Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
There is a new genre of finance video doing the rounds, and you have probably seen one. A familiar face, no studio lighting, an apology for the rough production because the message simply cannot wait. The warning is that artificial intelligence is about to lock the economic ladder in place, that ordinary people have a few short years to get onto the right side of it, and that the way to do that is to start an AI side business now. Somewhere near the end, a guest appears with a free event.
The format works because it is not built entirely on fiction. The statistics quoted in these videos are, for the most part, real. That is exactly why they deserve a closer look. When a true number is used to sell a conclusion it does not support, the reader who checks the source usually comes away with a very different picture, and often a better trade.
The Numbers Are Real, the Framing Is Not
Start with adoption. Anthropic has been publishing an Economic Index built from anonymised Claude conversations, mapped against the US government’s catalogue of occupational tasks. Its first report, using January 2025 data, found that 36% of occupations showed Claude being used for at least a quarter of their tasks. In the January 2026 update on its economic primitives, pooling data across reports pushed that figure to 49%. That is a genuine jump. It is also a measure of where the tool shows up in conversations, not of how many jobs it has replaced. The same report found that once you adjust for whether Claude actually succeeds at a task, the ranking of the most affected jobs shifts considerably, and the company’s own productivity estimate falls by roughly a third.
Then the labour market. The Stanford Digital Economy Lab’s August 2026 revision of its “Canaries in the Coal Mine” research, built on ADP payroll records, found that employment for workers aged 22 to 25 in the most AI-exposed occupations now sits about 19% below where it would have been had it tracked less exposed peers of the same age. A year earlier the gap was 15%. Those are sobering numbers. But the very first of the paper’s six headline findings is that there is no widespread, economy-wide displacement. The authors are also explicit that the patterns are descriptive rather than causal, that the gaps narrow once education is accounted for, and that some divergence was already visible before generative AI arrived.
Finally, the split of the economic pie. The US Bureau of Labor Statistics reported that the labour share of nonfarm business output fell to 52.8% in the second quarter of 2026, the lowest reading in a series that starts in 1947. For most of the post-war era it sat between 60% and 66%. The slide began around the turn of the century, long before anyone was prompting a chatbot.
Put those three together honestly and you get a story that is real but slow, uneven and concentrated. What you do not get is a countdown clock.
Who Is Actually Getting Squeezed
The most useful detail in the Stanford work is not the headline gap but the mechanism behind it. The adjustment is coming through reduced hiring of young people, not through firing the people already employed. Experienced workers in the same exposed occupations show no comparable shortfall.
The researchers offer a plausible explanation. Roles that lean on codified knowledge, the kind that can be written down, taught from a textbook and reproduced from documentation, are where young workers are losing ground. Roles that depend on tacit knowledge, the judgement built from years of watching real situations go right and wrong, are where experienced workers are holding steady or growing. Occupations where AI tends to complement people rather than automate them show flat or rising employment.
That matters for anyone planning a career or advising a son or daughter about theirs. It suggests the risk is less a matter of AI taking your job and more a matter of AI removing the bottom rungs where people used to learn the job in the first place. It is a pipeline problem. Pipeline problems play out over a decade or more, not in a single five-year sprint, and the response they call for is about building judgement and experience, not about racing to launch a side hustle before an invented deadline.
Urgency Is the Product
Notice the structure of the pitch. A grave warning, a personal history of spotting the internet and social media early, a sense that the window is closing, and then a guest with an event that costs nothing to attend. Anyone who has watched financial marketing for long will recognise the shape. Urgency is not incidental to that kind of funnel. It is the thing being sold, because people under time pressure skip the due diligence they would otherwise do.
The figures offered as proof deserve the same scrutiny as the macro data. Aggregate “tracked earnings” across every attendee of a programme tell you almost nothing about what a typical participant earned, what they spent to get there, or how many earned nothing at all. A large total can be driven by a handful of people who were already running sizeable businesses before they walked through the door. Even the presenters in these videos tend to concede, briefly, that not everyone made money. That concession is the most important sentence in the whole pitch.
