The Missing First Rung

A technology company dismisses junior programmers, automates routine maintenance, then advertises for senior engineers with ten years of experience. Its spreadsheet detects no contradiction: it has preserved the roof and removed the first rung. Automation can eliminate the routine work that produces future experts while preserving the expert roles that depend on it.

Expert judgment is largely compressed exposure. A senior worker recognizes a dangerous anomaly because thousands of ordinary cases have established what ordinary looks like. Routine work therefore has two outputs: the immediate product and a worker whose judgment has been altered by producing it. When a machine captures the first output, managers often assume the second will continue arriving from elsewhere.

In software, bug triage, test writing, maintenance, and minor features let beginners enter real systems under limited responsibility. The work supplied repeated cases, rapid feedback, and contact with the consequences of imperfect decisions. If machines take most of it, the human residue is architecture, security, and unfamiliar failure: work selected for ambiguity and cost. These are appropriate uses of expertise but poor raw material for creating it.

Medicine exposes an additional problem: automated routing can give trainees a statistically distorted education. A novice shown mainly exceptional cases may memorize rare patterns while misjudging thresholds and prevalence, because exceptions teach little without sustained contact with the baseline.

The shortage arrives late. Today’s senior workers are inherited capital, formed under an earlier production system. An organization can reduce junior payroll, raise output per employee, and use existing experts to handle exceptions for years; every visible metric improves while the rate at which expertise is replenished falls. By the time retirement, attrition, or expansion exposes the deficit, its cause lies far outside the reporting period. The institution then calls the result a talent shortage, as though the labor market had misplaced a product it stopped manufacturing.

Simulation and deliberate training can replace some incidental apprenticeship, perhaps efficiently. But exposure is not the only missing ingredient. Real responsibility calibrates judgment: it teaches which uncertainty warrants escalation, which error is reversible, and how incentives distort what people report. Simulation can reproduce cases and feedback; it reproduces consequence only when decisions eventually matter beyond the exercise. A replacement pipeline therefore needs supervised responsibility, not merely realistic prompts, and an explicit budget for training once training no longer masquerades as production.

The error is not automating junior work. It is accounting for that work only as output. An institution can improve its capacity to perform a task while quietly destroying its capacity to reproduce the people who can perform what automation leaves behind.