Competitive Compression

In the AI Story

The mechanism operates across the professional middle class — law, accounting, journalism, marketing, software development, financial analysis, middle management. What unites the affected workers is not their industry but their position in the skill distribution: skilled enough to have been valuable in the pre-AI economy, but concentrated in implementation tasks AI now performs comparably. A junior lawyer using AI for document review produces output that previously required several associates. The associates are not fired immediately; the premium their work commanded is compressed, and the compression spreads through the firm's billing structure, hiring patterns, and partnership trajectories.

The compression is experientially distinctive. It does not produce the sharp discontinuity of layoff or displacement. It produces the slow grinding of stagnation — bonuses that fail to rise, promotions that take longer, colleagues whose AI-augmented productivity makes one's own output look inadequate, career paths that lead to positions whose compensation structures have been quietly revised downward. The worker remains employed; her position is unchanged; the numbers on her pay stub show modest or no growth. The aggregate labor-market statistics show a functioning economy. Her relative distributional position, invisible to aggregate measurement, is eroding.

Milanovic's research on the globalization elephant documented this phenomenon's subjective dimension: populations experiencing relative stagnation while absolute conditions were stable experienced the condition as decline, because human beings evaluate economic position in relative terms. The AI valley produces the same subjective experience at compressed timescale. The professional whose income is unchanged but whose AI-augmented colleague is pulling ahead, whose industry is restructuring around capabilities she has not developed, whose children ask questions about the future she cannot answer — experiences the transition as loss, regardless of what aggregate productivity statistics report.

The political implications compound the distributional mechanics. The compressed professional middle class includes journalists, teachers, academics, creative professionals, public-sector workers — the populations that shape the narratives through which societies understand themselves. Their distributional experience will be disproportionately represented in public discourse. This proximity to narrative production is both advantage and risk: it can accelerate political pressure for institutional response, and it can center the discourse on the experience of the professional middle class while neglecting populations lower on the distribution whose AI-related displacement is more severe but less visible.

Origin

The mechanism has deep roots in labor economics — David Autor's skill-biased technical change literature, Daron Acemoglu's work on task displacement and automation — but its specific AI-era form was first documented in detail in 2023–2024 studies by researchers at Microsoft, MIT, and Stanford showing that AI tools raise the productivity of less-experienced workers more than experienced workers, compressing the wage premium for experience itself. Milanovic's contribution is situating the mechanism within the broader distributional framework, connecting it to the capital-labor split and the homoploutic elite whose gains compound as the middle compresses.

Key Ideas

Supply expansion, not displacement. AI does not fire workers; it expands the supply of workers capable of producing any given quality of output. Increased supply depresses price. The premium was scarcity; scarcity is reduced.

Invisible to aggregate statistics. Employment remains tight; average wages show modest growth; the aggregate picture is of a functioning labor market. The compression is in the trajectory, which aggregate statistics do not capture.

Experientially subtle, politically consequential. The compressed professional does not experience displacement but stagnation. Stagnation is slower to produce political mobilization than displacement, which buys time for institutional response — or forecloses it.

The narrating class is being squeezed. Journalists, teachers, academics, public-sector professionals — the populations whose work shapes public discourse — are disproportionately represented in the compressed valley. Their proximity to narrative production will shape how the AI transition is politically understood.

Compression is the valley. The AI elephant's valley is not a population of the unemployed; it is a population of the compressed. Institutional responses designed for displacement (unemployment insurance, trade adjustment assistance) are structurally mismatched to the condition compression produces.

Debates & Critiques

Some economists argue that compression is transitional — that once labor markets fully adjust, new equilibria will emerge in which AI-complementary skills command premiums reflecting their scarcity. Milanovic's framework is skeptical: the history of previous transitions shows that 'transitional' compression often becomes permanent when institutional moderation fails to arrive on time, and the AI transition's speed compresses the adjustment window to a degree that makes 'transition to new equilibrium' a plausible-sounding phrase for 'permanent downward shift.'