The Financial Times published a striking analysis on Sunday: professional services firms across law, consulting, and investment banking are not simply shedding junior employees as artificial intelligence absorbs their work. They are redesigning what a first-year analyst or associate is supposed to be — rebuilding graduate hiring pipelines, early-career training programs, and workplace culture around AI-augmented workflows. The piece, by FT senior business writer Andrew Hill, appeared on Techmeme on July 19, 2026, alongside companion reporting in Fortune and Exponential View.

What the FT reported is not new in outline, but the scale of the structural redesign it documents is. Entry-level job postings in the United States fell approximately 35% from January 2023 to mid-2025, according to labor research firm Revelio Labs, with AI playing a substantial role. The firms at the center of that statistic — elite law firms, major consultancies, Wall Street banks — are not treating that decline as a correction. They are treating it as a design specification for the workplace they intend to build.

The Pyramid Breaks: What the Old Model Was and Why It No Longer Works

For decades, the professional services talent model rested on attrition arithmetic. Firms hired large entry-level cohorts — at elite consultancies and law firms, fewer than two in a hundred would eventually make partner — because junior staff performed high-volume, repeatable analytical and research tasks that generated billable hours and freed senior professionals for client relationships and strategy. The pyramid was wide at the base and narrow at the top, and that shape was load-bearing.

The economic rationale for that base is now eroding from two directions at once. First, AI tools — built on large language models and, increasingly, on autonomous AI agents — absorb the specific tasks that defined junior professional work: document review, contract analysis, research synthesis, financial modeling, and first-draft generation. These are not peripheral tasks; they constitute roughly 60% to 70% of a junior professional's time. Second, those same tools make senior professionals more productive at the tasks that once required a junior team, compressing what previously required a team of four into work a senior professional can direct with AI assistance.

The result is what Harvard economists Seyed Hosseini and Guy Lichtinger called "seniority-biased technological change" in a 2025 working paper: at companies that have adopted generative AI, entry-level hiring has fallen sharply while senior employment at the same firms continued to grow.

McKinsey Tests AI Fluency in Final Rounds. Rivals Are Watching.

McKinsey & Company offers the most vivid illustration of where the redesign is headed.

CEO Bob Sternfels disclosed at CES 2026 in January that the firm's workforce now includes roughly 40,000 human employees alongside 25,000 AI agents — up from only a few thousand agents approximately 18 months earlier. Those agents collectively saved McKinsey 1.5 million hours of search and synthesis work in 2025 alone. Sternfels described the shift with his "25-squared" framing: client-facing consulting roles are up roughly 25%, while non-client-facing roles are down by a similar margin, even as output from that shrinking non-client side continues to rise.

That math changes what McKinsey is looking for in a new hire. Since January 2026, the firm has been piloting a new final-round interview format in which business analyst candidates are asked to use Lilli — McKinsey's proprietary internal AI platform — to analyze a case study and refine their conclusions. Interviewers evaluate how candidates prompt the system, assess its outputs, and apply judgment to produce a client-ready synthesis. The test is not about AI mastery; it is about judgment in the presence of AI. McKinsey is testing whether candidates can direct the machine, challenge its weak suggestions, and take their own position. That is now the baseline the firm considers necessary for day-one work.

BCG is reportedly developing a comparable AI-enabled interview component for Summer 2026. The Bain equivalent is understood to be in planning, according to recruiting specialists covering the 2026 consulting cycle.

Law Is Not Cutting Its Way Out: It Is Restructuring Around AI

In legal services, the redesign has taken a different shape.

Legal employment hit a 10-year high of approximately 1.24 million jobs in January 2026, according to Bureau of Labor Statistics data — an apparent paradox in a field where AI handles first-pass document review and contract analysis. The explanation is structural redistribution rather than net loss. AI is not eliminating legal work; it is reassigning it. The work that junior associates used to perform — the document-intensive, research-heavy tasks — is being absorbed by AI tools, while the work that paralegals and legal operations specialists perform is expanding.

Law firms are not merely cutting junior associate classes. Baker McKenzie cut between 600 and 1,000 business services roles in February 2026 — the largest single AI-attributed workforce reduction in the legal industry to date — with cuts concentrated in research, marketing, knowledge management, secretarial, and design functions rather than fee-earning lawyer positions. Clifford Chance reduced UK business services staff by 10% in late 2025. Irwin Mitchell eliminated all its litigation assistant positions around the same period.

What law firms are adding is equally telling. Legal operations — a role that barely existed five years ago — has become one of the fastest-growing positions in the profession, as firms need specialists who can manage how AI tools integrate with attorney workflows, evaluate AI output quality, and maintain audit trails for AI-assisted work product.

Why the Class of 2026 Is Rethinking Their Career Plans

The effect on graduate sentiment is measurable.

