41 个数字：AI 已主流化，但价值鸿沟正在扩大

A majority of American workers now use AI on the job, making adoption mainstream even though capability remains uneven.

Enterprises can report better production capability while still failing to generate returns that exceed AI spending.

Executive optimism is rising faster than organizations' ability to meter usage, redesign workflows, and prove business value.

AI spending and outcome data show a widening power-law gap between ordinary users and a small group of sophisticated operators.

Subscription and usage shares suggest that users increasingly combine multiple model providers rather than relying on a single default.

A meaningful share of coding-agent users now delegate work that would otherwise take many hours, indicating a shift from assistance to execution.

When employees perceive official tools as inadequate, they adopt unsanctioned systems even when they believe the behavior violates policy.

Productivity gains can be offset when employees spend more time checking, correcting, and supervising model output.

Occupation-specific use is concentrated in tasks that are already text-rich, repeatable, and easy to evaluate.

Owning AI tools is common, but only a small minority of companies consistently connect them to measurable business value.