Who Will Pay the Rent

At dinner in Indore, fifty-two-year-old Prakash listened to his engineering-student son declare that a universal basic income had become inevitable. Prakash drives an auto rickshaw, and his counter was immediate and practical: the government cooking gas subsidy already reaches his bank account, so the real issues are who will pay for it and will it cover the rent. That exchange reaches right into our listener's question about whether artificial intelligence will take our jobs, which careers stay relevant, and what happens to missed income when high-skilled roles are replaced.

Untangling that begins with task automation versus occupational extinction. Researchers emphasize that labor effects depend on how technology reshapes internal duties and skill demands rather than simply wiping away payroll titles. In an analysis of Claude interactions, researchers found that fifty-seven percent of usage augments human work, helping people learn or refine drafts, while forty-three percent automates tasks with minimal intervention. Technical capability alone does not equal commercially viable displacement once error rates and integration expenses come due, and researchers report that within-job redesign accounts for roughly forty percent of declining task exposure as employers reallocate daily duties.

Yet the scale of exposure is vast across the workforce. A study cited by Federal Reserve researchers estimates that four in five American workers belong to occupations where large language models could alter at least ten percent of work tasks, while nearly a fifth could see over half their duties affected. Across borders, the International Monetary Fund estimated that artificial intelligence could affect roughly forty percent of jobs globally, substituting for some while supplementing others.

Earlier waves of automation show that economic reabsorption carries severe friction. Economists long maintained that technological progress avoids permanent joblessness by generating fresh industries and lowering consumer prices, but the displaced rarely step directly into those new openings. A study of industrial robotics showed that adding one machine per thousand workers reduced the employment ratio by zero point two percentage points and cut wages by zero point four two percent. When mechanical switches replaced telephone switchboard operators, researchers found that a decade later, those women tended to remain trapped in lower-paying jobs or had dropped out of the labor force entirely.

Advocates of labor absorption point out that economy-wide destruction has not shown up in aggregate data. Researchers at Stanford found no evidence of widespread job loss through mid-2026, a finding mirrored by the International Labour Organization and the Yale Budget Lab. A Gallup survey in early 2026 found only one percent of laid-off workers blamed automation. Roger Lee, founder of Layoffs.fyi, observed that there is little evidence software directly replaces the staff let go, adding that he believes tech firms are redirecting budgets toward computing infrastructure while trimming headcount elsewhere. The broader disruption is gathering, but the real bottleneck is where the hiring door is actually beginning to close for younger workers.

Where that hiring door is closing turns out to be precisely where previous technology never reached. For generations, automation moved through factories and loading docks, sparing the office. Generative systems break that pattern by automating complex cognitive and linguistic duties across medicine, finance, and information technology. Researchers Jacob Dominski and Yong Suk Lee observe that roles relying on complex reasoning face larger drops in full-time employment, while manual physical work appears less affected. Looking at early outcomes, proponents of redistribution highlight findings that college graduates in the most exposed majors faced a thirteen percent penalty in initial earnings, while advocates of displacement point to Stanford estimates that employment for exposed junior staff trailed less-exposed peers by nineteen percent.

The pressure is mounting because these tools are no longer passive chat screens. By May 2026, two-thirds of responding Texas businesses reported using AI. In the wider economy, white-collar payrolls have contracted for twenty-nine consecutive months, a stretch economists describe as unprecedented outside a recession. Models have advanced from answering simple queries to autonomously completing complex human workflows, and layoffs tracker Roger Lee reported that AI was cited in about a third of tech layoff events in 2026.