Bloomberg Businessweek published an investigation in April 2026 tracking the career decisions of graduating Princeton students who had built their trajectories around elite consulting. Princeton seniors Liv Bobby and Ezekiel Akinsanya had met on their first day of college, discovered they both aspired to top management consulting firms, and co-founded an undergraduate consulting club. By graduation, Bobby was heading to Wall Street and Akinsanya was planning to return to England for government or military work. "I want to be integral, and right now it's not very clear that you will be integral as an analyst anymore," Akinsanya said.

That disorientation is not anecdotal. According to Monster's 2026 Graduate AI Readiness Report, nearly 90% of graduates in the Class of 2026 expressed concern that AI or automation could replace entry-level roles, up from 64% in 2025. The unemployment rate for college graduates ages 22 to 27 stands at 5.6%, nearly double the rate for all college-educated workers, per the Federal Reserve Bank of New York.

PwC's internal trajectory is one reason those numbers exist. Internal documents obtained by Business Insider showed the firm planned to cut entry-level hiring in its US audit and assurance division by 32% to 39% between 2025 and 2028. At the same time, PwC AI Assurance Leader Jenn Kosar told Business Insider in 2025 that the firm's new hires would be doing manager-level work within three years: "People are going to walk in the door, almost instantaneously becoming reviewers and supervisors," she said. The training that once focused on teaching first-years how to execute audit tasks now focuses on professional skepticism, critical thinking, and how to evaluate AI output.

How Does AI Actually Absorb This Work? The Technical Mechanism

Understanding why this redesign is happening — rather than simply that it is happening — requires knowing what AI agents actually do differently from the AI tools of three years ago.

Earlier AI tools were assistants: a junior professional still had to frame the question, run the search, read the results, and draft the synthesis. Modern AI agents are more autonomous. McKinsey's Lilli, for example, has evolved from a search-and-synthesis tool into an agentic orchestrator capable of decomposing a multi-part research task into sub-tasks, executing them across the firm's knowledge base, and returning a structured deliverable. That is the specific capability that produced 1.5 million saved hours of synthesis work in 2025 — work that previously required human time at every step.

The legal equivalent is document review and contract analysis: large language models like those underlying Clio, Harvey, and Lexis+ AI can read, categorize, flag, and summarize documents at speeds and per-document costs that make first-year associate review economically redundant as a standalone task. JPMorgan's COiN system has demonstrated this at scale in loan agreement analysis. The result is not that junior professionals have nothing to do; it is that the specific tasks that used to justify hiring large junior classes are no longer priced the same way.

The tradeoff built into this design is not trivial. AI systems make systematic errors — hallucinated case citations, missed nuance in contractual language, failure to flag context-dependent risk — that require human oversight. The firms that are growing are those that have figured out how to use AI to produce a first draft and humans to validate and improve it. The firms that are struggling are those that deployed AI and reduced human oversight simultaneously, discovering that AI confidence and AI accuracy are not the same thing.

The Experience Gap: Who Trains the Next Partner?

The structural concern that the FT article surfaces, and that MIT economist Andrew McAfee has articulated most directly, is not about this year's job market — it is about the partner class of 2034.

McAfee warned in May 2026 that eliminating entry-level roles without replacing the apprenticeship pathway they provided creates a leadership void that compounds over time. "How else are people going to learn to do the job except via on-the-job learning and training apprenticeship?" he told Harvard Business Review. "That's how you learn to do difficult knowledge work is by helping somebody who's good at that with the routine stuff. And when we put too much automation in that too quickly, we lose that apprenticeship ladder."

The historical model delivered more than execution capacity. Junior professionals watched how senior partners read a client, framed a problem, and decided when to push back — learning by proximity, not by instruction. AI agents do not provide that kind of mentorship. They produce outputs; they do not model judgment.

Firms that have recognized this are building new apprenticeship pathways. PwC's accelerated training program explicitly front-loads critical thinking and professional skepticism because it knows first-years will not develop those skills through repetitive document tasks. McKinsey's liberal arts hiring pivot signals a similar calculation: Sternfels has indicated the firm now prioritizes candidates with creativity and judgment over those who demonstrated mastery of the problem-solving tasks AI can replicate.

The firms that have not recognized the risk are hiring less and training less, assuming AI will fill the gap — a bet that may prove correct for the next two years and damaging for the decade after.

What Firms Hiring Differently in 2026 Look Like

The contrast between redesigners and laggards is already visible.

IBM, which announced in February 2026 that it would triple entry-level hiring in the United States, is the most prominent counterexample to the cutting trend. IBM's logic, articulated by Chief Human Resources Officer Nickle LaMoreaux, is that AI tools require human oversight and that eliminating the entry-level pipeline creates a long-term management problem: firms may save now and pay later when they run out of people who understand the domain work well enough to supervise AI doing it.

