Transparent model
How AI automation risk is computed
There is no official federal “AI automation risk” field. NEF Pathways adds one, derived only from published O*NET 30.3 work-context (CX, 1–5) and ability importance (IM, 1–5) ratings. Rank order is the product — scores are percentile-blended onto an 11–89 scale so the catalog is readable, not a forecast of job loss.
What raises risk
- Repeating the same tasks, repetitive motions, exactness, sitting, current automation, equipment-paced work
- Low originality and fluency of ideas
- Number facility and information ordering (structured cognition)
What lowers risk (human characteristics)
- Conflict and messy interpersonal situations — not merely “talks to customers”
- Disease exposure and care work, outdoor weather, hands-on tools
- Leading others, decision freedom, setting one’s own priorities
Scripted public-facing work (cashiers, tellers) is not treated like counseling. Customer contact only shields the score when conflict / care / leadership is also high.
The three fields
- AI Automation Risk %
- Rank of a weighted raw score across all 1,016 O*NET occupations. Higher means more routine, desk-bound, low-originality work relative to the rest of the file.
- Human Characteristics %
- Independent score — not forced to 100 − AI risk — from conflict, care, originality, leadership, outdoor/hands-on work, and autonomy.
- Task Orientation %
- How discrete, exacting, and repetitive the work is versus holistic or self-directed.
What this is not
Not Frey–Osborne probabilities, not a BLS forecast, not a claim that any one job will disappear. It is a transparent ranking so students, coaches, and career explorers can compare pathways in the same units. A production model should be validated against task-level exposure studies and refreshed when O*NET updates.