That shift fuels the argument that displacement is outrunning the labor market. Advocates of displacement point to research showing early-career employment fell twelve percent across exposed industries over the ten quarters following ChatGPT's release, driven primarily by reduced hiring. Dallas Fed economists Samuel Dodini and Tucker Smith documented an eight to nine percent drop in exposed postings by early 2026, citing “strong evidence that GenAI has decreased labor demand” for automatable tasks. Anthropic CEO Dario Amodei warned that “AI could wipe out half of all entry-level white-collar jobs – and spike unemployment to 10-20% in the next one to five years,” while an economic model by Brett Hemenway Falk and Gerry Tsoukalas warns that individual firms capturing automation savings externalize demand losses onto competitors.

Yet advocates of labor absorption see a very different trajectory unfolding. The Bureau of Labor Statistics projects total employment across all occupations to expand by 5.2 million jobs through 2034, anticipating that adoption of AI technologies will fuel strong job growth in computer and mathematical roles. Internationally, an International Labour Organization review of surveys across seven nations concluded that large-scale job displacement remains limited. Examining more than one billion job postings across twenty-seven countries, PwC reported that companies best able to use AI expanded headcount significantly faster than the least-exposed employers.

Displaced workers and recent graduates are adapting in real time. Researchers find that labor demand adjusts not only through shifts in hiring, but through within-job redesign, which reportedly accounts for 39.5% of overall reductions in AI exposure. When hiring cools, young workers seek shelter in school; exposed graduates were reportedly more likely to enter graduate programs within a year, and two-thirds of returning computer science majors chose advanced degrees in the same field. As researchers note, navigating these dislocations will require job-search assistance, wage insurance, and targeted retraining where viable career pathways exist—raising the question of which professions and skills are proving durable enough to withstand the pressure.

The durable work begins on the ground. While models replicate cognitive routines on a screen, studies by the Dallas and Philadelphia Fed find administrative roles more affected, with the Dallas study noting that the skilled trades are relatively less exposed, and an analysis by Jacob Dominski and Yong Suk Lee points out that occupations centered on manual physical tasks appear less affected. According to researchers summarizing Kording and Marinescu, software tends to substitute for virtual, intelligence-heavy tasks, nudging workers toward roles that demand physical human presence. In fact, a Federal Reserve Beige Book report noted ongoing wage pressures for field technicians servicing oil and gas equipment, where hands-on troubleshooting cannot be handled by a server.

The second sanctuary lies in caregiving and direct human relationship. Dallas Fed economists Samuel Dodini and Tucker Smith noted that while computer science majors face high exposure, nursing, education, and psychology rank among the least exposed, and Stanford researchers found that while early-career software roles contracted, home health aide employment among the youngest workers grew. As one Federal Reserve Worker Perspectives Project participant observed about the fear of automation replacing human beings, there remains the essential human touch that cannot simply be programmed away, saying there is still "the human touch, which is going to be necessary in general".

Even within technology, the systems themselves require physical and structural stewardship. Labor research notes that human personnel remain indispensable to ensure complex operations function properly and maintain basic safety and quality. Forecasting on the side of labor absorption, the U.S. Bureau of Labor Statistics projects data scientist roles will grow by 33.5 percent through 2034, with software developers expanding by roughly 268,000 positions, while information security analysts are also projected to see robust job growth.

For the professional who never leaves a desk, resilience boils down to discernment. Research highlights high-level tasks like strategic planning and synthesizing complex sensory data as areas that still require human expertise, while PwC reports that employers are placing heavier emphasis on leadership, creativity, and judgment. Crucially, as analysts citing Fleming and colleagues point out, effective adoption demands knowing when automated outputs can actually be trusted. Yet even if these roles endure, an uncomfortable question follows: what happens to the lost income and the government tax base when high-earning work is hollowed out?

Behind every white-collar paycheck sits a quiet fiscal reality. According to 2023 data cited from a Congressional report, about three-quarters of all United States federal tax revenue comes directly from labor. Individual income taxes contribute about 49 percent of federal collections, while payroll taxes account for approximately 36 percent. If widespread automation shifts national income away from human paychecks and toward corporate profits, it strikes directly at the employment-based income streams that fund Social Security, Medicaid, and nutrition assistance. The Budget Lab at Yale points out that output growth skewed toward capital will naturally reduce the share of gross income collected as tax, because investment returns are taxed at far lower effective rates than wages.

