AI is learning your old company. That’s the problem.

01The motor was in the factory. The productivity was not.

In 1899, electric motors of all kinds accounted for less than 5 percent of the mechanical drive capacity in American factories 1. By 1919, electric motors represented over 53 percent of the horsepower used to drive machinery 2. The technology spread. The productivity did not follow, at least not for years.

The economic historian Paul A. David showed that factory electrification did not have an impact on productivity growth in manufacturing before the early 1920s, four decades after the first central power station opened for business 1. In 1900, he wrote, an observer might well have remarked that electric dynamos were to be seen “everywhere but in the productivity statistics!” 1

That line was a deliberate echo. In 1987 the economist Robert Solow had written: “You can see the computer age everywhere but in the productivity statistics.” 3 It became known as the productivity paradox. David’s answer, in 1990, was that it had happened before, and that the delay had a cause 1.

02Swapping the engine changed almost nothing

For about two decades, most factories electrified the way most companies now adopt AI: by plugging the new technology into the old layout. Under the “group drive” system, an electric motor simply turned sections of the same overhead shafting that a steam engine used to turn 12. Retrofitting an old plant typically meant adding motors to the original equipment, while the belts, the shafts and the engines that powered them stayed in place 1.

This sort of overlaying of one technical system upon a preexisting stratum is not unusual during historical transitions from one technological paradigm to the next.

Paul A. David, 1990 1

David saw the same pattern in the computer era of his own day: old paper-based procedures kept alongside the new electronic ones, “sometimes to the detriment of each system’s performance.” 1

03Redesigning the factory changed everything

The gains arrived when engineers stopped asking how to power the old factory and started asking what a factory could be. With “unit drive”, each machine got its own electric motor 12. The heavy overhead shafting disappeared, and with it the reasons factories had been built the way they were 1:

• Lighter buildings, because nothing heavy had to hang from the ceiling anymore 1.
• Single-story plants instead of costly multistory ones, which had existed largely to keep the line shafts short 1.
• Machines arranged in the natural sequence of the work instead of in rows parallel to a shaft, so materials moved less 2.
• Changes to one line or one department without shutting down the entire power system 1.

Manufacturers found that the energy savings were the least of it. Reports summarized by Professor F. B. Crocker at a technical society meeting in 1901 said output often rose “20 or 30 percent or even more, with the same floor space, machinery, and number of workmen.” 2 Devine’s conclusion: manufacturers came to see the savings as far less important than the gains from increased production 2.

The redesign had to be learned, plant by plant, by a new cadre of factory architects and electrical engineers 1. When it spread, it showed up in the aggregate: David estimates that about half of the 5 percentage point acceleration in the total factor productivity growth of US manufacturing in 1919 to 1929, compared with 1909 to 1919, is accounted for statistically by the growth of one measure of factory electric motor capacity in that decade 1.

04Why it looks like nothing is happening

Economists now have a name for the gap. Erik Brynjolfsson, Daniel Rock and Chad Syverson call it the Productivity J-curve 4. General purpose technologies such as AI require large complementary investments: new processes, new products, new business models and new human capital. Those investments are mostly intangible and poorly measured, so in a new technology’s early years measured productivity growth is underestimated, and later, when the benefits are harvested, it is overestimated 4.

The pattern is already visible in today’s data. In McKinsey’s 2026 survey, the small group of AI high performers is distinguished less by tools than by intent and discipline: nearly three-quarters report fundamentally redesigning workflows because of AI, against one-quarter of other respondents 5.

05AI adopts the old shape of your company

And the market has already moved. On Cloudflare’s network in June 2026, bots and agents generated 57.5% of HTTP requests, against 42.5% from humans 6. Between January and April 2026, 68.01% of US Google searches ended without a click 7. The four pillars marketing stood on are cracking at the same time: a person on the other side, exposure as attention, behaviour as intent, persuasion as strategy.

06Redesign the work: where to start

This is what “AI-native” should mean. Not a company that uses more AI, but one whose work was redesigned around it, the way the single-story factory was designed around the unit-drive motor.