The world is not what it appears to be.
At least, that’s what I’ve been trying to persuade you of in the first two parts of this report. (I had planned it as three parts. It has since become four. The material kept growing – and so did the stakes.)
On the surface, everything looks more or less the way it has for decades. Stores are still open. Traffic still crawls. Your paycheck still lands on the same day. But just beneath that ordinary surface, the deepest layer of the global economy – the layer that runs on human knowledge and human thinking – is being quietly pulled apart and rebuilt. And when the rebuilding is finished, I believe we will look back and say it was a bigger event than the Industrial Revolution.
As I said in Part 1 of this report, the Industrial Revolution took two and a half centuries to run its course – starting around the middle of the 18th century and ending, roughly, in 1914, on the eve of World War I.
What is less well known is that it happened in two stages. The first began with machine tools replacing hand tools. The second began in the 1880s, when the American economy caught fire – an unprecedented explosion of growth that ran into the early 1920s. So the revolution may have taken 250 years from beginning to end, but the big changes – including the big opportunities for wealth building – took place in a fraction of that time. The heyday was those 40-odd years. That was when all the great fortunes were made.
Most historians date the AI Revolution to 1950, with Alan Turing’s paper on what became known as the Turing Test, followed by Arthur Samuel’s checkers-playing program in 1952 and the Logic Theorist program in 1956. Which makes it about 75 years old today – younger than the Industrial Revolution was when it hit its explosive stage, but moving at many times the speed.
What if the same pattern holds?
Here we are, living comfortably in 2026, with the world around us looking very much as it did five or 10 years ago – and yet I believe we are walking into the kind of moment businesspeople and investors faced in the early 1980s, when the danger of carrying on as if nothing was changing was counterbalanced by opportunities that made the people who acted very, very rich.
250 years of change is coming in the next five.
My prediction – and I will try to back it up in the pages that follow – is that this revolution will do its most important work in eight to 10 years. That the lion’s share of the change will come in the next five years. And that the deepest and most permanent changes – the ones that will decide who ends up on which side of the line – will happen in the next six to 18 months.
That is not a typo. Not six to 18 years. Six to 18 months.
In Parts 1 and 2 of this report, I laid out my thesis and gave you concrete, verifiable examples of industries and professions whose foundations have already been knocked out from under them. The sign painter, the videographer, the kid who made a national Porsche commercial in a week for a fraction of what it should have cost. In this installment, I’m going to complete the damage report: which jobs go first, the arithmetic that is collapsing the price of professional work, what the most credible people in the field are saying about how many jobs are at stake, and the three clocks that are ticking – one of which is ticking for you.
Underneath all of it sits a single question. I’ll ask it now, and by the end of this installment you will be able to answer it yourself: When knowledge stops being scarce, where does the value go?
I’ll warn you: This is going to be the hardest part of this report for you to read. It was the hardest to write. But you cannot make a good decision about an impending danger if you ignore it. So let’s take a hard look at what’s coming.
Muscle is safe… for now.
Let’s start with the one thing I can say for sure after weeks of studying, reading, and talking to the people who build these systems: AI is going to keep taking over work – but it is going to take over the thinking work far faster than the physical work.
So for the purposes of this report, I’m going to more or less set aside the jobs that are mostly “hands” – the barber, the gardener, the plumber, the manicurist. AI may nibble at the edges of that work, but it may also do the opposite: As more and more becomes automated, people may come to prefer the human touch and pay a premium for it.
The knowledge jobs are a different story.
The safest careers in America are now the most exposed.
Three years ago, Goldman Sachs put out a now-famous analysis estimating that generative AI could expose the equivalent of 300 million full-time jobs worldwide to automation, and that roughly two-thirds of occupations in the US and Europe were already exposed to some degree. At the top of the danger list was office and administrative support (about 46% of tasks exposed), legal work (about 44%), and architecture and engineering (about 37%). At the bottom of the danger list sat construction (about 6%) and installation and maintenance (about 4%).
I should say that I’m not entirely confident about those numbers, for a simple reason: They are too exact to take seriously. I asked Nigel – the AI I work with – where they came from. He said:
Your instinct is sharp, sir. These data come from a government database that breaks every occupation into its component tasks, with economists estimating, task by task, what a machine could do. On the one hand, there is a rationale behind the percentages. On the other hand, all of these estimates were made in early 2023. The capabilities of AI have more than doubled since then.
In other words, these numbers are educated guesses. My own guess is that they will turn out to be worse – especially for the credentialed people who, in every past economic upheaval, fared the best.
This shouldn’t surprise us. AI doesn’t replace muscle. It replaces the knowledge and the analytical thinking that muscle was never asked to do.
