My fellow pro-growth/progress/abundance Up Wingers in the USA and around the world:
Maybe this is an obvious point: If a) most “behind a screen” office tasks are fully automated by AI within the next 18 months, causing huge job market disruption, or if b) there’s a growing sense that self-improving systems are gaining real autonomy and slipping beyond human oversight, then, yes, c) AI will be the central issue in the 2028 presidential election. (Well, assuming no war with China over Taiwan.)
My working political assumption, however, is that the digital die has been already cast. Even if we’re still short of “you’ll know it when you see it” human-level AI, America’s technological future will be inescapable in the 2028 presidential election. And the debate will range well beyond the electricity use and water consumption of data centers.
How could it not, when predictions of imminent white-collar carnage have become a dystopian genre unto themselves? And that would be just the beginning. First, AI absorbs the white-collar staples: drafting contracts, managing projects, designing campaigns, analyzing accounts. Soon after, a professional bloodbath. Accountants? Lawyers? Middle managers? Everywhere redundant, sidelined, and obsolete.
Then what? Learn to code? Maybe learn to weld. And when the machines can do that, too? Pop on your VR goggles and hope your meager basic-income check doesn’t get delayed by a government shutdown.
But let’s take a step back.
The proper baseline here isn’t speculative fiction, even if conjured by AI company CEOs eager for new investment and high valuations, but economic history. General-purpose technologies—from electricity to the computer—have always disrupted work: restructuring firms and rendering skills obsolete. AI wouldn’t be much of a GPT if it didn’t do those things in a big way.
Yet the broader economic arc has been one of net gains. Productivity shocks reshuffle labor; they rarely annihilate it. There are more people creating movie special effects today than before CGI, for instance. When new tools reduce the cost of doing things, demand tends to rise for those things—and entirely new activities worth doing tend to emerge. Over time, employment expands and real wages rise thanks to higher labor productivity. Let’s start there, OK?
Tasks, not jobs
Today’s most extreme jobocalypse claims blur two distinctions. First, there’s the difference between capability and substitution. An AI that can draft a tight legal brief isn’t the same thing as a law firm that has reorganized itself to eliminate most of its lawyers. Jobs are bundles of discrete tasks: drafting contracts, yes, but also judgment, client trust, and making decisions that someone — not some thing — must ultimately take responsibility for. Replacing the full bundle across white-collar jobs across the economy means changes in regulation, risk allocation, and client expectations.
Consider coding. Today’s AI tools can already write a large share of software code, and an individual programmer can start using them almost instantly. But that doesn’t mean companies can suddenly operate with far fewer engineers. Inside big organizations, adoption takes time. The tools have to pass security and legal review, managers have to approve budgets, workflows have to change, and thousands of employees have to be trained.
As Anthropic CEO Dario Amodei has put it, there are effectively “two exponentials”: one in AI capability and another, slower one in economic diffusion and adoption.
New work emerges
Second, there’s the distinction between old and new tasks. Research cited by Goldman Sachs, drawing on work by David Autor and co-authors,1 finds that roughly 60 percent of American workers today are employed in occupations that didn’t exist in 1940 — implying that more than 85 percent of net job growth over the past eight decades has accrued to entirely new occupations. Over just the past 30 to 40 years, digital technologies have created or enabled more than 6 million computer-related jobs and another 8 million to 9 million roles in industries such as e-commerce, the gig economy, and digital content creation. Predicting which present jobs will be radically changed by a new technology is hard, but not nearly as difficult as predicting the future jobs that entrepreneurs will create.
The human premium
Push the thought experiment to its limits. Suppose artificial-general-intelligence-powered humanoid robots become near-perfect substitutes for white- and blue-collar labor. Would that utterly empty offices and factories? Not necessarily. When societies grow richer, they tend to demand more experiential and relational goods. The “human touch” often becomes more valuable as people get richer: Consumers pay for live music rather than recordings, personal trainers rather than apps, boutique service rather than self-checkout. Tomorrow: Human caregivers for elderly parent, not the latest CareBots.
Nor is work merely a subsistence mechanism. It’s a source of identity and meaning. A society in which millions suddenly became economically unnecessary wouldn’t simply adjust quietly—it would feel politically destabilizing because work is bound up with dignity and status. (Not to mention the Great Digital Depression.)
That intuition aligns with the skepticism of Ben Jones, a leading growth economist. Responding to the prospect of sustained, warp-speed growth of 30 percent annual from AGI or superintelligence, he argues it would require extraordinary “creative destruction,” rendering both physical and human capital obsolete at unprecedented speed. Owners of “vintage capital stock” and workers with “relevant vintage human capital,” he writes, “won’t want to see their capital assets or skills become worthless,” and would likely try to block such advances. Governments already struggle with worker displacement at ordinary growth rates. The politics of churn “beyond all historical precedent” are hard to imagine, Jones adds.
Choice, not collapse
Perhaps the more interesting question, then, isn’t whether jobs will exist but whether people will need them. If AI ushers in extreme abundance, reservation wages—the minimum pay people require to work—may rise dramatically. Some may opt for part-time roles, while others may pursue only truly meaningful or highly paid positions. The labor market could contract voluntarily.
As my AEI colleague Michael Strain, a labor economist, has put it:
Do human beings have a deep desire to contribute to the world or not? Are business leaders interested in and able to make productive use of the enormous talents, capacity, and energy of two-thirds of American adults? If yes and yes, then AI won’t take all our jobs.
The robots may indeed grow capable of doing almost everything. That doesn’t mean we’ll have nothing left to do. I hope that message makes it into the emerging national debate.
1 As Autor also argues, AI’s distinctive potential is not merely to replace expertise but to spread it. By combining rules, data, and experience into practical decision support, AI can help workers with solid training take on higher-level tasks once reserved for elite professionals. Think nurses handling more diagnostics, or paralegals drafting complex filings. The result is an expansion of the labor market’s middle, not a hollowing out. Early evidence from coding, customer support, and professional writing points in the same direction: AI tools tend to raise the performance of less-experienced workers toward that of their more seasoned peers, compressing skill gaps rather than eliminating jobs.
