The Case of the Vanishing Junior Associate

In early May, Axios published a piece that should have rattled every managing partner in the country. The headline was about Big Law, but the argument is not really about law. Artificial intelligence is wiping out some of the entry-level work that trains the next generation of elite lawyers, and Big Law's entire business model depends on armies of junior associates learning on the job.

The diagnosis generalizes. The legal profession is just where the data is most legible. The same pattern is showing up in software, finance, consulting, accounting, and journalism - and it has a structural shape worth taking seriously.

• The Axios framing was sharper than the usual "AI is coming for jobs" story. University of Houston Law Center professor Nik Guggenberger told Axios that junior work has always served two purposes: billing and training. "If more and more of that work that trains junior associates is being automated, then there's no real material anymore for them to train on." Stanford Law professor David Freeman Engstrom put the same point in plainer terms: firms are racing to "extract the knowledge of their lawyers" and embed it in AI workflows, client portals, and self-service tools, which could mean "getting ready for a world in which you need fewer human lawyers." (axios.com)

• The dislocation is already visible inside Big Law's own numbers. Baker McKenzie cut 600 to 1,000 business services roles across IT, knowledge management, marketing, secretarial, and design in February 2026, citing AI integration — the largest AI-attributed cut in the legal industry to date. Hiring patterns are quietly shifting underneath the surface too: large junior associate classes are under pressure, and the traditional model of onboarding 20+ first-years and training them through repetitive work is being questioned. If AI handles first-pass research and drafts, the training path that used to develop junior lawyers is narrowing. (reuters.com)

• In software, the numbers are starker. A recent Fortune essay by Yale's Jeffrey Sonnenfeld and colleagues drew on Stanford labor economist Erik Brynjolfsson's research to make the point unavoidable: employment among developers aged 22 to 25 has fallen nearly 20% from its late-2022 peak, and software development job postings on Indeed have fallen 53% from the same starting point. Headcount overall is still growing, but the entry rung is being shaved. (fortune.com)

• In finance and consulting, the pattern repeats. Anthropic CEO Dario Amodei predicted in May 2025 that AI could eliminate up to 50% of entry-level white-collar jobs within five years, particularly in finance, law, consulting, and tech, and warned that youth unemployment could rise to 10 to 20%. The 2025–2026 data strongly supports his warning: tech entry-level hiring dropped 30 to 50% in 2025, Wall Street banks are cutting approximately 200,000 roles concentrated in entry-level analyst positions, and nearly 55,000 U.S. layoffs in 2025 were directly attributed to AI. (cnbc.com)

• There is a more comforting version of this story, and it is worth taking seriously. In a January 2026 essay for Above the Law, Vanderbilt Law's T. Kyle Turner pushed back: "I do not subscribe to the belief that AI will reduce the need for new associates. Who will become senior associates and partners if we cut off the pipeline? Instead, I view AI as an opportunity to reframe what early practice looks like. Rather than spending their first years bogged down in repetitive but necessary tasks, associates will be able to focus earlier on rewarding intellectual work." (abovethelaw.com)

• But the optimistic case requires an assumption the data does not yet support. It requires that firms will actually invest in the new apprenticeship model the optimistic case demands. The recruiting data so far suggests they are not. Many firms are over-hiring juniors without rethinking the training path. If AI handles the repetitive work that used to develop junior lawyers, firms need a new model for how those associates build judgment and client skills. Hiring the same way without solving that creates a pipeline that does not lead anywhere.

• The most candid voices in legal training admit the curriculum has not caught up. One law firm's head of associate development told Thomson Reuters that her firm is working on a curriculum to teach associates how to accelerate the development of legal judgment, including the ability to supervise and efficiently validate AI-produced work. "It has not fully happened yet," she said. "But building the training infrastructure to operationalize this is a stated goal for the year ahead, including formalized curriculum around effectively and efficiently supervising AI output." (thomsonreuters.com)

• The conventional career algorithm has been stable for fifty years. As Shawn Kanungo put it bluntly in a recent essay: "Get good grades, go to a good school, get a good entry-level job, work your way up" was a stable algorithm for 50 years. In 2026, the first rung of that ladder is being chopped in half. 52% of workers aged 18 to 24 believe AI will hurt their long-term career prospects, and young adults are 129% more likely than older workers to fear AI-driven job loss — which suggests the people closest to the rung being cut have read the situation correctly. (linkedin.com)

Orthogonal Take

The conventional framing of this story is wrong in a specific way. It treats AI's impact on junior work as a question about jobs. It is actually a question about training infrastructure.

A few things follow from that reframe:

1. The first rung was always the point. Junior lawyers, junior analysts, junior associates, junior reporters, junior residents — the names differ, but the function is the same. The bottom rung is where professional judgment gets built. Take away the work that builds it, and you do not get faster paths to senior judgment. You get senior titles attached to people who have never developed it.
2. The optimistic case is plausible but contingent. Turner and the Thomson Reuters essay may be right that AI will let juniors do more interesting work, sooner. But that outcome requires firms, hospitals, banks, and editorial offices to absorb the cost of training that they have spent a century outsourcing to junior employees. The data so far suggests they are hiring fewer juniors without yet building the new training models the optimistic case requires.
3. Apprenticeship systems have collapsed before. They collapsed in the late nineteenth century not because they stopped being necessary, but because someone figured out how to extract their economic value without paying for them. The professions reinvented apprenticeship inside their internal hierarchies — the first-year associate, the surgical resident, the cub reporter, the junior analyst — and those structures have carried the load ever since. What happens next depends on whether the institutions that benefit most from mid-career judgment are willing to fund its creation.

The bet right now is that the market will sort it out. It usually does. But not before the people on the bottom rung find out whether the ladder has a second rung at all.