Here is roughly how I see the knowledge economy sorting out:
Medicine and healthcare. I see a big deflation coming in the areas of healthcare that rely heavily on pattern-reading and analysis: radiology, pathology, dermatology, and diagnostic medicine of almost every kind. As of earlier this year, the FDA had already cleared more than 1,500 AI algorithms for medical use – and more than three out of four of them are in radiology. When a machine can scan hundreds if not thousands of mammograms in one night without getting tired, the economics of paying a human radiologist $450,000 a year to do the same thing start to bend.
Law and accounting. Contract drafting and review, due diligence, legal research, tax preparation, bookkeeping, audit support – the enormous, well-paid middle of both professions is built on knowing precedent and processing documents, which is exactly what these machines do best. (I’ll show you the brutal arithmetic of this in a minute, and I’ll show it to you from inside my own family.)
Engineering, research science, and computer science. Design calculation, simulation, data analysis, and – remarkably – writing software itself. The head of Google has told investors that more than 30% of the company’s new code is now written by AI, and Mark Zuckerberg has said he expects AI to write half of Meta’s code within a year. The people who were supposed to be automating everyone else are automating themselves. (Hold that thought. I’m coming back to it because it’s bigger than it looks.)
The creative professions. This is the part I most need you to believe, because it challenges the comforting idea that creativity is inherently human and therefore safe from being replaced by AI. In its 2025 Future of Jobs report, for the first time ever, the World Economic Forum listed graphic designers among the fastest-declining jobs in the world – and named generative AI as the reason.
I gave you three stories about this in Part 1 of this report, remember? The Porsche ad, the muralist, the sign painter. But you may still be telling yourself that’s the flashy front end of the creative business. The artsy stuff. So let me tell you about my Number One Son.
He spent nearly 20 years in the movie industry as a senior executive at a post-production house – doing the painstaking, invisible technical work behind major films, colorization among it. Big pictures. Mission: Impossible was one of them. His company typically charged at least a million dollars for the work it did on a single film, and he earned in the high six figures. Two or three years ago he told me that his company’s future was in danger. He’d entered the industry just as it was switching from manual to digital finishing, so he had seen a technological wave up close before. “I’m seeing the same thing happening with AI that I saw when we switched to computer-based work,” he told me. “A process that used to require a highly trained specialist at a cost of hundreds of thousands of dollars was, within a 10-year period, being handled by a kid straight out of college on a laptop.”
He predicted, correctly, his own obsolescence.
I have spent 40 years watching industries rise and fall. I always watched from a safe and comfortable distance, because what I did – the analytical thinking, the creative writing – was untouched by every technological improvement of that era. Not this time. This time it came for my son and got him. And very soon it will be coming for me.
How My Son Cut Our Legal Bills by 80% – and What That Does to Every Professional’s Paycheck
Now let me tell you about Number Two Son. He runs the Ford Family Office. That means he keeps an eye on a dozen or so operating businesses, three sizable nonprofits, and the dozens of individual investments and deals that K and I have made over the years. With that many entities to look after, it won’t surprise you to hear that he is reading, or signing off on, some kind of legal document almost every day.
In my day, that kind of paperwork was done by a trusted attorney… at $800 an hour. But my son figured out that roughly 80% of legal paperwork is boilerplate – language lifted from the last deal, and the one before that. So now, instead of paying an attorney $800 an hour to read a 50-page document where 49 pages are boilerplate, he hands it to AI, gets back a clean summary and a list of the things worth flagging, and sends the attorney exactly one question about the single page that genuinely requires his experienced judgment.
Here’s what that means across the entire legal profession…
Take an ordinary commercial contract. A junior associate might spend six hours drafting it at $300 an hour, and a senior partner might spend half an hour reviewing it at $600 an hour. The client’s bill: about $2,100.
Instead, hand the drafting to AI. The machine produces the document in minutes. The only human step left is the senior partner’s half-hour review. The client’s bill: about $300. An 86% cut.
Move up to something meatier – a negotiated NDA for a Fortune 500 client. Junior work, six hours at $500. Senior review, half an hour at $2,000. Before AI: $4,000. After: $1,000 – a 75% cut.
Now take something genuinely hard – a complex deal that really needs 120 hours of research and 30 hours of senior judgment. Before AI: $120,000. Let the machine do the research and leave the judgment to the human, and the client’s bill falls to about $60,000 – a 50% cut.
Those numbers undergird the thesis of this entire report. The simple, routine legal job lost 86% of its value. The complicated, judgment-heavy job lost only 50%. The more a job is really just knowledge work, the closer its value falls toward zero. The more it depends on a trusted human exercising real judgment, the more of its value survives.
And notice – you just watched the answer to the question I asked above take shape: When knowledge stops being scarce, where does the value go? (Hold that thought.)
