Every STEM occupation placed on three axes at once: what AI
deployment buys, what it costs to take the human out of the loop, and how hard
the human capability would be to rebuild afterwards. The first two say whether a
handoff is attractive. The third says whether it is reversible — and that is
the question a planner actually has to answer.
x · AI efficiencyImportance-weighted share of the
occupation's task mass AI can carry. Moves with the scenario.
y · Risk of removing humansError cost, accountability
and judgment under uncertainty. Deliberately excludes interpersonal demand
and physical embodiment.
z · Reconstitution difficultyTraining depth (Job
Zone), workforce scarcity, and isolation from occupations that could
cross-train in.
LowModerateHighCritical· mark size = workers
The eight cells
Cell
n
Workers
Highest combined exposure
Name
Eff
Risk
Z
Score
All values as a table
Name
Field
Efficiency
Removal risk
Reconstitution
Cell
Severity
Workers
A 3D scatter is a poor instrument for reading exact values:
near marks hide far ones and depth is ambiguous. That is why the three face
views exist — they drop perspective and give an axis-aligned 2D scatter, which
is where values should be read. The 3D view is for the shape of the cloud.
Colour encodes severity on a validated single-hue ordinal ramp, not field;
position already carries all three axes, and seven fields cannot be given seven
distinguishable hues. Employment is collapsed to 6-digit SOC before any sum.