The AVJ Index

The AVJ (AI vs Job) Index measures how much pressure AI puts on a job today, rung by rung, on a dated 0–100 scale. Every score links to its formula and its evidence, so anyone can check it and argue with it.

The water-level scale

0–20 Dry Land: AI barely reaches the work.
20–40 Ankle Deep: AI helps with parts of the job.
40–60 Waist Deep: AI can do much of the routine work.
60–80 Neck Deep: AI can do most of the work at this rung.
80–100 Underwater: AI already does the work reliably.

One job, five rungs

AI does not hit a junior and a senior the same way, so every job is scored on five rungs of the career ladder: L1 Executor, L2 Professional, L3 Expert / Designer, L4 Lead / Manager and L5 Owner / Principal. Job titles are labels mapped onto these rungs.

How a score is built

Capability Atlas: what AI can do now. 23 human capabilities in five families (Body, Perception, Core cognition, Executive mind, Social), plus three human statuses that are not capabilities: legal accountability, real presence and human authenticity. AI is scored on a human-referenced scale: 2 Beginner, 4 Average adult, 6 Competent professional, 8 Expert, 9 Virtuoso, 10 Best humans, 11 Superhuman.
Two numbers per capability. Reliable is what AI does at least 80% of the time in typical use; only Reliable counts in the index. Peak is the best documented result; we report it as a future warning, never in the score.
Job demand. Each task and each rung needs certain capabilities at certain levels, starting from the O*NET database. For critical capabilities the weakest link decides whether AI can do a task; supporting capabilities decide how well. The weakest link also marks where human value remains.
Real-world evidence. What AI actually does at work (DOES), what companies pay for (PAYS) and what employment data shows (SHOWS). Layoffs count only after an AI Washing check.

Rules we keep

Every AI score is tied to a named benchmark or source, graded A (primary or peer-reviewed), B (credible secondary) or C (vendor claim or judgment).
Readings are dated. Old readings are never deleted or rewritten.
The same yardsticks apply to every job.
Adding to a list (a capability, an employer type) is a minor version. Changing a weight or formula is a major version, and old readings are recomputed side by side.

Your AVJ

The official AVJ describes the job. Your own risk also depends on who you work for and what they decide. Your AVJ will give two separate numbers, Threat and Opportunity, from your employer type, their motives and your own skills. It is coming later and never changes the official score.

Disagree with a number? Tell us. Corrections with evidence are listed in the changelog.