View this email in your browser

76report

March 7, 2026

Will AI Destroy the Economy?

With growing concern that AI could trigger a shock to the labor market unlike anything seen in modern history, markets have been grappling with some deeply disturbing questions in recent weeks.

Is AI likely to wipe out large portions of the white collar workforce?

And, if so, could this take down the entire U.S. economy?

Yesterday’s jobs numbers did nothing to help alleviate these fears. Nonfarm payrolls saw a decline of 92,000, versus consensus estimates of job growth around 60,000. The unemployment rate edged up slightly to 4.4%.

The Employment Situation Report from the Bureau of Labor Statistics was noisy. Strike activity in the health care sector had an outsized influence on the figures. But the relative softness in hiring underscores the idea that real structural change may indeed be underway.

A few years ago, questions like these might have sounded ridiculous. Now, they make intuitive sense.

AI chatbots are capable of generating solutions to virtually any complicated problem you throw their way—legal, financial, medical.

And as for the millions of well-paid men and women who report to office buildings or log into their laptops at home every morning… what are they really doing that AI cannot?

In the most basic sense, they are processing and responding to information they receive on their computers. This is exactly what AI does best… but with no coffee breaks, no vacations days, no typos, and no health care benefits.

As AI replaces human labor, it logically follows that companies will need to employ far fewer workers to bring their goods and services to market.

So who will have any money to purchase all these goods and services?

A hyper-productive, machine-driven economy sounds great… but 80% of the American labor force, some 120 million people, consists of service sector workers.  

The doomsday scenario

Attention to this topic intensified late last month following the release of a widely circulated note by a previously below-the-radar group called Citrini Research. Published on February 22, The 2028 Global Intelligence Crisis quickly went viral across Wall Street and Silicon Valley.

Written as a scenario analysis rather than a forecast, the piece imagines a near-term future in which increasingly capable AI systems begin replacing large numbers of white-collar workers. As layoffs spread, consumer spending weakens, prompting companies to adopt even more AI to cut costs.

The result, in the scenario, is a self-reinforcing economic loop: falling employment leads to weaker demand, which accelerates automation and pushes unemployment even higher.

The essay captured attention because it framed AI not as a productivity-enhancing technology wave, but rather as a potential macroeconomic shock—and one that our current system is not built to withstand.

Citrini’s creative writing exercise struck a chord, but it was not the first apocalyptic outlook on AI that has kept investors up at night.

Anthropic is the very rapidly growing private company that gave the world Claude—a leading AI model with surging revenues that are now closing in on $20 billion annually. Its founder and CEO, Dario Amodei, has been in the news lately for his skirmish with the Trump administration over the military’s access to Claude.

But Amodei has also been known for making bold forecasts about AI-driven job losses. Last May, he alarmed the investment community after predicting that AI could eliminate as much as half of all entry-level white collar jobs, driving unemployment rates to as high as 20% in five years.

Is it happening yet?

Ironically, it was economic researchers at Anthropic who just this week released a relatively reassuring report on AI’s impact on jobs. Labor Market Impacts of AI: A New Measure and Early Evidence is now gathering a lot of attention as well.

Despite widespread fears of job displacement, the report actually finds little evidence that AI has significantly affected unemployment so far.

Comparing unemployment trends across occupations since the launch of ChatGPT in late 2022, the researchers find that jobs that are “highly exposed” to AI disruption have not experienced meaningfully higher unemployment than less exposed ones.

There is, however, one subtle signal worth watching.

Hiring into highly exposed occupations—particularly among workers aged 22 to 25—appears to have slowed slightly in recent years. The study suggests the earliest labor market impact of AI may show up not as layoffs, but as fewer entry-level opportunities.

The Citrini report and the Anthropic study highlight the difference between AI’s theoretical effects and its current economic reality. The technology may eventually reshape large parts of the labor market, but for now the data suggests the transition is unfolding slowly.

But just because AI-related job displacement has not really happened yet, this does not mean it never will.

For the most part, we are in the early phase of AI where compute capacity is being built out and models are being trained.

Over the next several years, the economy will transition into widescale AI deployments. That is when the job displacement may truly kick into gear.

Basis for optimism

While we take these uncertainties seriously, we remain optimistic about the long-term impact of AI on the economy and markets. We continue to see AI as a technology that will drive productivity growth, which will in turn drive real economic growth and wealth creation.

AI is deflationary—in a good way. AI enhances the efficiency of businesses and thereby makes goods and services cheaper to produce.

Uncontrolled deflation caused by economic disruption and financial crisis (like the Great Depression) can lead to a dangerous downward spiral. This is the high-risk, deflation shock scenario Citrini describes.

