Restructuring Lag
Electricity won in 1900 and factory output sat flat for thirty years. We're living the replay
Hey.
In 1900, electricity had already won. The dynamo was proven, the wiring was cheap, the factories were plugging in. And then, for nearly thirty years, almost nothing happened to productivity.
Read that again, because it is one of the strangest facts in economic history. The most transformative technology of the era arrived, got adopted, spread through the economy, and the numbers that were supposed to move sat flat for a generation.
We are living the exact same decade right now. And almost everyone is drawing exactly the wrong conclusion from it.
The technology arrives in years. The advantage arrives in decades. The lag is not the wait. The lag is the window.
The Factory That Didn’t Change
Here is what actually happened in those factories, because the answer is the whole edition.
When electricity arrived, factory owners did the obvious thing. They ripped out the giant central steam engine and bolted a giant electric motor in its place. Everything else stayed identical. The same central shaft running the length of the building, the same forest of belts and pulleys hanging off it, the same machines arranged by their distance from the power source instead of by the logic of the work.
They had electrified the factory without changing the factory.
And it did almost nothing.
The economist Paul David told this story in 1990, and it is the best thing ever written about why powerful technology disappoints in the short run. The gains did not arrive until a new generation of owners asked a different question. Not “how do I plug electricity into my factory,” but “what would a factory look like if it had been born electric.” The answer was the unit drive: every machine with its own small motor, the central shaft gone, the whole floor rearranged around the flow of the work instead of the flow of the power. Single-story, spread out, lit, ventilated. That redesign, not the motor, is what finally doubled productivity. It took about three decades, because it was not a purchase.
It was a reorganization, and reorganization is slow, expensive, and invisible while it happens.
The technology was installed in years. The value took a generation, and it went almost entirely to the firms that used the quiet decades to rebuild.
Confusing Headline
Now look at the headline that confused everyone this year.
In February, the National Bureau of Economic Research surveyed six thousand executives. More than eighty percent said AI had produced no measurable impact on their productivity or their employment over three years. Three years. The most hyped technology of our lifetime, and four out of five companies cannot find it in their own numbers.
In those same years, aggregate American output per worker grew at nearly double its decade-long average.
Both of those are true at once, and the contradiction is not noise. It is a signal we have seen before. In 1987, the economist Robert Solow made the famous joke that you could see the computer age everywhere except in the productivity statistics. They named the paradox after him. Then, between roughly 1995 and 2004, the computer gains finally arrived in a flood, and they came not from the computers but from the decade of process redesign the computers had quietly forced. Same shape. Install fast, benefit slow, and the benefit accrues to whoever reorganized.
Here is the part the doom headlines skipped. One of the economists behind that survey, Stanford’s Nicholas Bloom, said the quiet part out loud: the steam engine, the electric motor, the computer and the internet all had enormous long-run effects and almost none in their first five years.
The man holding the damning number is telling you the number is normal.
So What Is the Restructuring Lag?
This is the Restructuring Lag: the gap between a general-purpose technology arriving and its value showing up in the numbers, a gap that closes only when organizations rebuild themselves around the new capability instead of bolting it onto the old shape.
And the lag is not dead time. It is the single most valuable window a business ever gets, because advantage built during the lag is cheap, because it is not yet legible to your competitors and not yet priced by the market.
Costly Misread
Here is what most people overlook, and it is a costly mistake.
When the numbers stay flat, the consensus reads it as deflation. The bubble was overblown. AI was hype. Wait for proof before you commit. That reading feels sober and data-driven, and it is exactly backward.
The proof, by definition, arrives at the precise moment the arbitrage closes.
Once AI shows up cleanly in the aggregate productivity statistics, the reorganization will be common knowledge, the playbooks will be published, and the advantage will be competed away to nothing. The flat numbers don't mean the opportunity is fake. They are telling you the window is still open and most of your competitors are still bolting electric motors onto steam-engine workflows.
Betting a Company
I am betting a company on exactly this. With Belkins Home I am not adding an AI feature to an agency. I am trying to rebuild the agency from the studs around what these systems can now do, which is slow and painful and, for a while, looks like no progress at all. I made this case in an earlier edition, The Great Restructuration: the real arrival of AI is not a feature you bolt on; it is an organization you rebuild. Belkins Home is me living inside that argument.
That feeling, the feeling that you are pouring in effort and the dial is not moving, is not failure. In a Restructuring Lag, it is the sensation of being early. It is what the inside of the window feels like.
