AI productivity tools deliver measurable gains, but the benefits concentrate among high-skill workers while lower-wage employees face job displacement. New data from three years of AI adoption reveals a stark divide that contradicts vendor marketing promises.

Research shows AI boosts productivity most dramatically for knowledge workers in software engineering, data analysis, and professional services. These workers use AI to automate routine tasks, accelerate coding, and handle information synthesis. Productivity gains reach 30-40 percent in some domains. Simultaneously, AI adoption creates downward pressure on wages and employment prospects for administrative staff, junior analysts, and routine knowledge workers whose core responsibilities overlap with AI capabilities.

The paradox emerges clearly: workers whose jobs are most threatened see productivity improvements as AI tools displace their responsibilities. A software engineer gains efficiency through AI-powered code completion. An administrative assistant faces automation of scheduling and document management. The same technology that enhances one worker's output threatens another's relevance.

Companies deploying AI rarely retrain displaced workers. Instead, attrition absorbs the impact. Hiring freezes and layoffs follow productivity gains as businesses realize they need fewer people to handle the same workload. This pattern holds across industries from finance to customer service.

Vendor narratives stressed universal productivity uplift. The reality proves selective. AI excels at tasks requiring pattern recognition and information synthesis but struggles with novel problem-solving and human judgment. This creates a skills-based split where AI augmentation works best for workers already commanding higher salaries and market power.

The wage impact follows predictably. Workers in roles where AI provides substantial augmentation see compensation pressure as their individual output increases without corresponding pay increases. Workers in roles where AI provides displacement face redundancy. Middle-tier roles compress fastest.

Three years of productivity data reveals a technology that amplifies existing inequalities rather than leveling opportunity. The gains are real and substantial for some. The costs fall hardest on