But deflationary pressure, i.e, falling prices, is precisely what gives central banks the ability to print more money and inject more liquidity into the economy, boosting nominal growth.

Crucially, from an investor standpoint, a more dovish central bank means higher asset prices. Anticipation of this dynamic, as we discussed in late January, has potentially even contributed to rising gold prices (Gold Hits $5,000: This Is Scarcity, Not Risk Aversion).

To be fair, we are still very much in the early stages of the AI revolution. Vast uncertainties remain, and the more pessimistic scenarios need to be examined carefully.

Credible voices in the investment world, who are quite wary of the risks ahead, should not be dismissed.

A real life debate

To provide a window into how one of these voices is processing all these risks, we decided to share a text exchange we recently had with a prominent investor.

Trish and Rob have many valuable contacts across the financial world and regularly draw upon their insights.

What follows was a real-time exchange with this highly accomplished investor, who has responsibility for several billion dollars of assets. He prefers to remain anonymous, but we will use the alias David.

David:

Did you guys see this? Today’s sell-off triggered by this article imagining a world in 2028. [Link to The 2028 Global Intelligence Crisis]

Rob:

Yes. The missing piece is that if people lose their jobs to AI and therefore their consumer spending power goes down, the money doesn’t disappear. It becomes higher corporate profits, which gets recycled elsewhere, or lower prices (enabling easier monetary policy).

What this scenario in fact describes is extreme productivity growth—more output per worker. It’s exactly what you want: non-inflationary growth.

Imagine the reverse happened. Firms need to hire more workers to produce the same output, which happens with heavy regulation, for example. You get stagflation.

David:

I don’t see how mass unemployment of white-collar workers is a positive. Prices will come down, productivity goes up, some companies will thrive (semis), but if the consumer economy breaks and all the assumptions underpinning it, that’s a serious problem.

And the point on governments relying on income-tax receipts of high earners is important. Plus Social Security relies on new workers funding retiring ones. That system is going to break anyway without this.

Rob:

They ultimately won’t be unemployed. Wealth will be created. They will work elsewhere. It’s like saying factory automation is horrible for the economy because all those factory workers will lose their jobs.

Corporate profits will either soar and be reinvested in the economy one way or another, or prices will come down and interest rates will go to zero. The first-order effects are easy to see, but the second- and third-order effects are more abstract but no less real.

David:

There is no place for knowledge workers to work if AI is all-knowing.

Rob:

If health care for example gets cheaper because it’s more efficient I can spend less on health insurance and take more vacations.

David:

And I don’t buy that we will just supervise AI.

Rob:

Yes, everyone becomes a retiree to some extent. I think ultimately it’s good for stocks: earnings up, interest rates down. The Fed and the government will take care of aggregate demand.

There will of course be winners and losers as there were with factory automation. All consumers win. A sliver of the population is screwed.

Here the sliver may be larger, but there will be opportunities created by all this wealth. You just can’t argue productivity growth is bad. That’s totally irrational. It implies negative productivity growth is good.

The economy breaks when unemployment is too low. In 2022, it was 3.5%, like an all-time low. The result was stagflation, interest-rate hikes, and a market crash.

A separate concern may be that people are getting way ahead of themselves talking about 5–6% annual real GDP growth. A massive pickup is like 1% above the long-term trend of 2%. A 5% baseline is bonkers.

At the end of the day prosperity is how much stuff your economy can produce. AI means we can produce more stuff.

David:

This is productivity hitting white-collar jobs. Industries that structured themselves to benefit from productivity but never bear the costs because they protected themselves through regulation, complication, brand, and “moats.” This is different.

While stocks may benefit from lower rates and higher margins initially, many of these businesses become commoditized, which dooms profits. Especially when their supply is limitless and demand is finite.

Rob:

You are describing heaven on earth. We simply ask Claude for anything we want and it delivers. No middlemen extracting rents. The key is to figure out which moats survive AI. Physical ones may be most resilient. Railroad tracks, for example.

David:

And populism will come back to bite on this. There won’t be sympathy for the out-of-work lawyers.

Rob:

Affluent boomers make out like bandits. Already retired sitting on 401ks. The challenge will be for younger people to acquire wealth. Although there will be more wealth to acquire.

Honestly, I see massive growth in wealth management broadly writ. What else does society do when machines do everything other than fuss over their wealth?

Trish:

Just catching up on this! Very interesting article here. Jamie Dimon just dismissed fears over how AI will hit JPMorgan, by the way.

So, who is the sliver that is screwed? I would think it’s the middle-class desk worker. They aren’t a sliver though. They’re the underpinning of an entire class.

