Williams's Keywords method applied to the word at the center of the AI labor transition. Skill in Old English meant discernment—the capacity to distinguish, to judge. By the Middle English period it acquired the meaning of practical ability: doing something well with hands or mind. The industrial revolution sharpened the word toward execution: skilled labor required apprenticeship, unskilled labor did not. The computer era narrowed further: technical skill meant proficiency in specific languages, frameworks, tools. Each narrowing served economic functions—organizing labor markets, distributing reward, producing professional identities. The AI transition is forcing a reversal. When the machine handles implementation, skill must expand to recover what the narrowing suppressed: judgment, taste, architectural instinct, the capacity to decide what should be done rather than merely how to do it. The expansion is uncomfortable because institutions (compensation structures, credentialing systems, hiring practices) remain organized around the narrowed meaning. The lag produces real crisis for practitioners whose identities were built on technical execution.
The industrial narrowing of skill served the factory system's need for a distinction between workers requiring training (skilled, higher-paid, hierarchically superior) and workers requiring minimal instruction (unskilled, lower-paid, easily replaced). The distinction was economic but also moral: skill was associated with virtue—patience, dedication, the willingness to undergo discipline in pursuit of mastery. The digital era's further narrowing—technical skill as the dominant usage—served the technology industry's need for a workforce defined by measurable capabilities. The narrowing excluded from 'skill' the very capacities (judgment, discernment, strategic thinking) that the word's earlier history had included. These broader capacities were either treated as innate (the province of founders and executives) or irrelevant (the province of humanities and management).
The AI transition's restoration of the broader meaning is not comfortable. Ascending friction—the elevation of difficulty from execution to judgment—is real, but the new friction is disembodied. The old skill was embedded in practice: the hand-weaver's discernment about cloth quality was inseparable from the capacity to weave. The AI-age skill—judgment about what should be built—floats free of the embodied practice that previously grounded it. Whether discernment can survive the loss of its practical substrate is an open question. Williams would have noted that the answer depends on whether institutions develop methods for cultivating judgment—something the apprenticeship system provided for execution but that no equivalent structure currently provides for the vertical shift the AI transition demands.
The keyword's transformation reveals a political contest. The broader meaning (skill as discernment) is more democratic—it includes capacities that the narrowed technical meaning excluded. But the broader meaning is also less legible: judgment cannot be tested on a coding assessment, taste cannot be certified by a boot camp. The illegibility creates governance problems. How do institutions distribute reward when the valuable capacity resists measurement? The dominant culture's response may be to develop new measurement systems (judgment rubrics, taste portfolios) that restore legibility by converting the illegible back into the measurable. Or it may abandon measurement and rely on proxy signals (years of experience, institutional affiliation, demonstrated past performance). Either path has political consequences for who gets recognized, rewarded, and included.
Williams did not treat skill as a standalone entry in Keywords (1976), but the semantic trajectory is implicit in his entries on labour, work, art, and creative. The AI-transition application of the method to skill follows Williams's framework precisely: track the word's changing meaning, show how the changes reflect social transformations, reveal what the current usage conceals.
Skill meant discernment first. The word's oldest meaning—judgment, the capacity to distinguish—was suppressed by industrial and digital narrowings.
Industrial capitalism narrowed it to execution. Skilled vs. unskilled labor organized markets, hierarchies, and identities around technical capability.
AI forces vertical expansion. From execution to judgment, from how to do to what should be done—a recovery of the original meaning under radically different conditions.
Institutions lag the expansion. Compensation, credentialing, hiring remain organized around the narrowed meaning, producing crisis for practitioners.
The keyword is a battlefield. Whose meaning prevails will determine who is valued, trained, and rewarded.