How Fast Is Work Changing?

Occupational drift landscape

The federal government defines each occupation by a set of canonical task statements. Our agents analyze live job postings across the U.S. and compare every duty sentence against these standards. When employers consistently require duties that aren't in the federal definition, we call it occupational drift.

This isn't keyword scraping. When work changes, the task structure shifts — duties stretch, recombine, and expand in ways that a keyword search would miss entirely. AI may be driving some of these shifts, but you won't find it by searching for "AI" in job descriptions. You find it by measuring how the duties themselves are evolving relative to the federal standard. That's what we do.

—

Occupations with
verified drift

—

Independent duty
clusters detected

—

Employers
confirming

The drift spectrum

Not all occupations are changing at the same pace. We group them by how many independently verified duty clusters have emerged beyond the federal standard. Every signal below was confirmed by multiple employers across multiple U.S. metro areas.

Which occupations are changing fastest?

Ranked by number of independent duty clusters — duties confirmed by multiple employers across multiple metro areas that the federal O*NET standard doesn't capture. Click any occupation to see the full evidence trail, including the actual employer duty sentences and a semantic map of how the task space has shifted.

View all occupations with drift →

How we measure drift. Our agents collect live job postings across U.S. metro areas, parse every sentence, and classify each as a duty, qualification, or boilerplate. Duty sentences are embedded using JobBERT and semantically compared against all canonical O*NET task statements for the occupation. Sentences that don't match any federal task proceed through a five-gate validation pipeline: quality filtering, multi-employer clustering, LLM-generated labeling, specificity checking, and national persistence tracking. A duty cluster is only reported when it has been independently confirmed by multiple employers in multiple metro areas with diverse sentence language. This is semantic measurement against a stable ontology baseline — with sentence-level traceability to the actual employer who wrote it.

Sources: O*NET® 30.2 Database (federal task standards) · BLS Employment Projections · NCES CIP-SOC Crosswalk · semanticinsight.tech agent pipeline (employer-confirmed duties from live job postings) · Data updates continuously.