Future-Oriented Job Analysis · Stage 1 of 3

Audit the AI's verdict on a job

FOJA lists the tasks O*NET publishes for an occupation and asks an AI model to judge each one: how far could AI do this task by about 2030, and what would still need a person? Students then audit those judgements. Choose an occupation, then change any mode or score to see what your judgement does to the totals.

Real FOJA output. Six occupations, 132 tasks, scored by Claude in 2026.Developed by Song Zhaoli, NUS Business School.

All six at a glance

Model scores before any edits, weighted by task importance. The occupation you are viewing is in bold.

Exposure (replace plus augment)Replacement

What each number means

Automatability, 0 to 1: Can AI perform the core function of the task? 0 is not at all, 0.5 is about half of it or only under supervision, 1 is fully and reliably without a person.
Replacement: The importance-weighted automatability of the tasks marked replace.
Exposure: Replacement plus the tasks marked augment. The gap between the two is usually the interesting part: most professional work is far more exposed than it is replaced.

The three modes and the human bottlenecks

Replace, augment, unaffected: Replace: AI does the task instead of a person. Augment: AI helps, but a person still does or signs off the work. Unaffected: AI has little bearing on it.
Bottlenecks: The human capability that keeps a task out of AI's reach: perception and manipulation, creative intelligence, social intelligence, judgment under ambiguity, accountability and liability, or physical presence.
After Stage 1: Stage 2 holds your audited scores fixed and re-weights them under four futures: AI capability, adoption speed, regulation and cost of compute. Stage 3 turns the future you choose into three job descriptions about ten years out. This page shows Stage 1 only.