The Woman Who Beat Her Eviction with ChatGPT
Lynn White outside her RV home
Last fall, NBC News told the story of a woman named Lynn White who was facing eviction in California. She’d worked with a lawyer, and she lost. So she turned to ChatGPT, taught herself the procedure, represented herself in court, and overturned the eviction – escaping some $55,000 in penalties and another $18,000 in back rent. Her words: “It was like having God up there.”
A New Mexico business owner named Staci Dennett, served with a summons over a disputed debt, told ChatGPT to “pretend [it’s] a Harvard Law professor and rip my arguments apart,” then negotiated her own settlement – and the opposing attorneys actually wrote to compliment her.
Lawyers, as a profession, are not amused. Axios ran a piece about AI legal advice driving lawyers to distraction. One partner called it “the WebMD effect on steroids” – clients turning up convinced they already know the answer. He was right to be upset, and the reason is simple: The machine really did give Lynn White and Staci Dennett and countless others the answer.
I went, looking to be talked down.
The legal firms themselves feel what’s coming.
Clifford Chance, one of the largest law firms on the planet, is cutting around a tenth of its London staff and openly pointing to AI as part of the reason. Fifty of the hundred most prestigious law firms in America are the paying customers of a single AI legal tool now valued at $11 billion. And when a major new AI legal tool was released, the share prices of LegalZoom and Thomson Reuters dropped the same day – the market doing the arithmetic out loud.
Whatever you think about my argument, understand that the people with the most money at stake have stopped debating it. They are repricing it.
Now… how big does this get? And how fast?
To come up with realistic numbers, I did what any reasonable skeptic does: I went looking for the serious people, the scientists actually building these systems, certain they’d roll their eyes at the hype. I expected to come back reassured.
I came back more frightened than when I started.
Dario Amodei runs Anthropic, one of the handful of companies actually building the most advanced AI in the world. He warned last year that AI could “eliminate half of all entry-level, white-collar jobs and spike unemployment to as much as 20% in the next one to five years.” And he named the people in the blast radius: paralegals, payroll clerks, financial advisers, and coders among them. Axios, reporting his warning, called what’s coming a “white-collar bloodbath.”
Think about who is talking here. The man building the machine is telling you what he believes his own machine is going to do.
The World Economic Forum estimates that by 2030 some 92 million jobs will be displaced even as 170 million new ones are created – an enormous churn that lands hardest, and first, on exactly the clerical and knowledge roles we’ve been discussing.
And the damage is already showing up exactly where the theory says it should – at the bottom of the ladder, among the young. A Stanford study found workers aged 22 to 25 in the most exposed jobs have already seen their employment fall by double digits, while older workers in the very same roles held steady. The last hired are the first the machine replaces.
Keep in mind that this is a story of many chapters, and not all of them are bad. There is no doubt in my mind that virtually every major industry will be affected by AI. Some will disappear entirely. Most won’t. And once you look inside those industries – down to the individual businesses and the individual jobs – the picture gets harder to predict, because industries are not single entities that move in lockstep. They are thousands, sometimes millions, of people making individual decisions to continue with one thing or begin something else.
Keep in mind, too, the strongest argument against everything I’ve been telling you: 60% of the jobs Americans do today didn’t even exist in 1940. Every technological revolution before this one destroyed work and then created more of it.
But as I showed you in Part 2 of this report, all of those past technologies pushed people up – into knowledge work. This is the first one pushing in the other direction. And even if the optimists are right about new jobs eventually appearing, “eventually” is not much comfort to a 55-year-old radiologist, or a laid-off paralegal, or a computer-science graduate with a diploma and no offers during the 10 or 15 years it takes the new world to sort itself out.
The transition is happening now. But the timing, and the danger, is not equal for everyone.
The three clocks: Which one is ticking for you?
The first clock: the next six to 18 months. This one is for the jobs and businesses that are already basically destroyed – the local artists, the videographers, the sign painters, everyone whose product is knowledge work delivered digitally with nothing physical in the way. For them, the window to adapt to the new world is not five years. Unless they put into effect what they need to survive in the next six to 18 months, they will be left behind – possibly forever. I’ll tell you exactly what “adapting” means for them in the next installment of this report (and it is not what most of them think).
The second clock: 18 months to three years. This one is for the analytical professions – law, financial services, research, accounting, consulting. These businesses are already in the adapting phase, but they are not in danger of disappearing in the next year or two. What is in danger is their pricing (as my son’s attorney has already discovered). To stay in these industries and make decent money, the people in them must begin now to learn to take full advantage of AI. In practice that will mean producing five to 10 times more work in the same amount of time and at the same expense, just to make as much as they are making today. Standing still is a pay cut on a conveyor belt.
