Are you running an electric motor in a water-turbine factory?

Textile mills were built beside rivers. Not for the view — the river was the power source. A water turbine turned a central shaft, the shaft ran overhead line shafting the length of the building, and belts came down off it to drive each machine.

Everything about the mill followed from that. Its location was wherever the water fell hard enough. Its shape was long and narrow, because machines had to sit near the shaft. Its floors were stacked to get more machines within reach of one drive. And the workers had to live wherever the river happened to be, which was usually nowhere anyone wanted to live.

Then electricity arrived, and almost nothing happened

The obvious move was to take out the water turbine and put in an electric motor. Mills did exactly that. The motor drove the same central shaft, turning the same belts, running the same machines in the same positions in the same building.

It worked. It also changed remarkably little. The economist Paul David documented this in 1990 in The Dynamo and the Computer: decades passed between electricity becoming available and electricity showing up in productivity figures.

What actually unlocked it

The change came when somebody built a new factory instead of retrofitting an old one — and designed it around what electricity made possible rather than around what the river had forced.

Every machine got its own motor. That killed the central shaft, which killed the constraint that machines had to sit near it. Now the floor could be laid out in the order the work actually happened — single storey, wide, arranged by process instead of by proximity to a belt.

And the bigger one, which is the part people forget: power now arrived down a wire. The factory no longer had to be at the river at all. It could be built near a town, which meant it could hire people who had a choice about where they worked.

The gain was never in the motor. It was in everything you could stop doing once the river no longer decided where the factory stood.

Now ask it about your own company

Your business also has a river. It is human attention. Every process you run was shaped by the assumption that thinking is expensive and scarce — so work queues behind whoever can approve it, someone fills in a form afterwards to tell the system what already happened, and the month closes three weeks late because a small number of people have to look at everything.

That is the line shaft. It is the reason your quotations wait, your reports get rebuilt by hand, and one senior person is the bottleneck on everything that leaves the building.

So when a company buys AI, the only question that matters is which of the two things it is doing. Bolting a motor into the water-mill: a subscription, a chatbot on the intranet, staff pasting company material into a consumer tool when they remember to. Everything downstream unchanged.

Or redesigning around what is now cheap. Taking one flow — quotations, tenders, compliance filings — writing down what makes an output correct, building that in as a check, and removing the person from the middle of it entirely.

The evidence that most are doing the first one

MIT’s Project NANDA studied enterprise adoption across structured interviews with 52 organisations, survey responses from 153 senior leaders, and a systematic review of over 300 publicly disclosed AI initiatives and found around 95% of organisations reported no measurable effect on profit and loss.

That figure gets quoted as proof AI is overhyped. It is not. It is the retrofit problem, measured. The motor works fine. Nobody moved the building.

Do not consume AI. Make it work.

Consuming AI is buying access and hoping people find a use for it. The output depends entirely on who is at the keyboard, nobody checks it against anything, and nothing accumulates.

Embedding AI is putting it inside a piece of work that happens whether anyone remembers or not, running on your own knowledge rather than the public internet, with what comes out checked against a written standard before a person ever sees it.

The mills that redesigned did not do it all at once. They built one new plant, proved it, and then never built another the old way. Start with one flow, measure the turnaround before and after, and let that number decide whether there is a second.

You already know how to use electricity. Nobody had to sell you on it. The only real question is whether you are running your motor in the old mill.