, they made upwards of 48,000 pins a day, about 4,800 each. Working alone, he wrote, each of them “certainly could not” have made twenty, “perhaps not one pin in a day.”
That is the trade. Split work into small steps and output can soar. So can the boredom.
Smith knew it. Hundreds of pages later, in Book V, the same author warned that a man “whose whole life is spent in performing a few simple operations” generally “becomes as stupid and ignorant as it is possible for a human creature to become.” He did not treat this as a rare side effect. Unless government took pains to prevent it, he wrote, this was the state into which “the great body of the people” must fall.
The book that made the case for splitting work also contained the bill.
Most workplaces now split work by desk instead of by bench. Picture a claims office where one person checks the form, a second checks the policy, and a third approves the payment. Each desk is fast. The claim that is wrong in a way none of the three checklists covers can pass through all of them, because nobody’s job is to look at the whole claim. The person on the second desk may have noticed something odd. Noticing odd things wasn’t her step.
An assembly line for AI
The same trick works on AI, and the reason sounds a lot like Smith’s.
Anthropic’s guide to building AI agents describes a pattern it calls prompt chaining: break a task into a fixed series of steps and give each step to a separate model call, each one checking or building on the last. The aim, it says, is “to trade off latency for higher accuracy, by making each LLM call an easier task.” A related pattern splits one job into parts handled side by side, because, in the guide’s words, models “generally perform better when each consideration is handled by a separate LLM call.”
A model in step three of a chain doesn’t get bored, and it won’t quit for a better offer down the road. Smith’s human cost doesn’t apply.
His condition still does. The guide recommends chaining only when the task “can be easily and cleanly decomposed into fixed subtasks.” Pins could be. Plenty of work can’t, and splitting it anyway leaves the same gap as the claims office: every step done well, and no one holding the whole.