The MBB Triad as Limit Case of Algorithmic Expropriation: Proprietary Agentic AI, Knowledge Capture, and the Failure of the Up-or-Out Contract

Abstract The premium strategy consulting industry — represented by McKinsey, Boston Consulting Group, and Bain & Company (the “MBB triad”) — offers a uniquely transparent laboratory in which to observe the full mechanism of algorithmic expropriation of intellectual capital under the Agentic AI regime. Three structural conditions converge in these firms with a sharpness rare in other sectors. First, the private partnership form fuses ownership and labour at the partner level (4–7% of headcount across the triad) while excluding sub-partner workers from residual claims, materialising a vertical Insider–Outsider asymmetry that classical labour theory predicts only at sectoral level. Second, the up-or-out career contract systematically extracts knowledge from associates and consultants over five-to-ten-year tenures and discharges them (“counsels them out”) without intellectual capital compensation upon exit. Third, the deployment of proprietary Agentic AI systems — McKinsey's Lilli (launched August 2023, 72–75% adoption, 500,000 prompts/month), Bain's Sage (powered by OpenAI, trained on five decades of proprietary knowledge base), and BCG-X workflows (approximately 4,000 specialists actively developing AI workflows as of 2025) — crystallises the accumulated intellectual capital of generations of consultants into structural assets owned by the partnership in perpetuity. Drawing on primary press releases, annual reports (where available), and triangulated journalistic sources covering the period 2018–2026, this paper documents three findings. First, the three firms exhibit a structural asymmetry between aggregate revenue trajectory and sub-partner headcount trajectory across the period: at McKinsey, where the documentary base sustains the most robust quantitative reconstruction, revenue-per-employee rises approximately +36% from the FY2022 peak once the December 2025 wave is executed, concurrent with cumulative headcount reduction of approximately 25% from peak levels (Project Magnolia 2023, expanded December 2025: cumulative ~12,000 jobs eliminated); at Boston Consulting Group, sustained workforce expansion concurs with progressively skewed functional composition toward AI/digital profiles; at Bain, the magnitude of the asymmetry across 2018–2025 is subject to structural uncertainty over the FY2018 base figure that the paper articulates with methodological transparency, while the direction of the phenomenon is convergent with the two prior firms. Second, the three firms display three distinct displacement regimes — public forced layoffs at McKinsey, AI-tilted continuous growth at BCG (33,500 employees, twenty-two consecutive years of revenue growth), and hybrid hidden-plus-voluntary mechanisms at Bain — yet all converge in the same structural outcome: intellectual capital crystallisation in proprietary Agentic AI assets, accompanied by explicit doctrinal articulation of the displacement logic (“do I need armies of analysts making PowerPoints? No” — Kate Smaje, McKinsey, 2025) and of the redistributive intent (“preserve the compensation pool for partners” — declared objective of Project Magnolia, as reported by Bloomberg). Third, the MBB triad operates a reflexive propagation of the extractive paradigm: the firms that have built proprietary Agentic AI systems on a century of uncompensated intellectual capital transfer are simultaneously the principal vendors of that methodology to the broader corporate world. We argue that the MBB case constitutes the limit case of algorithmic expropriation: even within organisations where ownership and labour are vertically integrated at the partner level, the residual cohort of sub-partner knowledge workers experiences the full force of intellectual capital appropriation, demonstrating that internal Coasian arrangements between firm and worker cannot resolve the externality. The implications for contractual reform of knowledge work — particularly through tokenisation mechanisms preserving residual rights over transferred intellectual capital, formalised in the companion theoretical paper — are discussed in light of the MBB evidence.