The optimists are winning, for now.
The Summary
- New Gallup polling shows 30% of American workers now use AI daily or weekly at work, but anxiety about job replacement is rising alongside adoption.
- Two-thirds of AI users report productivity gains, but the gap between managers and individual contributors tells a different story about who benefits.
- Non-users cite preference, ethics, and privacy concerns, not just lack of access.
The Signal
The productivity numbers look good on paper. Two-thirds of workers using AI say it makes them more efficient. Forty percent work at organizations that have formally adopted AI tools. But drill into the manager versus worker split and you see the real dynamic: 70% of leaders using AI report efficiency gains versus just over half of individual contributors.
That gap matters. It suggests AI isn't a rising tide lifting all boats equally. Managers get leverage. Workers get... what, exactly? Faster task completion that justifies headcount reduction?
"The productivity gains are concentrating at the top, where decisions about headcount get made."
Scott Segal, the 53-year-old social worker quoted in the poll, uses AI daily to connect elderly patients with healthcare resources. He finds it useful. He also thinks it will replace him and is "planning ahead." That's the tell. The people actually using AI in the trenches aren't blind to where this goes. They're the ones with the clearest view of their own replaceability.
The non-user cohort is equally revealing:
- Some prefer working without AI (selection bias for roles where AI adds friction, not flow)
- Others cite ethical opposition (likely healthcare, education, creative fields)
- Privacy concerns remain a blocker (especially in regulated industries)
This isn't technophobia. It's rational risk assessment. If you work in a "replaceable field or trade," as Segal puts it, why would you enthusiastically train your replacement?
The Implication
The divergence is widening. Thirty percent adoption with rising replacement anxiety means we're past experimentation and entering the sorting phase. Some roles get augmented. Others get automated. The difference isn't always obvious until it's done.
If you're in the individual contributor tier and seeing managers get more productivity lift than you are, that's not a usage problem. That's a structural signal. Watch what gets measured. If your AI-assisted output just means your team gets smaller next quarter, you're not being augmented. You're being optimized away.