KPMG has articulated a parallel vision internally. "We want juniors to become managers of agents," said Niale Cleobury, KPMG's global AI workforce lead, describing the firm's approach as accelerating juniors into an oversight role rather than cutting them. BCG has told candidates that roles tied to AI transformation, model operations, data strategy, and human-in-the-loop change management are now the fastest path into top-tier consulting.

The talent proposition at these firms is concrete: an entry-level professional who can prompt AI effectively, evaluate its output critically, and communicate findings to clients is more valuable — and commands a faster promotion track — than one who spent their first three years doing work AI now does for free.

Is the Career Ladder Permanently Shorter, or Just Differently Built?

The question the FT piece ultimately surfaces — and that the data does not yet definitively answer — is whether the talent transformation professional services firms are undergoing will produce a better junior experience or simply a smaller one.

The optimistic case: entry-level professionals who join a firm that has genuinely redesigned the role will do more interesting work earlier, develop client skills faster, and reach senior positions in less time. PwC's Kosar puts it directly: in three years, a first-year hire will "feel more like the managers of my day."

The structural concern runs alongside that optimism. The PwC audit and assurance hiring cut of 32% to 39% through 2028 means that the entering class will be substantially smaller. Fewer people entering means fewer people completing the development pathway, regardless of how accelerated that pathway becomes. If the firms that are cutting deepest are also the ones failing to redesign training, the talent pipeline does not just compress — it breaks.

The class of 2026 and beyond faces a professional services landscape in which firms are still hiring graduates, but the number of available seats has contracted, the required skills on day one are different, and the value of choosing the right firm has increased. Firms offering AI-augmented junior roles with genuine training and mentorship will attract and retain stronger candidates. Those still running the 2019 model — large classes doing work that AI can now do — are accumulating a competitive disadvantage that will compound over the next decade.

Frequently Asked Questions

How is AI changing what entry-level consulting and law jobs actually involve?

Firms are shifting junior professional roles from task-execution to AI oversight. Where a first-year associate once spent the majority of their time on document review, research synthesis, financial modeling, or first-draft creation, they are increasingly expected to direct AI tools through those tasks, evaluate and correct AI output, and communicate findings to clients. McKinsey now tests this explicitly in final-round interviews using its Lilli AI platform. PwC's AI Assurance Leader Jenn Kosar has said the firm expects new hires to be performing manager-level oversight within three years. The adjustment is real, but the transition is not yet complete at most firms — many entry-level professionals are being asked to supervise AI output before they have fully learned the underlying craft, which creates a training gap that leading employers are just beginning to address.

Is McKinsey actually using AI in its job interviews?

Yes. Since January 2026, McKinsey has been piloting a final-round interview format in which business analyst candidates use Lilli, the firm's proprietary internal AI platform, to work through a consulting-style business problem in real time. Interviewers assess how candidates formulate prompts, challenge weak AI outputs, and synthesize findings into a client-ready recommendation. The pilot began in select US offices and is expected to expand more broadly. BCG is understood to be developing a comparable component for its Summer 2026 recruiting cycle. The change reflects how the firms' actual work has changed — McKinsey now operates with approximately 25,000 AI agents alongside 40,000 human employees, and consultants routinely collaborate with AI tools on every client engagement.

What is the experience gap problem, and why does it matter for the long term?

The experience gap problem describes what happens when AI absorbs the routine work that used to teach junior professionals how to do their jobs. Historically, a first-year lawyer learned to think like a senior attorney by spending thousands of hours on document review and research — building intuition, spotting patterns, and watching how senior partners used those raw materials in client situations. AI now does much of that routine work, which is good for efficiency but removes the proximity-learning that produced the next generation of senior professionals. MIT economist Andrew McAfee has framed this as losing the "apprenticeship ladder" — the pathway through which new hires gradually build the judgment that senior work requires. Firms that cut junior cohorts without rebuilding this pathway may save money in the next two years while depleting the partner pipeline they will need in the next decade. The partnership class of 2034 is being shaped right now by decisions firms are making about 2026 hiring.

What should a graduate entering professional services in 2026 specifically look for?

The specific signals that distinguish firms genuinely redesigning for the AI era from firms simply cutting are: explicit AI literacy training in onboarding (not just access to tools); a clear articulation of how entry-level professionals supervise and validate AI output; accelerated promotion tracks based on demonstrated judgment rather than years served; and evidence that senior partners have redesigned what they expect from junior colleagues rather than simply assigning the same work to smaller teams. IBM, which is tripling entry-level hiring in 2026 while redesigning roles around human oversight of AI, represents one approach. McKinsey's Lilli pilot and PwC's manager-track acceleration represent another. Firms where "AI integration" means fewer junior hires with no described redesign of the remaining roles are the ones where the gap between promised career development and actual day-one experience is likely to be widest.