Everyone is waiting for AI to show up in the numbers. By the time it does, the window to build the advantage will already have closed.
Everything above is the diagnosis, and it is free. What follows is the prescription, the part I would think twice about handing to your competitor: what to do about the lag depending on where you sit, and the five moves to start on Monday.
For my readers
So stop waiting for proof, and start closing your own lag. Here is what that looks like depending on where you sit.
If you are a founder or an operator, do not buy AI. Reorganize around it. The gain was never in the tool, it was in the workflow you are willing to tear up and rebuild. Take your most important process, the one shaped by constraints from five years ago, and ask the unit-drive question: what would this look like if it had been born today, AI-native, with no legacy steps to protect. Then build that, and delete everything that only existed to work around a limit that no longer applies.
If you are an employee building a career, the skill that pays in a lag is not prompt tricks. It is process redesign. Anyone can add a tool to a task. The valuable person is the one who can look at how the work is done and rewire it from scratch. Be the architect of the new factory, not the one who plugged a motor into the old one.
If you are an investor or a builder, the returns are hiding inside the lag, which is why they are cheap. Bet on the unglamorous companies quietly rebuilding their operations, not the ones with the most impressive demo. In the electrical era the fortunes were made by the manufacturers who redesigned their factories, not by the people who sold the motors.
If you lead an organization, your job is to protect the reorganization through the period when it looks like nothing. The restructuring will dip your numbers before it lifts them, the classic J-curve, and there will be enormous pressure to kill it for a clean quarter. Holding the line through the dip is the entire job. It is also where almost everyone fails.
Five Moves for Monday Morning
That is the worldview. Here is the Monday-morning version. Five moves to close your Restructuring Lag.
Pick one core workflow. Not the whole company. One process that matters and that you secretly know is shaped like a 2019 org chart.
Ask the unit-drive question. If this were designed today, from zero, with AI as a native capability and not an add-on, what shape would it take. Draw that, honestly.
Delete the legacy steps. Find every step that exists only to work around an old constraint, the handoff, the review, the reformatting, and remove it. Most of the gain is in the deletion, not the addition.
Measure cycle time, not tool adoption. “We rolled out AI to the team” is a vanity metric. “This process now takes a third of the time” is the real one. Track the outcome, not the install.
Budget for the dip. Tell whoever you answer to, in advance, that the numbers will look worse before they look better. Name the J-curve out loud so nobody panics and kills the window at month four.
If you do one thing, redesign one workflow from zero. Adding AI to an old process is bolting a motor to a shaft. The money is in the new factory.
We’re Somewhere Around 1905
Back to that factory floor, around 1920, when the gains finally arrived. They did not arrive evenly. They poured into the firms that had spent the flat, frustrating, unprofitable decades rebuilding around the new reality, and they bypassed entirely the firms that had simply swapped the engine and waited.
We are somewhere around 1905 right now. The technology has won, the numbers are quiet, and the consensus is starting to mutter that maybe it was all overblown. That muttering is the sound of the window staying open a little longer for you.
The lag is closing. The only question that matters is whether you spend it building the new factory or waiting for permission that, by the time it comes, will be worthless.
Thanks for being here, and thanks for reading something that argues for patience in a feed that sells urgency. Let’s build through the quiet part.
Post-Credit Scene
Five things worth your attention this week, all about the long, slow, unglamorous gap between a technology arriving and the world reorganizing around it.
Paper
The Dynamo and the Computer, Paul David (1990). The handful of pages this whole edition stands on. David’s account of why electrification took thirty years to show up in the numbers is still the clearest lens anyone has built for what is happening to us now. Free to find, worth an afternoon.
Book:
Technological Revolutions and Financial Capital, Carlota Perez. Her installation-versus-deployment framework is the macro map of every lag like this one, including the bubble and the crash that tend to sit right in the middle of it. Dense, and worth it.
The Rise and Fall of American Growth, Robert Gordon. The pessimist’s case, that the big productivity gains are behind us. Read it as the steelman against this entire edition, so your optimism has to earn its keep.
Essay:
The Dynamo, the Computer, and ChatGPT, James Pethokoukis (Faster, Please!). The piece that drags David’s analogy fully into the AI moment. The fastest way in if a few pages of 1990 economics is a hard sell.
Documentary:
American Factory (Netflix). Not about AI at all, about a Chinese glass plant reopening in Ohio, and quietly the best film on what actually happens when old work and new systems collide on a real factory floor. The human texture the economics leaves out.
Thanks for reading.
Vlad