Between 2023 and 2025, U.S. professional services firms have reduced entry-level hiring and slowed senior promotions while adopting AI tools that automate traditional junior tasks. For example, McKinsey’s partner promotions fell from 400+ in 2021 to ~200 in 2024, and Morgan Stanley’s managing director class declined over 20% from 2022 to 2024. The implication: AI-driven “synthetic leverage” may preserve margins while weakening long-term leadership pipelines.
1. Is the Apprenticeship Layer in U.S. Professional Services Actually Contracting?
Yes. Between 2023 and 2025, multiple U.S. professional services sectors show simultaneous reductions in entry-level intake and senior promotions, coinciding with AI-driven productivity adoption.
For decades, U.S. professional services firms followed a structural model: hire broadly at the entry level, promote selectively over time, and rely on repetitive, supervised work to build future leaders. The current data suggests that this apprenticeship layer is compressing at both ends.
In consulting, McKinsey reduced partner promotions from more than 400 in 2021 to roughly 200 in 2024, despite overall headcount growth during the pandemic expansion. Industry analysts report that entry-level hiring across McKinsey, Bain, and BCG has slowed materially from 2021–2022 peaks, reducing the size of incoming junior cohorts.
In U.S. investment banking, promotion tightening is visible at senior levels. Morgan Stanley’s managing director class declined more than 20 percent between 2022 and 2024. Goldman Sachs reportedly reduced vice president promotions by roughly one-fifth in 2025, while extending analyst-to-associate promotion timelines. A senior executive at Deutsche Bank summarized a common automation rationale: “The easy idea is you just replace juniors with an AI tool.”
In U.S. accounting, Big Four firms reduced graduate hiring and slowed equity partner admissions while maintaining or increasing per-partner payouts. Firms introduced non-equity partner and managing director tiers to retain senior professionals without expanding the equity pool. This structural adjustment results in fewer entrants at the base and fewer admissions at the top.
In U.S. law firms, average summer associate class sizes have fallen to some of the lowest levels in decades. At the same time, lateral partner and associate hiring has increased, indicating a shift toward acquiring mid-level experience externally rather than developing it internally.
Executives have explicitly linked these decisions to AI productivity. At JPMorgan Chase, leadership has described a “strong bias” against reflexively hiring additional staff due to AI-enabled productivity gains. Internal tools can now generate pitch materials in seconds that historically required hours of junior analyst work, prompting internal discussions about how apprenticeship models in banking may change.
Across sectors, three concurrent trends are observable between 2023 and 2025:
Reduced entry-level intake.
Slower promotion at senior gates.
Increased reliance on lateral mid-career hires.
Historically, apprenticeship was embedded in operating workflow rather than designed by learning functions. If AI removes the workflow layer that produced experiential reps, responsibility for reconstructing that development architecture shifts, implicitly or explicitly, to L&D and talent leaders.
This pattern differs from prior cyclical downturns, where firms temporarily reduced hiring but preserved the long-term pyramid structure. The current compression coincides with automation of the repetitive tasks that traditionally trained junior professionals.
The forward-looking risk is probabilistic rather than certain: if cohorts entering between 2023 and 2025 are smaller and complete fewer experiential repetitions, leadership bench depth in five to ten years may be thinner than under prior models.
2. Why Does AI-Driven Compression Threaten the Economics of the Traditional Leverage Model?
AI threatens not only hiring levels but the economic and developmental logic of the traditional leverage model.
U.S. professional services firms historically operated on a leverage-based profitability model articulated by David Maister. In this model, income per partner is a function of leverage (staff-to-partner ratio), utilization, billing rates, realization, and margin. Leverage, defined as the ratio of junior staff to equity partners, has been the primary driver of profit expansion.
The economics are well established:
In law, the “rule of three” holds that a non-partner timekeeper should generate approximately three times their compensation cost: one-third salary, one-third overhead, one-third profit.
In consulting and accounting, staff-to-partner ratios of 10:1 or higher are common.
In investment banking, analysts and associates perform modeling, diligence, and presentation work at compensation levels far below managing directors, who capture the majority of revenue.
This structure generated both margin and managerial development. Junior professionals accumulated pattern recognition and judgment through high-volume drafting, modeling, reconciliation, and supervised correction. The repetitive nature of this work produced cognitive reps that underpinned later promotion.
AI tools now automate portions of this base-layer activity.
At JPMorgan Chase, internal large language model tools can generate multi-page pitch materials in seconds, work that historically consumed hours of junior analyst time. In consulting, observers estimate that internal AI systems can perform a substantial share of traditional junior tasks, including research synthesis and first-draft slide production. In accounting, leaders question how foundational skills will develop when AI agents handle initial execution.