US regulators have been circling this exact problem. In January 2025 the Federal Trade Commission proposed extending its Business Opportunity Rule to business coaching and other money-making opportunities, which would require sellers to hold written substantiation for any earnings claim and hand it over on request. The proposal passed on a 3 to 2 vote, and one of the dissenting commissioners now chairs the agency, so its future is far from certain. The logic behind it stands on its own regardless. If someone quotes you an earnings figure, ask for the median outcome and the share of participants who lost money. If they cannot or will not supply it, you have your answer.
The Contrarian Problem With Owning the Winners
Strip away the side hustle and the video’s underlying thesis is one that plenty of serious investors share: capital is taking a larger slice of output than labour, so own capital. The data supports the first half. EY-Parthenon’s chief economist Gregory Daco, speaking to Fortune in September, put corporate margins at a record 14.9% of GDP alongside that record low labour share, and said he does not see an obvious floor under labour’s slice. William Blair’s economists made a similar point in their Economics Weekly on record profits and lagging pay, noting that hourly earnings growth has slowed to its weakest pace since 2021 even as margins sit at their highest since 1947.
Here is the contrarian catch. When a thesis is this widely understood, it is usually already in the price. Investors buying the dominant owners of AI infrastructure today are paying for margins at record levels and implicitly assuming they hold or expand further. Bridgewater laid out the risk in a 2019 paper on peak profit margins, estimating that US equities would have been roughly 40% lower without two decades of margin expansion, and warning that a reversion toward historical norms would leave valuations badly exposed. The forces it identified then, weaker labour bargaining power, lower corporate taxes and rising concentration, are the same forces pushing labour’s share down today.
There is also a political dimension. A record low labour share paired with flat real wages is precisely the type of statistic that draws the attention of tax authorities and regulators. The Fortune analysis notes that if more of each dollar of output flows to data centre owners and their shareholders, the tax base grows more slowly than the economy itself, which complicates the fiscal arithmetic for governments already carrying heavy debt. Margins that are partly a product of policy can be partly reversed by policy. Owning the winners at any price is not the same trade as owning productive assets at a sensible price.
What a Sensible Response Looks Like
None of this means the right answer is to do nothing. It means the right answer looks far less dramatic than the video suggests.
[Also Read: Greed Is Not Good]
First, own productive assets, steadily and broadly. The case for holding a stake in capital is strong, and it does not expire in five years. Regular contributions into diversified, low-cost holdings capture the shift toward capital without forcing you to bet everything on a narrow cluster of AI names at peak margins. Valuation discipline matters more, not less, when everyone agrees on the story.
Second, protect and deepen your tacit knowledge. If the evidence shows experienced workers who use AI as a complement are holding their ground, the move is to become one of those workers. Use the tools heavily in your existing role, become the person who can tell when the output is wrong, and invest in the kind of judgement that is hard to write down.
Third, treat any side venture as an experiment with capped downside. Building something small with AI tools can be a perfectly sensible way to learn. Paying for courses, masterminds or coaching on the strength of aggregate earnings screenshots is not. Set a budget you can afford to lose, measure actual net profit rather than revenue, and walk away if the numbers do not show up.
Fourth, ignore deadlines you did not set yourself. Long-term wealth is dominated by time in the market and the avoidance of large mistakes. A self-imposed sprint is the ideal environment for large mistakes.
The Clock Is a Sales Tool, Not a Forecast
The research behind this debate is genuinely important. Entry-level hiring in exposed occupations is weakening, labour’s share of output is at a record low, and AI use is spreading across a growing range of jobs. Those trends deserve attention from anyone thinking about their career and their portfolio.
But the researchers themselves describe a slow, uneven and still uncertain shift, and none of them put a date on it. The only party in this story with a precise deadline is the one with something to sell. A countdown converts a real long-term trend into short-term fear, and short-term fear is rarely profitable for the person feeling it.
The better position is unglamorous: keep accumulating assets at reasonable prices, keep building skills that are hard to automate, and keep your wallet closed whenever someone tells you the window is about to slam shut.
