Despecification is the process by which AI tools convert highly specialized professional skills into generic capabilities accessible to anyone with the tool. A backend engineer's years of specialization become less organizationally specific when a designer can use Claude to build backend features competently. The skill itself may not decline, but its scarcity value collapses because alternatives multiply. This is not the elimination of expertise but the reduction of its specificity: the knowledge that was deployable only in narrow professional roles becomes deployable across roles through AI mediation. In Williamson's framework, despecification weakens the bilateral dependency justifying employment relationships—the firm has less reason to retain the specialist when the market (or AI-augmented generalists) can supply adequate substitutes.
The mechanism operates through three channels. First, capability generalization: AI provides competent performance across domains previously requiring specialized training, enabling workers to operate outside their historical specializations. A Napster designer who never touched backend code can now implement complete features end-to-end because Claude handles the translation between design intent and technical implementation. Second, skill-floor elevation: the minimum capability required to perform professional work rises while the ceiling remains roughly constant, compressing the range between novice-with-AI and expert-without-AI. Third, context-independence: skills that were valuable precisely because they were embedded in particular organizational contexts (knowing this codebase, this customer base, this regulatory environment) become less specific as AI provides general-purpose bridges across contexts. A lawyer using AI to draft briefs develops capabilities portable between firms because the AI generalizes across legal contexts rather than specializing to any single one.
The despecification is asymmetric across professional strata. Junior and mid-level professionals whose value resided primarily in technical execution face the most severe erosion: their specialized training in particular languages, frameworks, or analytical methods becomes less organizationally specific when AI performs those tasks adequately. Senior professionals whose value included substantial judgment alongside technical skill face more complex dynamics. The technical dimension despecifies (the ability to write code or run analyses), but the judgment dimension respecifies (the ability to evaluate whether the code or analysis serves purpose). The net effect depends on whether the professional can credibly claim and demonstrate judgment value to the organization—a transition many find wrenching because it requires reconceiving professional identity from 'doer' to 'evaluator,' from 'maker' to 'director.' The senior software architect in The Orange Pill who felt like 'a master calligrapher watching the printing press arrive' was diagnosing this identity crisis with precision.
The historical precedent clarifies the stakes. When electrification arrived in factories, the specific skills of the line-shaft mechanic—who understood the mechanical power transmission through belts, pulleys, and gears—were despecified by the electric motor allowing individual machine control. The mechanic's knowledge became obsolete not because machines no longer needed maintenance but because the type of maintenance shifted from mechanical transmission systems to electrical systems, and the old specific knowledge did not transfer. Some mechanics retrained; many did not or could not. The transition took decades, softened by generational turnover and the gradual expansion of electrically-skilled labor supply. AI's despecification is compressing that timeline from decades to years, and the workers experiencing it do not have the luxury of generational adjustment. A 45-year-old programmer whose React expertise is being despecified cannot wait for the next generation to acquire AI-era skills. She must acquire them herself, mid-career, while supporting a family and managing the psychological disruption of skills built over twenty years becoming less valuable than a subscription.
The term is new to this volume, synthesizing Williamson's asset specificity framework with the empirical reality of AI's impact on professional knowledge work. Williamson analyzed despecification in industrial contexts—how process innovations (continuous casting in steel) or technological substitutions (electric motors for line shafts) could reduce the specificity of human and physical capital. But his examples unfolded over decades, and he did not systematically address the speed of despecification as an independent variable affecting governance. The AI transition has made speed the critical factor: when specificity collapses faster than institutional adaptation, the workers bearing the adjustment costs have no governance structures to protect them, and the distributional consequences are severe and concentrated.
Specificity is relative to alternatives. A skill is specific not in absolute terms but relative to available substitutes—AI expands substitutes, reducing relative specificity.
Execution despecifies, judgment respecifies. The bifurcation is structural: technical skills become generic (anyone with AI can perform them), evaluative skills become more organizationally embedded.
Despecification weakens bargaining power. Workers whose value was specific to particular technical domains face eroding bilateral dependency with employers as AI-augmented alternatives proliferate.
Speed prevents adjustment. Despecification happening over years allows retraining and institutional adaptation; despecification happening over months produces displacement the governance system cannot absorb.
The policy response is capability investment. Public investment in judgment development, evaluative training, and the cognitive infrastructure for strategic thinking—the only response adequate to despecification at scale.