Smart people disagree on the AI job apocalypse, and even the prophets of white-collar doom—Dario Amodei and Sam Altman—have walked back their predictions.

But the best explanation for why AI won’t kill off jobs across the economy comes, perhaps unexpectedly, from a Dutch software company that sells its products to law firms. It also explains why the entry-level market hiring struggle is painfully real.

Wolters Kluwer is a 190-year-old Dutch information services company that sells AI-powered software to law firms. In a piece published earlier this month, the company cited two economic concepts: the “lump of labor fallacy” and the Jevons Paradox.

The “lump of labor fallacy” was coined by English economist David Frederick Schloss in 1891, as he noted that many workers and employers believed there was a fixed amount of work to be done in an economy. You can see this everywhere over the past four years, even among AI kingpins such as Amodei and Altman, as they warned that if AI eliminates a category of tasks, the workers who performed those tasks would simply be displaced with nowhere to go.

Wolters Kluwer alluded to the fallacy by noting that AI is freeing up attorneys to spend more time on strategy, counseling, and judgment-driven work but isn’t resulting in smaller legal teams.

“Legal teams are increasingly looking for junior professionals who arrive AI-trained and ready to work alongside these tools,” it said. “They need people who can validate AI output, manage workflows, and apply their expertise to the outputs rather than the inputs.”

The Jevons Paradox is an even older bit of economic lingo. Coined in 1865 by the English economist William Stanley Jevons, it has been invoked regularly by Apollo Global Management chief economist Torsten Slok to argue that AI will create more jobs, not less. Amodei even referenced it himself in May while retreating from his own AI jobpocalypse claims.

This paradox applies when a resource becomes cheaper or more efficient to use and total consumption of it tends to rise, not fall. When steam engines became more fuel-efficient in the 19th century, coal consumption didn’t drop—it multiplied, because cheaper engines proliferated everywhere.

Applied to legal work, Wolters Kluwer said AI that cuts the cost of research and document review doesn’t reduce demand for legal services, but rather expands the universe of what clients expect law firms to deliver. Efficiency creates appetite, not surplus.

“Efficiency gains driven by AI are likely to increase expectations about the work you can produce rather than reduce demand,” the firm argued, calling AI a “task machine, not a job machine.”

Wolters Kluwer added that AI “excels at completing individual workflows but lacks the judgment required to perform an end-to-end job as a person would,” citing internal research findings that AI produced professional-quality output on individual tasks roughly 50% to 60% of the time across various roles. When tasked with executing a complete project end-to-end, though, the success rate drops to around 2%.

This perfectly fits the pattern of a labor market where the entry-level workers who do one task at a time struggle to get hired, and the rest of the AI jobpocalypse just doesn’t really show up in the data.

The question the webinar doesn’t ask

The entry-level job market is the worst it has been in 37 years. Entry-level positions across professional services have dropped 29% since January 2024. Finance and information services—the industries that have historically provided an on-ramp for most college graduates—shed an average of 9,000 jobs per month since 2023, compared to adding 44,000 per month before the pandemic. A Stanford study found workers ages 22 to 25 in highly AI-exposed occupations experienced a 13% drop in employment since 2022. Before walking it back, Amodei warned that AI could eliminate roughly half of all entry-level white-collar jobs within five years.

Gen Z is not struggling because of bad attitudes or unrealistic expectations. The first rung of the career ladder is structurally disappearing. And the Wolters Kluwer framework explains why—although it declines to say so.

The document frames AI’s impact as a pyramid: AI handles tasks at the base, humans retain judgment at the top. Legal teams are growing, it notes, by hiring professionals who can validate AI outputs and focus on higher-value strategic work. It also describes a profession that has decoupled entry-level hiring from its own growth.

Firms don’t need fewer senior lawyers—they need more of them, better-leveraged, handling more sophisticated work for more demanding clients. While Wolters Kluwer sees demand expanding, it doesn’t look closely at where in the value spectrum that is the case. Compounded across an industry over a decade, it describes a profession that has stopped training its own replacements.

PwC calls this “seniorization,” based on an analysis of more than 1 billion job postings. The Big Four firm’s 2026 AI Global Jobs Barometer found that entry-level roles in highly AI-exposed occupations have become seven times as likely to require skills that have historically appeared later in a worker’s career. These are skills like strategic decision-making, stakeholder management, leadership, and judgment.

This is not a new pattern. It is the oldest pattern in economic history.

The medieval plow dramatically increased agricultural output across Europe. Peasants didn’t benefit—the surplus went to build cathedrals. The spinning jenny automated textile production and led to longer hours at lower wages for the workers it was supposed to liberate. The internet created more wealth than any technology in modern history and concentrated it among a small number of platform companies while generating, for most workers, gig roles, delivery routes, and content moderation queues.

The question has never been whether technology creates wealth. It is always about who captures this wealth, and under what political and institutional conditions it becomes broadly distributed.