The AI Wealth Pitch: Does It Really Benefit Workers?

Ramaswamy’s headline promise in the Wall Street Journal is seductive: workers can build wealth in the AI era. Here's the thing: that’s a thesis I want to be true. But there’s a difference between saying “you can” and building a world where that’s not just technically possible for a sliver of people with great timing and a Delaware C-corp lawyer on speed dial.

Let’s give the column its due first. It pushes a claim too many tech and political types dodge: people shouldn’t be left behind by automation. That’s better than the usual “learn to code” shrug. The piece channels a kind of American bootstrap optimism—if you reskill, adapt, and plug into the AI boom, you won’t just keep a job, you’ll build wealth.

Sure, but optimism is not a capital structure.

The argument treats “workers building wealth” as mostly a matter of willpower and retraining, not a question of where the returns from AI actually land. Talk about hustle culture and reskilling is comforting; it would make an excellent TED talk. It doesn’t, on its own, rewire who owns the pipes, the platforms, or the equity.

Who gets the rent?

Ramaswamy’s framing leans on a familiar Silicon Valley assumption: markets will re-price labor and capital so that new winners emerge across the board. In theory, that’s how Schumpeterian disruption works. You blow up the old incumbents, and in ride the founders with laptops and an API key.

And yes, in a world where a tiny team can ship something valuable in a weekend with AI copilots, you do get new openings. The indie developer using a large-language model as a tireless cofounder is real. The solo consultant who suddenly looks like a 10-person shop is real.

But look: software has a long track record of concentrating returns. Platform dynamics and network effects mean the lion’s share of productivity gains often flows to the owners of the network, not to the users generating the value. Uber drivers don’t own Uber. Creators don’t own YouTube. People training models with their labor and data generally don’t own the models.

If you want a science-fiction mirror, Charles Stross’s Accelerando imagines acceleration that mints enormous winners and leaves everyone else bartering for scraps in the shadow of post-human capital. It’s a cautionary tale, not a product roadmap.

History rhymes here. The industrial revolution didn’t just create wealth; it created concentrated fortunes and a long stretch where many workers lost bargaining power before labor law, unions, and social insurance caught up. I’ll be honest: rhetoric about “workers building wealth” can’t substitute for a hard look at who gets equity, who gets profit shares, and who just gets a log-in.

Education is necessary but insufficient

A big part of Ramaswamy’s optimism rides on retraining and adaptability. On the basics, I’m with him. Skills matter. People using AI will beat people who aren’t using it.

But training is not the same thing as capital.

You can train a huge cohort of workers to be power users of AI tools; that doesn’t mean they’ll own the algorithms or the platforms that monetize those skills. We’ve already lived through a version of this with the “learn to code” era: plenty of newly minted developers ended up in contract roles with little security while equity stacked upward to founders and a narrow band of early employees.

Policy levers matter if the goal isn't just to keep people employed but to give them a direct line into the upside. That could mean expanding meaningful employee ownership models beyond the usual tech darlings. It could mean tax and regulatory regimes that don’t punish small, employee-owned firms while making it easy for giant incumbents to roll up markets. It could mean default profit-sharing schemes in AI-intensive sectors, the way some manufacturing firms built gain-sharing into their DNA.

These ideas aren’t sci-fi; they’ve been floating around policy circles for years. What’s missing in the column is the friction: without ways to translate productivity gains into household balance sheets—equity, profit-sharing, low-cost capital—skills become a ladder leaning against the same glass wall.

The time lag problem

There’s another issue: time.

Retraining takes sustained effort. Markets reprice in real time. A worker pushed out of a stable role by an AI deployment doesn’t magically get a two-year runway to discover the perfect “AI entrepreneur” program.

If you want people to gamble on new skills and new ventures, something has to cushion the landing: income support during transitions, subsidized training, or models where companies deploying automation co-fund the reskilling of people they displace. Yeah, no one wakes up excited about bureaucratic programs. But the market-only answer assumes a level of frictionless mobility and savings that just doesn’t show up in actual households.

Who falls through the cracks

Ramaswamy paints a broad possibility but glides past how uneven the exposure will be.

Some fields—software, certain professional services, parts of creative work—really could see democratized gains. A smart paralegal wielding AI tools could spin up a niche legal-research firm; a mid-career marketer might turn into a one-person agency with decent margins.

Other sectors look more like a straight substitution problem. In logistics, routine clerical work, and parts of customer support, AI isn’t a sidekick—it’s a replacement. Those workers face not just a skills gap but a geography and network gap. Cities with capital, dense tech ecosystems, and social networks wired into new AI ventures will have on-ramps. Rural counties and smaller businesses without those connections will mostly be AI customers, not AI owners.

That’s not a conspiracy theory; that’s how capital and networks tend to behave.

The counter-argument—and the missing institutions

A fair rejoinder to all this is the classic market story: technology kills some jobs, creates others, and raises living standards over time. We saw that with electricity, with cars, with the internet. Entrepreneurship, in this view, remains the great equalizer if you’re willing to jump.

The problem isn’t that this story is totally wrong. The problem is that it quietly assumes everyone has a ticket to the next act.

It also assumes that the institutions that softened prior transitions—unions, broad-based public education, progressive taxation, social insurance—will either reappear in upgraded form or aren’t needed this time. That’s a bet, not a plan.

A more grounded version of Ramaswamy’s thesis would put ownership on equal footing with education. It would treat worker equity, co-ops, platform cooperativism, and portable benefits less like afterthoughts and more like infrastructure—closer to how we think about roads or broadband.

Ramaswamy’s column plants a flag on worker opportunity in the AI boom. Whether that headline ages well will depend less on how many people learn to prompt a model, and more on who gets a share of the models they help build.