At KPMG, an AI workforce leader stated: “There is a question around how to develop those core skills when you bring an agent into the mix,” later acknowledging, “I probably don’t 100% know the answer to that question.”
The Thomson Reuters Institute has framed this as a structural pipeline issue: when routine research and drafting are automated, “how are firms supposed to bring up senior associates if the work that traditionally transformed younger associates into more senior lawyers has been automated away?”
Some industry commentary describes AI as “synthetic leverage”: technology substituting for the base of the pyramid without adding headcount. One cited estimate suggests that if a 10-week engagement can be executed in six weeks using AI, labor cost may fall 30–40 percent while pricing remains stable.
This draft’s analysis distinguishes two functions of junior labor:
Economic surplus generation.
Judgment formation through repetition.
AI appears to preserve the first function while weakening the second. The leverage ratio may hold economically, but experiential accumulation declines.
Under the traditional model, firms did not need a formal system to manufacture judgment; the pyramid produced it automatically. If synthetic leverage replaces labor-based leverage, L&D leaders inherit a design challenge that operations never formally budgeted or owned.
This asymmetry, margin preserved, reps reduced, creates a structural tension that differs from prior cyclical slowdowns.
3. Are Firms Successfully Rebuilding Apprenticeship Models for the AI Era?
Firms have begun redesigning development pathways, but no publicly disclosed longitudinal evidence yet demonstrates that these replacements replicate prior apprenticeship outcomes.
The absence of measurable promotion-readiness or time-to-competency benchmarks means that most firms are redesigning apprenticeship models without a feedback loop. For capability leaders, this converts a structural economic shift into a governance risk: development effectiveness cannot be assumed simply because productivity improves.
In accounting, the CEO of the American Accounting Association described the issue as “the big question right now,” asking how professionals will understand underlying work if AI performs foundational tasks. At KPMG, an AI workforce leader stated: “I probably don’t 100% know the answer” to how core skills will develop under AI mediation.
In banking, senior recruiters have warned that skipping document review and modeling reps could be “detrimental to young bankers.” Executives at JPMorgan Chase have discussed how AI productivity gains may alter apprenticeship structures.
The Thomson Reuters Institute compared this transition to the elimination of draftsmen in engineering following CAD adoption, a shift that required redesigned mentorship and education models.
Emerging responses fall into three categories:
1. Auditor ModelJunior professionals review and interrogate AI-generated output rather than produce first drafts. Firms describe training early-career professionals to identify hallucinations, test assumptions, and validate AI conclusions.
2. Structured AI Academies and Immersion ProgramsLatham & Watkins implemented a mandatory AI academy for first-year associates. PwC introduced multi-day AI immersion programs combining technical and human skills. EY accelerated client exposure and cross-team rotations.
3. Accelerated Client ExposureSome firms argue that AI frees senior capacity, allowing earlier participation in client-facing or strategic work.
These initiatives demonstrate awareness. However, based on publicly available information, no major U.S. firm has published:
Longitudinal time-to-readiness comparisons between AI-era and pre-AI cohorts.
Error-rate or quality metrics for AI-mediated development models.
Promotion velocity data isolating AI-era cohorts.
Quantified development reinvestment relative to AI productivity gains.
Externally, technology firms have recruited former investment bankers and consultants to train AI models on financial modeling, slide creation, and analysis, tasks that historically defined junior apprenticeship work. This indicates that the underlying task layer is being absorbed into training data rather than eliminated.
The current assessment, based on available evidence, is conditional: firms are capturing AI-driven margin improvements immediately, while developmental consequences are deferred. If smaller cohorts between 2023 and 2025 experience fewer experiential repetitions, promotion pipeline fragility may surface within five to ten years.
The redesign of junior development has begun. Whether those redesigns reproduce the judgment density of the traditional apprenticeship model remains unproven. If AI productivity gains flow directly to margin while development redesign remains underfunded or unmeasured, L&D functions may be tasked with correcting leadership pipeline gaps years after the economic benefits have already been booked.
Learning and Development Executive Intelligence is for CHROs, CLOs, and senior L&D buyers investing in internal talent development, training, and reskilling.
This is one of our six education and learning-related publications spanning K-12, Higher Education, and Workforce. Our education newsletters reach tens of thousands of senior decision-makers across the U.S. and key international markets.
Ping us at [email protected] if you’d like to learn more, explore Enterprise Subscriptions, or would like to partner in other ways.
The Intelligence Council is a next-gen B2B media and business intelligence platform built for people who make strategy, allocate capital, and carry operating risk.