The third clock: three to five years. This one might surprise you, because it is ticking for the workers everyone assumes are the winners – the data scientists, the fintech engineers, the machine-learning specialists, the coders. The official forecasters will tell you these are the fastest-growing jobs in the world. The World Economic Forum’s latest report puts big data specialists and AI specialists at the very top of its growth list, and the US government projects data scientists to grow 33% over the decade.
I am doubtful, and here is why: Every one of those projections quietly assumes that today’s AI capability stays frozen for 10 years. It won’t. AI is self-learning. It is already writing a third of Google’s code. Entry-level tech hiring is already falling, and I believe the data and fintech people will be in the second stage of redundancy. In five years their numbers will be decimated, and their compensation will deflate right along with everyone else’s.
The first wave of the AI Revolution took the paralegals. The second wave comes for the people who built the machine that took the paralegals’ jobs.
The Next Great Transfer of Wealth: 2026 to 2030
I don’t need the most extreme version of any of this to be right. Strip out every speculative claim and what remains is still the largest economic event of your lifetime – because the value that AI strips out of the knowledge professions does not vanish. It moves. And it concentrates in the hands of the small number of people and companies that own and direct the machines.
I call it the Great Transfer of Wealth.
Here is one scenario of the way it might play out…
There will be a couple of dozen people/companies throughout the world who – in three to five years – will each have hundreds of millions of users for their services, users who got addicted to those services for free and then gradually grew accustomed to paying for them with tokens. The tokens will start out cheap, just like the federal income tax did, but will gradually increase. Those few dozen people/companies will have the money and power of developed nations – each with something like a trillion dollars’ worth of captured revenue that they can control freely. And it may be that the nations with the weapons will see this coming and try to take over their assets (mostly their followers). In fact, if you watch the fights over chips and the governments rushing to build their own “sovereign AI,” you can see that struggle starting now.
Why would this not happen?
I’ll give you the counterargument: The price of running capable AI has been collapsing – by one Stanford measure, roughly 280-fold in 18 months – and a dozen labs are competing furiously, with open-source models nipping at the frontier. Taxes ratchet because you can’t switch governments, but a user can switch chatbots in an afternoon. So maybe the token squeeze never gets as tight as I fear.
But notice what that counterargument does not touch. The thinking may be digital, but it runs on physical objects that a tiny handful of companies – and one contested island – know how to make, powered by electricity that has to come from somewhere, in buildings that have to stand on land. Whoever owns that physical layer collects a toll no competition can wish away. That is where the Great Transfer of Wealth flows, no matter how the chatbot wars turn out.
And there – if you’ve been keeping track – is the answer to that question: “When knowledge stops being scarce, where does the value go?”
When knowledge stops being scarce, the value obeys a kind of economic gravity. It drains out of whatever the machine can now do for pennies, and it flows toward whatever is still scarce – the judgment, the trust, the land, the power, the machines themselves, and the people/companies that own them.
The technology is the least of it. Follow the gravity.
Where the Vanished Trillions Are Pooling Right Now
So there is the reckoning: The professions we thought were the safest being exposed to automation first. The price of expensive thinking collapsing in my own family’s office and in a California eviction court. And behind it all the largest transfer of wealth in human history, already gathering.
If you’ve read this far and felt the floor move a little under you, good. That fear is not your enemy. It is the most useful thing you own right now, because here is what all of the bad news I’ve been telling you has been building toward…
Every dollar that vanished from the junior legal associate’s billable hours, from the radiologist’s salary, from my son’s post-production invoices, flowed somewhere – and it is pooling, right now, in a handful of places that most people are not looking at: in certain kinds of work, certain kinds of assets, and one particular kind of ownership that I have spent my entire 40-year career building. (Which is how I know exactly what it looks like.)
In the next installment of this report, I’m going to show you where the value went – which jobs and businesses will actually grow in this new world, what will never change no matter what the machines learn to do, and the surprising thing that will separate the people who come through this richer from the people who merely worked harder.
Meanwhile, I’m quietly putting together a small group of people who intend to be on the right side of this transfer and help each other get there while the getting is good. (More on that soon.)
The window is open. It will not be open long.
Economics
Two Quick Reads
“What Mamdani Can Learn from Hugo Chávez’s Government-Run Grocery Store Debacle”
“Venezuela’s government-run stores led to endless lines, empty shelves, quotas, and corruption.” Read more here.
David Stockman on “Why NONE is the Right Target for Inflation and Unemployment”
“When it comes to the Fed’s dubious 2.00% target for inflation and 5%, 4% or whatever % target for unemployment, no, Congress didn’t make them do it!” Read more here.
Politics
How Ohio Became America’s Political Bellwether State
This video explains how Ohio, by consistently reflecting the image of an “average American,” has become so influential in presidential elections.
Worth Considering
From the Red Bull Dance Your Style USA Finals
Can you believe the human body could move like this?
