Aug 28, 2026 · 55 min
AI agents put entry-level work under pressure
AI agents got so good that this top Meta exec quit her job
The episode examines whether workplace disruption is already reaching recent graduates before labor statistics can capture it, and what institutions should do next.
- 1AI agents can compress product-development work once divided among researchers, designers, engineers, and product managers.
- 2Routine, preparatory tasks assigned to recent graduates may be especially exposed as companies deploy increasingly capable agents.
- 3Clara Shih argues that schools, employers, government, and workers must share responsibility for adapting to the transition.
Don't miss
Clara Shih explains how AI agents compressed a product-development process involving researchers, designers, engineers, and product managers into work handled by one or two people.
The brief
The episode opens with a troubling labor loop: workers record manual tasks to train robots that may eventually automate the same jobs, often without clear terms for their data.
Clara Shih says AI agents changed her view of automation after she saw them compress product development from a team of specialists into work handled by one or two people at Meta.
Her central warning concerns entry-level roles, where routine research, preparation, and artifact-making may disappear before official employment data registers a broad crisis.
Shih’s New Work Foundation offers free resources for graduates and maps how AI is changing fields including marketing, accounting, and software engineering.
The proposed response is neither blanket optimism nor fatalism: teach practical agent skills while requiring coordinated action from employers, schools, government, and workers.