The hype is real. The spending is real. The productivity is not showing up in the numbers.
A recent survey of 6,000 companies found that more than 80% see no measurable gains from their AI investments. US total factor productivity grew about 0.07% in the year to early 2026. That is a rounding error. Meanwhile every earnings call mentions AI, every company has a pilot running, and every consultant has a deck.
That gap is the most interesting thing happening in business right now.
This has happened before
Economists have a name for this pattern: the productivity J curve. When a genuinely transformative technology arrives, measured productivity does not jump. It dips. Companies pour money and attention into the new thing, output barely moves, and the sceptics declare the whole thing overhyped. Then, years later, the curve bends sharply upward.
The clearest example is electricity. When electric motors became available, factory owners did the obvious thing: they swapped out the steam engine for an electric motor and kept everything else the same. Same building, same layout, same workflow. Productivity stayed flat for roughly two decades. Long enough for serious people to conclude that electricity was a disappointment.
They were wrong. They were just early.
The gain was never in the motor
The breakthrough came when a new generation of managers stopped treating electricity as a replacement part and started treating it as a reason to redesign the factory. Instead of one giant motor driving everything through shafts and belts, they put a small motor on every machine. Suddenly the factory floor did not have to be arranged around the power source. It could be arranged around the flow of work.
That is when productivity exploded. Not when the motor arrived, but when the work was rebuilt around it.
The lesson is uncomfortable because it means the hard part was never the technology. The hard part is reorganization, and reorganization is slow, political, and painful.
We are dropping motors into steam factories
This is exactly where most companies are with AI today. They have bolted a chatbot onto the old factory. Same workflow, same handoffs, same org chart, with an AI tool sprinkled on top. The result is a local speedup here and there, and a flat line in the aggregate numbers.
The companies that will bend the curve are the ones redesigning the work itself: rethinking what a team is, what a role is, which steps still need a human, and which processes should not exist at all. That work has barely started anywhere.
The dip is where we are.
Who gets there first
If the gain comes from reorganization rather than adoption, then the advantage goes to whoever can reorganize fastest. That is rarely the giants. A large enterprise can buy all the AI it wants, but it cannot quickly rewire its org chart, its reporting lines, or the careers that depend on the current structure.
A smaller team can rearrange the entire floor in a quarter.
That makes the flat productivity numbers something unusual: for the underdog, they are a window. The gains are available. Most people just haven't reorganized to collect them.
A bet, not a certainty
The J curve is a pattern, not a law. Economists still disagree about how much of the current dip is measurement lag, how much is misallocated investment, and how much is simply hype. I am describing a bet, not a certainty.
But it is a bet I am making in my own companies. New power sources have always paid off late, and always after the work was redesigned to use them. There is no reason AI should be the exception.
The productivity isn't showing up yet. That is not evidence that it won't. It is evidence that the reorganization hasn't happened yet, and the first to do it will take ground the rest of the market has already decided isn't there.