> ## Content Index > Fetch the complete content index at: https://piotrmechlinski.com/llms.txt > Use this file to discover other available public pages before exploring further. # AI Didn't Take Your Job. It Took Your Successor's. - URL: https://piotrmechlinski.com/writing/ai-took-your-successor/ - Published: 2026-06-12T22:00:00.000Z - Updated: 2026-09-08T07:53:59.000Z - Description: AI can raise junior output while removing the work that built judgment. Redesign apprenticeship before efficiency weakens your succession bench. - Author: Piotr Mechlinski - Tags: Judgment The machine may not remove your seat. It may remove the work that prepared someone else to take it. That is the succession risk hidden inside the AI productivity case. Companies see entry-level analysis, drafting and coordination as cheap work waiting to be automated. The saving is immediate. The capability loss arrives later, when the organization needs people who can challenge a forecast, handle a difficult client or make a decision after the system fails. The old corporate pyramid used repetitive work as an accidental apprenticeship. AI is dismantling parts of that apprenticeship. Leaders now need to design a better one on purpose. ## The evidence is early and uneven The labour story is more complicated than the headlines. A [Stanford Digital Economy Lab working paper](https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/?ref=piotrmechlinski.com), revised in August 2026, found no evidence of widespread economy-wide job displacement in its US payroll sample. It did find that employment among workers aged 22 to 25 in highly AI-exposed occupations stood 19 percent below the path it would have followed if it had kept pace with less-exposed peers. The difference appeared mainly through lower hiring. The authors call these patterns early descriptive indicators, not causal estimates. They also note important qualifications around education, earlier trends and differences between their analysis sample and national benchmarks. Another field study, later published in *The Quarterly Journal of Economics*, examined an AI assistant used by 5,179 customer-support agents. [The largest productivity gains went to novice and lower-skilled workers](https://www.nber.org/papers/w31161?ref=piotrmechlinski.com). The authors found suggestive evidence that the tool spread some practices used by stronger performers, while remaining careful about whether workers were learning or simply following recommendations. Together, the studies support caution in both directions. AI can help less-experienced people perform better. Early-career opportunity may also be weakening in some exposed occupations. Neither finding answers the question every executive team must face: Which work will develop judgment when output becomes cheap? ## Output can hide the missing apprenticeship A polished first draft creates the appearance of senior capability. The analyst may still lack the experience behind it. They have not selected the evidence, noticed the contradiction, defended the assumption or lived with the result. If the system makes those choices repeatedly, output quality can rise while independent judgment remains untested. This is the Diamond Trap. The base of the organization narrows. AI-augmented juniors produce work that once required more experience. The middle looks efficient and capable. Yet the pipeline may contain fewer people who have made small decisions, repaired mistakes and learned when the model is wrong. Much routine work is a skill desert and should disappear. The danger begins when a company automates a task without checking whether it carried a judgment rep the company still needs. Map that learning function before removing the task from the role. ## Engineer constructive friction Preserve the learning function and redesign how it happens. Busywork can disappear. 1. **Require a first view.** For selected decisions, ask the junior employee to frame the problem, name the assumptions and form an initial judgment before opening the AI output. 2. **Turn production into audit.** Give the model's answer to the employee and require them to verify sources, locate weak logic, test a countercase and defend what they would reject. 3. **Give people bounded decisions.** Assign real choices with limited exposure, a named mentor and an after-action review. Judgment grows when a person owns a consequence small enough to survive. 4. **Rehearse system failure.** Run cases with missing data, conflicting recommendations or no model access. The exercise must show whether the future leader can still orient, choose and explain, without turning the session into theatrical pressure. The design should vary by role. A finance analyst needs different reps from a field engineer or customer-service lead. What matters is that each path connects current AI-enabled work to the decisions the next role requires. ## Rewrite entry level IBM offers a useful correction to the old article's claim. The company did not report that it had already tripled entry-level hiring. In March 2026, [IBM said it planned to triple US entry-level hiring during the year](https://www.ibm.com/think/news/entry-level-roles-get-reset-ai?ref=piotrmechlinski.com) while shifting those roles toward analysis, problem-solving, AI review and earlier client exposure. That is a plan, not proof of a successful model. The direction is still important. Keep the entry point. Change the work. Use AI to increase the difficulty a new employee can handle, then add review and responsibility that reveal whether capability is growing beneath the output. Track the bench beside the saving. Can the person explain why the recommendation is sound? Can they work when a source is missing? Have they made a bounded decision and reviewed its consequences? Are senior people transferring judgment, or merely accepting faster deliverables? The executive question is who will be ready to lead when you leave, and which work will prepare them. Your move.