The Churning
How a company prepares for the AI years
Every company is churning right now. Boards want the prize: lower costs, new products, a return on the AI money. Most companies bought the AI and cut people. Few changed how the work flows, and that is the only part that pays.
When electricity came, factories put one big motor where the steam engine used to be, and for forty years little changed, because nobody wanted to tear down a factory that still worked1. The gains came only when they redrew the floor. A company’s floor is its teams and its process, and that is far cheaper to redraw. An old story says what to prepare for.
The old story
In an old story from India2, gods and demons churn the ocean for a prize: a drink that lets them live forever. A mountain is the stick and a giant snake is the rope. When the mountain starts to sink, a giant turtle holds it up. The first thing out is not the prize. It is poison. One god holds it in his throat, and his throat turns blue. Only then come the treasures, and a fight over who drinks.
Companies are living the same story, in the same order.
Hold the mountain
Nothing moved until the turtle held the mountain. For a company, the mountain is the boring part: clean data, tests, clear rules for what an agent may touch, and logs of what it did. An agent cannot read what nobody wrote down.
The mountain is also the human in the loop, who can tell the prize from the poison. The model sounds just as sure when it is wrong, so stay awake. It is also a patient teacher: ask, and it explains its work. An honest human gets better every week. With the mountain steady, pick which end of the snake to hold.
Head or tail
The demons grabbed the head of the snake, where the glory is, and churned with fire in their faces. The gods took the tail. The shiny chatbot is the head. The support queue is the tail. Nobody claps for the tail, but it pays back first.
Not every product needs AI. Sort them into three piles. Where AI changes what the customer gets, invest. Where it only cuts cost, automate quietly. Where a customer can rebuild it with an agent in an afternoon, close it. Count each quarter: cost in, return out. Two red quarters, stop. Then the churning starts, and the first thing out is not what anyone ordered.
Poison first
The first thing AI brings is not the prize. It is poison: code that works at lunch and breaks by dinner, and work that is “95% done” until a human finds the missing part.
Companies react in two ways. Some swallow it: AI everywhere, no rules, and soon their name is in the news for a data leak. Others spit it out: they ban the tools, and their best people use them on their phones anyway. The god in the story did neither. He held the poison where he could see it. A company can do the same: name one person who owns what goes wrong, such as a leaked password, a wrong answer to a customer, or a shortcut nobody checked. Give that person three things: a log of what every agent did, an off switch that works, and a human yes before anything that cannot be undone. Then take turns on blue throat, like on-call. The agents watch, dig and suggest. The human smells trouble, makes the call, and says sorry when needed. Once the poison is held, the treasures can come.
Who gets the prize
The story ends with a trick. A stranger promises to share the prize fairly, then pours it all for the gods while the demons are looking away. The demons did half the work and got nothing. Companies do the same: AI saves money, the gains go to the top, and the people who did the work get cut. Next time, nobody wants to churn. Share the prize with the people who did the work.
What to do on Monday
- Fix the mountain. Clean data, tests, access rules, logs.
- Keep a human in the loop. Awake, honest, checking the work.
- Take the tail. The boring cost before the shiny chatbot.
- Sort the products. Invest, automate, or close.
- Count every quarter. Cost in, return out. Two red quarters and it stops.
- Name who holds the poison. One owner, and a weekly blue-throat shift: agents watch, humans decide.
- Share the prize. With the people who did the work.
Before the prize comes the poison. The companies that prepared for it are still standing when the prize is poured.
-
“The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox” Paul A. David, 1990 ↩