That fiscal threat is why advocates of redistribution warn that high-skilled automation could trigger deep structural harm. Proponents of that view point to Census Bureau research showing that after large language models became available, graduates in the most exposed decile of college majors saw initial full-quarter earnings drop by 13 percent, alongside a five-percentage-point decline in their likelihood of initial employment. The Budget Lab modeled national labor share falling from 53.7 percent in early 2026 to 51.3 percent by 2030 under rapid adoption, shrinking labor-tax revenues below federal baselines. Quoted in a 2026 paper, BlackRock chief executive Larry Fink sounded a similar note of caution: “History suggests that transformative technologies create enormous value—and much of that value accrues to the companies that build and deploy them, and to the investors who own them. There’s a real risk artificial intelligence could widen wealth inequality if ownership does not broaden alongside it”

Market proponents, however, argue that high-skilled automation does not simply destroy income; it can expand it. Across job postings in 27 countries, PwC found wage growth reached 24 percent at companies most able to use artificial intelligence, compared with 17 percent at the least-exposed firms. Pay in professionalized roles grew faster than in democratized categories. Researchers David Autor and Neil Thompson argue that automating high-expertise tasks lowers barriers to entry, enabling broader pools of workers to perform complex work. Another study of university syllabi found graduates with AI-exposed coursework enjoyed higher starting pay and shorter job hunts, while a 2025 paper cited by the Federal Reserve documented statistically significant wage gains in exposed occupations without significant drops in total jobs.

The ground-level picture for high earners is therefore not a uniform collapse, but a sharp generational wedge. Experienced professionals who wield senior judgment command substantial premiums, yet new entrants face immediate friction. Researchers Cody Orr, Lee Tucker, and Lawrence Warren observed that the 13 percent earnings decline for graduates in exposed majors is comparable to graduating into a major economic recession, with Dallas Fed economists Samuel Dodini and Tucker Smith finding Texas graduates from exposed fields saw first-year pay fall about 5 percent relative to less-exposed majors. Entry-level openings most exposed to automation are now seven times more likely to demand traditionally senior skills like leadership and face-to-face interactions. That divide leaves governments confronting a pivotal question: how to re-engineer public revenue and support displaced workers when the traditional wage base begins to splinter.

When the talk turns to lost income, the quickest promise on the table is often universal basic income. Supporters reportedly argue a guaranteed floor cushions workers through transitions, giving them room to retrain or launch new ventures. Back in Indore, that was the grand fix Prakash's son brought home from engineering school, arguing guaranteed checks were bound to arrive. But Prakash cut straight through the romance, asking who will pay for it and will it cover the rent. Run the actual math, and his skepticism proves well earned. Modeling published in 2026 found that funding a full-coverage universal basic income would cost 10.7 trillion dollars each year, representing about 36 percent of the nation's economic output. The authors note that funding that benchmark at current revenue levels would require an economy of over sixty trillion dollars, while even a fifteen-thousand-dollar annual payment is estimated to cost more than five trillion dollars.

That fiscal hurdle has led lawmakers to ask whether companies replacing human labor should pay for the transition directly. Proponents of redistribution call for robot taxes, which are reportedly designed to replace lost worker income tax and fund retraining or safety nets. Senator Bernie Sanders argued that "instead of providing billions in tax breaks to companies that are throwing workers out on the street and replacing them with new technologies, we should enact a robot tax on large corporations and use the revenue to improve the lives of workers who have been harmed." Representative Greg Casar proposed taxing large artificial intelligence developers on token sales or product revenue, with the rate rising automatically if national unemployment climbs. Theorists Brett Hemenway Falk and Gerry Tsoukalas have reportedly modeled corrective Pigouvian taxes on automation to deter excessive displacement. Proponents of redistribution point to South Korea, which reduced tax deductions for automation investments in 2017—a reform researchers found decreased automation spending, raised employment, and reduced wage inequality. Yet market proponents push back; policy analyst Will Rinehart argued in 2025 that existing corporate profits, capital gains, and property levies already capture the returns of automation without selectively penalizing innovation.

