90% of US working hours are spent on tasks not covered well by AI benchmarks
Mercor estimates 90% of U.S. working hours involve tasks poorly covered by current AI benchmarks.
Mercor's Benchmark Map links 29 public, agentic AI benchmarks to the U.S. Department of Labor's O*NET taxonomy of 923 occupations and 18,797 tasks. Only 208 occupations have at least one task matched to five or more benchmark prompts, and six of 22 job families have none. Combining that coverage with estimated task time shares, Mercor says 90% of U.S. working hours fall on tasks with little or no benchmark coverage. Coverage is skewed toward high wage-bill, AI-exposed work, such as lawyers' tasks, while similar work by legal secretaries and court clerks is barely covered. The authors call the results provisional because the 29-benchmark sample is incomplete.
- Only 208 of 923 occupations have a task matched to at least five prompts.
- Six of 22 job families have no occupation meeting that coverage bar.
- Coverage favors large, AI-exposed wage bills over similar lower-paid support roles.
- The estimate maps 29 agentic benchmarks onto 18,797 O*NET tasks.
- Countries
- United States
From the original article
11 min read at mercor.com
Each tile represents one of the 923 O*NET occupations making up the US labor force, grouped into 22 job families. Occupations are shaded according to the share of their tasks which are represented in our initial list of AI benchmarks. A single benchmark prompt can be mapped to multiple O*NET tasks.
By speckx · mercor.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.