Rob:

I guess I would argue that people will ultimately work. Roles will change. Some people will be stranded with useless skills that were once valuable. More wealth means more overall demand.

David:

Someone who works in a call center is accustomed to ups and downs and periodic unemployment and hardship. Lawyers and accountants are not. They won’t be able to handle it because they’ve never been forced to. And cannot adapt.

Rob:

Radiology has become a case study for why AI won’t replace human workers. As sophisticated services get cheaper, demand for them will grow. You are counting one side of the ledger—the negatives—and ignoring the positives.

Basically the argument is that a little productivity growth is good but too much is catastrophic.

David:

I bet the positive health benefits from disease detection and prevention are outweighed by mental health destruction of unemployed aimless people.

Rob:

There is no example in economic history of a productivity boom leading to depression.

David:

It’s like Hillary telling coal miners: “Don’t worry, you will be retrained and better off.” My argument is that AI’s impact and speed is unprecedented. So relying on economic precedent may not be reliable.

Trish:

My own great-grandfather went into severe depression during the Depression after losing his job at the mill in Manchester. He couldn’t get out of bed. His daughters dropped out of school and got work waitressing. He died a few years later.

He was what we’d now call clinically depressed. I bet this happens a lot.

It’s why there’s a bad opioid addiction now with middle aged men across America. Much of the fentanyl crisis and increase in mortality rates in the USA is attributable to diseases of despair.

David:

I support and welcome productivity. But the consequences of this AI-driven one are directed squarely on the upper middle class.

Rob:

In 1980, clerical workers and secretaries made up roughly 18% of the U.S. labor force. Since then the number of secretaries has dropped significantly. Office support roles fell from 12.7% of all jobs in 1980 to only 6.8% by 2022 due to technology and automation.

Trish:

I’ll tell you. I could use a secretary. A real one. Like they had in the ‘50s. That’s a lost art!

David:

Estimated active lawyers in the U.S. in 1980 — ~574,000. In 2024/2025 — ~1.32–1.37 million. The number of lawyers in the U.S. has more than doubled since 1980.

Rob:

Despite LexisNexis! More wealth equals more money to fight over. Which means more lawyers. The more a lawyer can do, the more a radiologist can do, the more their services will be sought.

Trish:

This is like the argument people make with Microsoft Excel and bankers. Tech tools make analysts at investment banks far more productive. As tech got deployed, many more analyst jobs were actually created, not fewer.

So it’s a good case study. When workers can get more done, more workers get hired, at least when you look at Wall Street over the past 30 years.

Rob:

Also, growth in financial advisors has happened despite index funds and robo-wealth tools. The number of financial advisors in the U.S. has basically doubled in the last 30 years.

David:

I think the model is flat-screen TV pricing over time. Cheaper. More abundant. Commodity.

Rob:

You are talking about unit pricing falling. Yes. The essence of productivity gains. But unit sales of televisions have tripled since 1980.

David:

Don’t get me wrong. Lawyers and Wall Street will do whatever they can to protect their turf. Seeking laws requiring humans do the work and provide advice. Politicians will likely comply.

Rob:

The wealth—the claim to production—is still there. And has only grown. It gets recycled into new domains. People with certain highly specialized skills are at risk, though.

Trish:

The problem is this could force the income inequality issue big time. Only 62% of the population owns stocks. 38% does not. To the point about political populism.

Rob:

I agree this favors capital over labor. Even before AI we had this problem.

Trish:

We should publish this exchange. With a note at the end summarizing that people should invest. Very entertaining exchange!

The need to be invested

As Trish mentioned, one of the core dynamics at play here is a rotation of economic value from labor to capital (technological equipment).

AI likely means that businesses in general become more efficient and profitable, precisely because they need fewer workers. Individuals should therefore aim to become business owners, i.e., owners of stocks.

Of course, not all stocks will succeed. Certain business models will simply disintegrate. Others will become more valuable. This happens with every technology wave.

We view the long-term impacts of AI as a net positive for stocks, while also acknowledging real uncertainties.

At the same time, we are highly aware that AI is probably going to be the most significant structural factor that drives investment outcomes in the years ahead.

So we are bullish on the long-term trajectory for stocks—but also highly attuned to the need to determine which stocks are likely to benefit from AI disruption, and which are likely to struggle, as we manage our Model Portfolio allocations across all sectors.

Stock picking, as opposed to broad market exposure, may matter more now than ever.

Click HERE to learn more about our Model Portfolio subscription plans.

Disclaimer

FOR SUBSCRIBER USE ONLY. DO NOT FORWARD OR SHARE.