Instead of penalizing specific machines, another school of thought proposes rebuilding public finance so the entire population shares directly in the gains of capital. Former Labor Secretary Robert Reich reportedly suggested those windfalls could flow through wealth taxes, or through a sovereign wealth fund holding shares in AI companies and distributing yearly dividends to every citizen. In April 2026, OpenAI laid out a similar blueprint, where proponents of redistribution proposed that government shift its revenue foundation away from payrolls toward corporate income and investment returns. Under that plan, a national wealth fund capitalized with contributions from technology companies would hold equity across adopting industries, paying out returns as a universal citizen dividend. Economists Anton Korinek and Lee Lockwood note that sovereign funds can capture returns that automatically scale with artificial intelligence, while consumption taxes must reportedly carry far more fiscal weight as payroll collections erode.

Yet sovereign funds take time to build, while displaced workers face immediate bills. Research shows that nearly half of the persistent earnings drop following technological job losses reportedly happens when workers slide into lower-paying occupations because their skills do not match newly created work. Federal Reserve Governor Lisa Cook emphasized that the societal benefit of AI will ultimately depend on worker adaptability, how well workers are retrained or redeployed, and how policymakers choose to support the hardest-hit groups. Analysts point out that American public spending on training and employment services sits at less than one-fifth of the average among advanced economies relative to GDP. To bridge that gap, labor economists advocate for targeted wage insurance to cushion pay cuts for workers moving into lower-paying positions while retraining. OpenAI's blueprint similarly recommended data-driven automatic stabilizers that trigger wage insurance and cash support the moment measured automation displacement crosses defined thresholds. With those policy blueprints mapped, the final question is what this revolution truly means for the work ahead.

The answer to what happens next begins by untangling task automation versus occupational extinction. Researchers report that generative AI is not causing mass net unemployment across the aggregate economy. Instead, researchers report that firms adjust labor demand by redesigning tasks inside existing roles and reallocating hiring. Advocates of labor absorption point to findings from the International Labour Organization that large-scale job displacement remains limited, while the Yale Budget Lab reported in September 2026 that AI usage measures show no connection to changes in employment or unemployment. Researchers at Stanford likewise found no evidence of widespread economy-wide displacement in data through June 2026, and federal projections still forecast overall employment to rise by about 5.2 million jobs through 2034.

What happens to the lost earnings of high-skilled work, however, is a contested split rather than total erasure. Market proponents emphasize that professionalized roles have seen 42 percent faster salary growth as AI tools elevate complex work. Yet proponents of redistribution note that shifting output toward capital concentrates gains. Modeling from the Yale Budget Lab suggests rapid AI adoption could push labor's share of output down to 51.3 percent by 2030. In 2026, BlackRock chief executive Larry Fink warned that value accrues to the companies and investors that own transformative technology, creating a real risk of widening inequality unless ownership broadens. According to cited 2023 data, about three-quarters of federal tax revenue comes from labor, though that reliance is contested, leading proponents of redistribution urge shifting revenue emphasis away from wages toward corporate profits, investment returns, and automated labor charges.

That tension returns to the dinner table in Indore where Prakash asked his son whether safety nets would truly cover the rent, and who would pay. As former Labor Secretary Robert Reich noted, the real question is where the political will is expected to come from to do it. While code can draft documents or process figures, workers in Federal Reserve focus groups observed that technology can help, but it remains to be seen how AI is used because there is still the essential human touch that will be necessary. In the end, software can model the world, but it is human hands and human judgment that keep it running.