The Risk-of-Replacement Grid

Definition

Job automation risk does not track the traditional low-skill/high-skill axis. Lee proposes two X-Y grids — one for cognitive labor, one for physical labor — built from two dimensions each: for cognitive work, "optimization-based" versus "creativity- or strategy-based"; for both, "asocial" versus "highly social." The resulting four quadrants are the Danger Zone (high risk: dishwasher, entry-level translator), the Safe Zone (low risk: psychiatrist, home-care nurse), Human Veneer (optimizable work with a necessary social front, like bartender or schoolteacher), and Slow Creep (dexterity- or creativity-bound work like plumber or entry-level graphic designer).

In the Book

The chapter builds the grid to answer "what jobs are really at risk?" after dismissing AGI/superintelligence as the near-term threat. Lee stresses that occupations are bundles of tasks, not single points on the grid — a job lands in a quadrant by its core tasks, but reorganizing which tasks a human keeps (the "Human Veneer" mechanism) can shift how much of a job survives even when the underlying computation is already automatable. He later applies the grid directly: paralegals shedding research work to focus on client contact, Amazon Go eliminating cashiers while upgrading concierge roles — both moves from Danger Zone tasks toward Human Veneer roles within the same profession.

Why It Matters

The grid gives a portable way to ask "how exposed is this activity to automation?" that doesn't require knowing the occupation's title, prestige, or education requirement — only what its core tasks actually consist of on two independent axes. That reframing generalizes past AI and jobs: any time a credential-based or prestige-based ranking obscures what's actually being measured, decomposing the activity into its task-level dimensions can reveal a different risk profile than the surface category suggests.