Before HR Automates a Task With AI, Ask What It Used to Teach - SHRM

TEXT START: Editor's note: Posts published on From the Workplace are written by outside contributors and do not necessarily reflect the views or opinions of SHRM.

The Dissection

This is a legitimate warning disguised as an HR design exercise. The article correctly identifies that junior-level “busywork” often functions as an apprenticeship layer: drafting, researching, analyzing, and making mistakes teach judgment.

But it narrows a structural labor shock into a workflow problem. Its proposed solution is to replace deleted practice with simulations, reviews, coaching, and deliberate learning moments. That may preserve capability formation temporarily. It does not preserve mass economic participation, wages, or bargaining power.

The article accepts AI’s destruction of the apprenticeship floor, then asks managers to decorate the basement.

The Core Fallacy

The central error is treating capability development as equivalent to economic indispensability.

Learning to review an AI-generated memo may produce judgment. It does not follow that the organization will need millions of humans to exercise that judgment. Under P1, AI systems can increasingly generate, compare, critique, revise, and document outputs. Human oversight is not automatically a durable moat; it is another cognitive layer exposed to automation.

The article also treats “AI reshapes jobs” as evidence against replacement. That is a category error. Reshaping roles can mean fewer workers performing broader supervisory functions. New roles and upskilling may be transition effects, not proof that productive participation remains structurally secure.

The article identifies a real second-order consequence—weakening talent pipelines—but mistakes the consequence for an escape route. A better training system does not create durable labor demand when the production system is designed to require fewer people.

Hidden Assumptions

• Firms will sacrifice short-term automation gains to preserve long-term human development.
• Managers will have the time, skill, and incentive to provide intensive coaching after automation removes routine work.
• Simulations and peer reviews can reproduce the contextual learning gained from real responsibility, real mistakes, and real consequences.
• Human review and approval will remain economically necessary rather than becoming another AI function.
• Reskilling leads to actual positions instead of a larger pool of trained people competing for fewer seats.
• “More capable employees” will retain bargaining power even when AI capital performs most economically valuable cognition.
• Organizations can redesign development without confronting ownership of the AI systems, distribution of returns, or the collapse of the wage-consumption circuit.
• The reported adoption and upskilling figures describe the destination rather than the lag phase of the transition.

These assumptions are not demonstrated. They are managerial hopes embedded inside an apparently practical checklist.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The partial truth is real: removing entry-level work can destroy the mechanism by which workers acquire competence. The transition-management function is also real: organizations need methods to operate while human roles are compressed.

The anesthetic lies in presenting this as a solvable HR responsibility. The article gives leaders a humane vocabulary—“better practice,” “feedback loops,” “stronger people”—without confronting the harder question: stronger for what labor market, and with what claim on the productivity created by AI?

It shifts attention from displaced workers and concentrated AI ownership to the quality of the remaining training process. The machine takes the work; HR is told to preserve the lesson.

The Verdict

This is an accurate diagnosis of apprenticeship erosion and an inadequate diagnosis of systemic collapse. It sees the talent pipeline being amputated but assumes better instructional scaffolding can replace the missing limb.

Under the Discontinuity Thesis, the likely result is not stronger people at scale. It is fewer junior positions, compressed career ladders, delayed entry into meaningful work, and a smaller class of humans retained as accountable servitors around AI capital. The article is not pure copium because its operational warning is valid. But as a systemic answer, it is a polished hospice memo for the human talent pipeline.