AI productivity gains are real but acutely unequal. Three years of deployment data now shows a stark pattern: workers in routine, well-defined roles experience spectacular efficiency improvements, while knowledge workers and specialists often see diminished output or no gains at all.

The disparity runs counter to industry marketing. Vendors promised universal productivity uplift. Instead, AI excels at automating standardized tasks. Customer service representatives using AI-assisted tools process inquiries 35-40% faster. Data entry workers complete jobs in half the time. These gains concentrate among workers in roles most vulnerable to displacement.

Knowledge workers face different pressures. Lawyers, engineers, and analysts must verify AI output, adding friction to workflows. Context-switching between thinking deeply and checking machine-generated work creates cognitive overhead. Some researchers report AI tools slow them down when tasks require domain expertise or creative problem-solving.

The productivity equation inverts by role. Those with the least bargaining power gain the most from efficiency tools but also face the greatest job security risk. Those with the most job security struggle to benefit. The workers AI helps most are exactly those whose roles are most threatened by automation.

This pattern creates a cruel calculus. A customer service rep becomes 35% more productive but faces potential workforce reduction. An engineer stays similarly productive but keeps their role intact. The same technology produces opposite career outcomes.

Companies deploying these tools rarely grapple with this contradiction directly. Productivity metrics drive adoption decisions. HR decisions follow separately. In this gap, the real story unfolds: AI productivity gains funnel through different labor markets with different consequences.

The evidence suggests AI's productivity benefit tracks closely with job displacement risk. That correlation isn't coincidental. The easier AI finds a task to optimize, the closer that task sits to automation. Workers whose roles improve most measurably under AI also inhabit the roles most likely to vanish.