On Sunday, July 19, the Financial Times ran a piece by Andrew Hill making a careful argument: the big law firms, consultancies and investment banks are not simply firing junior staff as AI eats their work. They are redesigning what a first-year is supposed to be — rebuilding graduate pipelines, training programs, and the whole idea of an entry-level role around AI.
That’s the polite framing. Underneath it are numbers.
The base of the pyramid is load-bearing
For decades the professional-services model ran on attrition arithmetic. Firms hired big entry-level classes — at elite shops fewer than two in a hundred would ever make partner — because juniors did the high-volume grind: document review, contract analysis, research synthesis, financial modeling, first drafts. That grind billed hours and freed seniors for clients. The pyramid was wide at the bottom and narrow at the top, and the wide bottom was the point.
AI attacks that base from both ends. It absorbs the exact tasks that defined junior work — the tasks that made up roughly 60% to 70% of a junior’s time — and it makes seniors productive enough that four juniors compress into one senior with a good prompt. Harvard’s Seyed Hosseini and Guy Lichtinger gave this a name in a 2025 working paper: “seniority-biased technological change.” At firms that adopted generative AI, junior hiring fell hard while senior headcount kept climbing.
The scoreboard: entry-level postings in the US fell roughly 35% from January 2023 to mid-2025, per Revelio Labs. The unemployment rate for college grads aged 22–27 sits at 5.6%, nearly double the rate for all college-educated workers. The firms at the center of that decline are not treating it as a dip. They’re treating it as a spec.
McKinsey is now testing whether you can boss the machine
The vivid example is McKinsey. CEO Bob Sternfels said at CES in January the firm runs about 40,000 humans alongside 25,000 AI agents, up from a few thousand agents 18 months earlier — agents that saved 1.5 million hours of search-and-synthesis work in 2025.
So the interview changed. Since January, McKinsey has piloted a final round where business-analyst candidates use Lilli, its internal AI platform, to work a case in real time. Interviewers watch how you prompt it, how you catch its bad suggestions, and whether you can turn its output into something client-ready. The test isn’t AI mastery. It’s judgment in the presence of a confident machine. BCG is reportedly building the same thing for Summer 2026; Bain is said to be planning one.
Law didn’t cut its way out — it moved the work sideways
Legal employment actually hit a 10-year high of about 1.24 million jobs in January 2026, per BLS data. That’s not a reprieve; it’s redistribution. AI took the document-heavy junior tasks, and the growth moved into paralegals, legal-ops and “manage the AI” roles.
The cuts landed on the support layer. Baker McKenzie cut 600 to 1,000 business-services roles in February — the largest AI-attributed reduction in legal to date — in research, marketing, knowledge management, secretarial and design, not fee-earners. Clifford Chance trimmed UK business services 10%. Irwin Mitchell deleted its litigation-assistant role entirely.
The part that compounds
Here’s the LostJobs read. The scary version of this story is not that the class of 2026 has fewer seats. It’s the class of 2034.
PwC’s internal documents show a plan to cut US audit and assurance entry hiring 32% to 39% between 2025 and 2028. MIT’s Andrew McAfee put the problem plainly to Harvard Business Review: “How else are people going to learn to do the job except via on-the-job learning and training apprenticeship?” You learn hard knowledge work by doing the routine stuff next to someone good at it. Automate that away too fast and you lose the apprenticeship ladder — and the partners it produces a decade out.
Not everyone is betting the same way. IBM is tripling US entry-level hiring in 2026, on the theory that AI needs human oversight and that gutting the pipeline is a bill you pay later. KPMG’s line is “we want juniors to become managers of agents.”
So the real question a 2026 grad should ask a firm isn’t whether it uses AI — they all do. It’s whether the firm redesigned the junior job or just deleted it and kept the org chart. Those two look identical in a press release. They diverge over ten years, and by then the choice is already made.
Sources
- AI Reshapes the Entry-Level Pipeline Across Law, Finance, and Consulting (TechTimes, July 19, 2026)
- Is AI responsible for the rise in entry-level unemployment? (Revelio Labs)
- Baker McKenzie to cut business services roles as AI use grows (Legal Cheek, Feb 2026)
- Automating Gen Z entry-level jobs could backfire, MIT researcher warns (Fortune, May 1, 2026)
