The AI revolution isn't replacing jobs yet, but it's already creating winners and losers inside companies. New research from Anthropic reveals a growing divide between employees who've mastered AI tools and those still struggling to adopt them, raising fresh concerns about workplace inequality even as fears of mass displacement haven't materialized. The findings, shared exclusively with TechCrunch, mark one of the first data-driven looks at how AI adoption is reshaping productivity dynamics across enterprise workforces.
Anthropic just dropped data that confirms what many in tech suspected but few could prove: AI isn't killing jobs en masse, but it's splitting workforces into haves and have-nots based on who can actually use these tools effectively.
The AI company's internal research, drawn from usage patterns across enterprise customers, shows experienced users extracting dramatically more value from AI assistants than their less-practiced colleagues. While Anthropic didn't release specific productivity metrics, sources familiar with the data say the gap between power users and casual adopters is widening month over month, creating what one researcher called "a tale of two workforces."
"We're not seeing the job replacement narrative play out yet," a person close to the research told TechCrunch. "But we are seeing inequality emerge in who benefits from these tools, and that's arguably more concerning for the average worker."
The findings land as companies from Goldman Sachs to Walmart race to deploy AI across their operations, often with minimal training or support for employees expected to integrate these tools into daily workflows. Anthropic's data suggests that hands-off approach is backfiring, creating internal disparities that could reshape team dynamics and performance reviews.
What separates power users from stragglers isn't rocket science. According to the research, experienced users ask more specific questions, iterate on responses, and understand how to structure prompts for complex tasks. They've developed what researchers call "AI fluency" - an intuitive sense of what these tools can and can't do. Meanwhile, less experienced colleagues often give up after a few failed attempts or stick to basic queries that barely scratch the surface of what's possible.
