The Consulting Pyramid Is Broken: What Replaces It

The staffing model that built modern consulting — a broad base of junior analysts supporting a thin layer of senior partners — is under structural pressure that goes beyond a cyclical downturn. Firms are cutting graduate hiring, freezing starting salaries, and laying off thousands of employees who cannot be reskilled for AI-driven delivery. The question is no longer whether the pyramid changes. It's what replaces it, and what that means for organizations buying advisory services.

Key Takeaways

• MBB starting salaries have been frozen for three consecutive years at $135,000–$140,000 (undergraduate) and $270,000–$285,000 (MBA); Big Four salaries have not increased since 2022
• PwC cut graduate hiring in 2025 and abandoned its target to add 100,000 employees globally by 2026 — a goal set before generative AI
• Accenture reduced headcount by approximately 22,000 in 2025, including 11,000 in a single quarter, as part of an $865 million restructuring
• McKinsey reduced its workforce from over 45,000 to approximately 40,000, with a further 10% reduction in non-client-facing roles expected over 18–24 months
• One analysis estimated that AI tools can now perform roughly 80% of a junior analyst's typical research and slide-generation work
• Only approximately 25% of McKinsey's global fees are linked to outcomes; the rest still come from traditional billing

The Leverage Economics That Made It Work

The consulting pyramid was an economic engine, not just an org chart. Partners sold work and maintained relationships. Mid-level consultants managed delivery. A broad base of graduates conducted research, built models, and assembled the analysis that senior staff leveraged into client presentations. The margin structure was straightforward: bill premium rates for hours mostly performed by the least experienced team members. A partner selling a project at €3,000 per day while staffing it with analysts costing the firm €400 per day was the core arithmetic.

This worked because the analytical work was genuinely time-intensive and difficult to source externally. Clients needed people to gather data, run scenarios, and synthesize findings, and the only way to do this at scale was to employ large cohorts of graduates capable of managing the volume.

Generative AI broke that dependency. Tasks that once took a junior analyst days — research synthesis, data analysis, first-draft writing, slide generation — can now be completed in hours. McKinsey's internal tool Lilli, deployed to over 7,000 consultants by 2025, reportedly saves consultants 30% of their time on research and knowledge synthesis. BCG's Deckster automates presentation formatting. The base of the pyramid is being automated faster than firms can redeploy the people who occupied it.

What the Numbers Actually Show

The hiring data tells a consistent story across the industry. PwC's trajectory is instructive. The firm set a target five years ago to increase global headcount by 100,000 by 2026. In 2025, it cut graduate hiring and acknowledged publicly it would miss that target — attributing the miss directly to generative AI's impact on productivity.

Accenture's restructuring was more aggressive. The firm reduced headcount by approximately 22,000 in 2025, including over 11,000 in a single quarter. CEO Julie Sweet stated explicitly that employees who could not be reskilled would be "exited." The firm simultaneously grew its AI and data professional headcount from 40,000 in 2023 to nearly 80,000 by late 2025. This is not a hiring freeze — it is a workforce replacement.

McKinsey cut its workforce from over 45,000 to approximately 40,000, with Bloomberg reporting an expected further 10% reduction in some areas over 18–24 months, primarily in non-client-facing roles. One consulting firm that was hiring 15 people in its incoming class in 2021 scaled that back to three or four for 2026. The pattern extends across professional services to law firms and accounting practices.

What Replaces the Pyramid

Several structural models are emerging to replace the traditional hierarchy. Each reflects a different set of assumptions about the role of AI and the economic basis of professional services delivery.

Model	Structure	Basis	Who Is Adopting
Diamond	Wider middle, thinner base	More mid-level orchestration	Most mid-to-large firms
Obelisk	Fewer layers, minimal juniors	Senior-led delivery	Restructuring-focused firms
Box	Senior matched 1:1 with junior	Experienced professionals, less leverage	Alvarez & Marsal
Inverted Pyramid	Senior teams + AI agents at the base	AI replaces junior analyst capacity	AI-native agencies

The most radical model flips the pyramid entirely: small teams of senior and mid-level consultants with minimal junior support, backed by AI systems for data processing and analysis. This is where AI-native agencies are emerging — firms built around AI-augmented delivery from day one rather than retrofitting it onto a headcount-dependent model.

A less discussed but equally significant shift is happening at the organizational level. The expert-network industry grew from virtually nothing in 2010 to an estimated $3–4 billion in 2024–2025, with projections to more than triple by the early 2030s. Complex, non-recurring work is increasingly sourced on demand rather than maintained as permanent internal capacity.

The Pricing Problem Nobody Wants to Solve

The structural issue firms are avoiding is pricing. If AI makes a junior consultant 10x more productive on research and analysis, the firm either passes those savings to clients, finds new work to bill, or pockets the margin improvement. So far, most firms are choosing the third option.

EY leaders have acknowledged the pressure to move toward "service-as-software" pricing, but few firms have made meaningful progress. Partner compensation structures built over decades around selling human hours are difficult to unwind. The Deloitte Australia incident — where the firm refunded part of a A$440,000 government contract after AI-generated errors including fabricated citations were discovered in the deliverable — illustrates the flip side: AI is being used to reduce delivery cost, but the quality controls have not kept pace.

What This Means for Buyers

Organizations purchasing advisory services should be asking different questions than they were two years ago. Not "how many people will you staff on this project" but "what is the actual delivery model and what role does AI play in it." Not "what are your day rates" but "how do you price outcomes and what happens if we don't achieve them."

The pyramid was a structure that served firms, not clients. Its replacement — whatever form it takes — should reverse that relationship. The Advisory Model Evaluation Guide covers the diagnostic questions, evaluation criteria, and red-flag signals for assessing whether a firm's staffing and pricing model is designed around client outcomes or internal margin protection.