关键实验数据来自为期8个月的研究：在未被强制的情况下，使用AI工具的员工工作速度提高、任务范围扩大，并主动延长至夜晚与清晨，体现出“内在激励”增强。长期背景对比显示，美国员工敬业度约仅三分之一，且该比例在20年内几乎未变化，表明传统激励投入（数十亿美元级）未显著改善参与度。

行为与生产率指标呈现分化趋势：个体报告生产率提升最高达5倍，同时“难以停止工作”的现象增强。另一项研究显示认知偏差显著——开发者实际效率下降19%，却主观认为提升24%，显示AI改变感知与真实产出之间的偏离。与此同时，AI降低工作“厌倦阈值”，使工作时间自然延长，削弱了传统的时间边界。

管理与结果之间存在显著统计差异：在监控环境下，42%的员工计划一年内离职，而未被监控群体为23%，差距达19个百分点。多数员工认为监控既不提升效率也削弱信任。整体趋势显示，AI既可通过减少低价值任务将参与度从约33%向上推升，也可能因监控与过度使用引发疲劳与流失，形成双向分化路径。

Key experimental evidence comes from an eight-month study: without mandates, workers using AI tools increased speed, expanded task scope, and voluntarily extended work into evenings and early mornings, indicating stronger intrinsic motivation. Long-run benchmarks show only about one-third of US workers are engaged, a ratio largely unchanged over 20 years, despite billions spent on engagement efforts.

Behavioral and productivity metrics diverge: individuals report productivity gains of up to fivefold alongside increased inability to disengage. Another study shows cognitive miscalibration—developers were actually 19% slower while believing they were 24% faster, indicating a gap between perceived and real output. AI also lowers the “tedium threshold,” naturally extending working hours and eroding traditional time boundaries.

Management approach correlates with outcomes: 42% of monitored employees plan to leave within a year versus 23% of unmonitored peers, a 19-percentage-point gap. Majorities report surveillance neither improves productivity nor preserves trust. The overall trend shows AI can raise engagement above the ~33% baseline by removing low-value tasks, but can also induce fatigue and attrition under monitoring and overuse, creating a bifurcated trajectory.