What “AI exposure” means

The one-sentence definition

AI exposure is how much of an occupation's tasks overlap what today's AI can help with — measured by reading the work, task by task, against what the models can do.

It is a property of the work, not a verdict on the worker. And it is not:

• Not a prediction the job disappears. Overlap is not removal.
• Not a measure of adoption. It says what AI could touch, not what anyone is actually doing in practice today.
• Not “jobs lost.” A high-overlap occupation can be — and often is — growing and hiring.
• Not automation. Most overlap shows up as copilot work: AI assists a task, a person still does the job.

A worked example: Customer Service Representatives

Take one of the most-exposed occupations in the country. Here is its honest coordinate — and why “most exposed” does not mean “most endangered.”

↔94th-percentile task overlap — yet about 341,700 openings a year (-5.5% projected, BLS) . What exposure means →

High task overlap, and still hundreds of thousands of openings a year. The work is being reshaped — routine questions handled by AI, people moving to the harder, human parts — not erased. That is the shape of most “exposed” work. See the full breakdown on the Customer Service Representatives page: what's already handed to AI, what stays human, and where it's heading.

How to read any exposure figure

1. Read the band as “how much of the work overlaps,” not “how likely the job ends.”
2. Look for the counterweight. Across the site, an exposure number is paired with its openings and its copilot-vs-hand-off lean. High overlap + high openings is the most common, and most hopeful, pattern.
3. Check what stays human. Every role page names the low-overlap tasks — the durable core the numbers say AI is least likely to touch.
4. Remember it's two studies, not a crystal ball. Exposure here folds two published measures (below). They estimate potential overlap, not the future.

Then why measure it at all?

Because the alternative is rumor. “AI is coming for your job” is everywhere and almost never attached to a number, a source, or a definition. Exposure, read honestly, replaces fear with a coordinate: here is how much of this specific work overlaps what AI can do today, here is what that does and doesn't mean, and here is the source. That's the whole project.

Sources for this page

Every figure above traces to a named public dataset and the exact release below — not hand-written opinion. See the full methodology for what each measure does and does not mean.

• O*NET 30.3 U.S. Department of Labor / National Center for O*NET Development
• BLS Employment Projections 2024–2034 U.S. Bureau of Labor Statistics
• “GPTs are GPTs” (Eloundou et al.) arXiv 2303.10130 OpenAI / academic
• AI Occupational Exposure (AIOE) Felten, Raj & Seamans academic

@misc{singulariki-exposure-means.html,
  title  = {What AI exposure means},
  author = {{Singulariki}},
  year   = {2026},
  note   = {O*NET 30.3; BLS Employment Projections 2024–2034; “GPTs are GPTs” (Eloundou et al.) arXiv 2303.10130; AI Occupational Exposure (AIOE) Felten, Raj & Seamans. Accessed June 5, 2026},
  url    = {https://singulariki.com/exposure-means.html}
}