Methodology

FutureProof shows three numbers for each US occupation: the median wage, the job outlook, and how exposed the job is to AI. Every number is built reproducibly from authoritative public sources, joined on the standard US occupation taxonomy (SOC 2018). This page explains, in plain language, where those numbers come from and what they do and don’t mean.

The two AI-exposure axes

We show AI exposure on two separate measures, and we never combine them into a single score. One answers “what is AI actually used for now?” and the other answers “what is exposable in principle?”

• AI exposure today (observed) — from the Anthropic Economic Index (AEI): the share of an occupation’s tasks that show up in real AI usage today. We publish it as a 1–100 relative rank across US occupations. Used under CC-BY with attribution.
• AI horizon (theoretical) — from the Felten, Raj & Seamans AI Occupational Exposure (AIOE, 2021): how exposable an occupation’s required abilities are to AI in principle, as tools improve. Also published as a 1–100 relative rank.

Both are a relative rank among US occupations, not a probability of job loss. A high rank means “AI overlaps substantially with this job’s tasks,” not “this job disappears.”

Wages & outlook

• Median wage — BLS Occupational Employment and Wage Statistics (OEWS), May 2025 release: the national annual median wage for each occupation.
• Outlook & growth % — BLS Employment Projections, 2024–34: the projected percent change in employment over the decade, which we bucket into growing, stable, or declining.

Automation vs augmentation

The AEI splits real AI usage into augmentation (AI assisting a person, ~57%) and automation(AI doing the task, ~43%). A reliable per-occupation split does not exist yet, so where we can’t show an occupation-specific figure we show this dataset-wide split and label it as such.

Limitations

• Exposure is a relative rank among occupations, not a job-loss probability.
• The observed measure (AEI) reflects current AI usage and will shift as adoption changes.
• The horizon measure (AIOE) is theoretical — it describes what’s exposable in principle, not what AI does today.
• “Human-edge” skills are editorial, mapped by occupation family rather than derived per occupation.

Attribution

Data: BLS OEWS, BLS Employment Projections, Anthropic Economic Index (CC-BY), Felten, Raj & Seamans AIOE (2021).

FutureProof is not affiliated with, sponsored by, or endorsed by Anthropic. Full provenance: the AEI dataset on Hugging Face and the AIOE source.