Automating a job's easy parts can make everything left tougher — fewer people qualify, but those who do earn more.

This explores whether automating the easy, beginner parts of a job leaves behind work that needs more expertise, and what follows from that. The short answer from the corpus is yes, and the effects are less obvious than they first seem. When automation removes inexpert tasks, the work that remains needs more expertise. Wages rise for the people who can do it, but fewer people qualify, so the workforce shrinks. When automation removes expert tasks, the reverse happens: requirements drop, wages fall, and less-skilled workers can enter. Employment moves in the opposite direction from the number of tasks removed. Counting automated tasks tells you little, because what matters is which end of the skill ladder gets cut Does automation raise or lower the skills that remaining work demands?.

Raising the bar has a second cost: it removes the steps people used to climb to reach it. Interviews with South Korean software engineers describe generative AI pulling basic tasks into workflows run by seniors with AI. Juniors lose the hands-on struggle that used to turn them into experts, and seniors and juniors don't see the problem the same way Does generative AI prevent juniors from getting entry-level work?. A related analysis names two ways this goes wrong. The 'missing junior loop' means nobody is being trained to check AI's output in the future. The 'codifier's curse' means experts who label data for AI end up encoding their own judgment into systems that replace them How does automating junior work destroy verification oversight?. Anthropic's own engineers show the same tension on a small scale. They report large productivity gains, yet most can fully hand off only 0–20% of their work, and they worry that skipping routine coding weakens the skills they need to catch Claude's mistakes Does AI assistance erode the skills needed to oversee it?.

The corpus also shows that AI doesn't always remove the easy parts. In the BCG experiment, GPT-4 helped lower-performing consultants most and narrowed the skill gap Does AI help lower-skilled workers catch up to top performers?. In that setting AI acted more like a ladder than a bar-raiser. In another set of experiments, AI took over the engaging parts of tasks and left people with dull leftovers, so after working with AI they felt more in control but more bored and less motivated Does AI collaboration drain motivation when workers return to solo tasks?. Which slice of work the AI takes, whether the beginner part, the expert part, or the interesting part, decides what's left for people.

Whether this actually reshapes jobs depends on how AI exposure is spread across a role. When it hits only a few tasks, workers shift to the tasks that remain and job losses stay modest Does concentrated AI exposure enable workers to adapt and reallocate?. Change may also come slower than expected. Microsoft's 66-firm Copilot trial found people saved time on email but did no different kinds of work, because shared work changes only once teams reorganize around the tool Does Copilot change what knowledge workers actually do?. Nielsen adds a useful reframing. Studies that take AI away to measure skill loss test a situation that rarely happens at work. The better question is how people build higher-level expertise when AI is always available Does removing AI tools actually measure real skill loss?. That turns the entry-level problem from a worry about lost skills into a design question: what should the new first rung of the ladder look like?