Training Ladder

What is the Training Ladder?

The training ladder is the traditional career development model in professional services—law, consulting, finance, medicine—where junior employees exchange low-value work for mentorship from experts. An analyst at McKinsey produces PowerPoint slides; a junior associate at a law firm reviews contracts; a medical resident does patient intake. The work itself isn't highly skilled, but it creates an economic exchange: the junior's labor subsidizes their training.

LSE economist Luis Garicano formalized this as the "AI Becker Problem," building on Gary Becker's foundational work on human capital. In Becker's original theory, companies won't invest in general training (skills that transfer to other employers) unless workers somehow "pay" for it. The training ladder is the implicit contract that solves this: juniors accept low wages and boring tasks; in exchange, they gain tacit knowledge—expertise that can't be taught in school.

Why AI Threatens the Training Ladder

AI language models can now perform many of the tasks that juniors traditionally used to pay for their training:

• Contract review - AI can analyze legal documents as well as a junior associate
• Basic research - Deep Research tools can do what McKinsey analysts spend weeks on
• Slide creation - AI can generate presentations that require only senior editing
• Financial modeling - Routine spreadsheet work is increasingly automated

When AI does this work better and cheaper, the junior's "currency" becomes worthless. Why would a partner hire an analyst when Claude can do the basic research? This breaks the economic logic of the training ladder.

Key Characteristics

Tacit Knowledge Transfer: The training ladder isn't just about task completion—it's how professionals learn things that aren't written in any manual. How to handle a difficult client. How to navigate organizational politics. How to know which details matter. This knowledge transfers through proximity to experts, not through reading.

The Currency Problem: As Garicano puts it, juniors pay for training "not in dollars, but in menial tasks." If AI devalues menial tasks, juniors have no currency. Firms might pay juniors to learn, but this reverses the traditional economics entirely.

Market Failure Risk: If training ladders collapse, who produces the next generation of experts? Current seniors retire; AI can't (yet) provide the judgment that comes from decades of experience. The economy may face a skills gap in 15-20 years.

Evidence of Training Ladder Disruption

Two papers from August-September 2025 show systematic evidence of junior hiring decline:

1. "Seniority-Based Technological Change" (Likenberg & Hosseini): Analysis of 62 million workers shows AI-exposed occupations have stable or growing senior employment but declining junior employment—specifically through reduced hiring, not layoffs.
2. "Canaries in the Coal Mine" (Erik Brynjolfsson et al.): Workers aged 22-25 show clear employment drops in AI-exposed versus non-AI-exposed professions.

Why the Training Ladder Matters

The training ladder problem suggests AI's impact on knowledge work may be subtler and more long-term than mass layoffs:

1. Silent de-hiring: No dramatic firings, just fewer entry-level positions posted
2. Expertise erosion: Current experts remain productive, but the pipeline of future experts dries up
3. Tacit knowledge loss: Institutional memory and professional judgment stop being transmitted
4. Inequality amplification: Senior experts gain leverage; juniors lose their path to expertise

Organizations deploying AI agents need to consciously solve for training—what Garicano calls designing the "ratio" where expert productivity gains are large enough to justify continuing to hire and train juniors.

Related Reading

• Supervision Threshold - The autonomy level at which AI stops needing junior oversight
• Human-in-the-Loop - The supervision model that keeps juniors relevant
• Enterprise AI - Where training ladder effects are most visible