困在AI裡的工作：非解放而更多任務

• Generative AI tools lower activation energy for task initiation, enabling workers to begin projects during traditionally protected time (breaks, lunch, meetings) by queuing prompts for asynchronous processing[1][3] • Multi-workflow juggling increases cognitive load as employees manage multiple AI-enabled processes simultaneously, each with different interfaces, output quality expectations, and integration requirements[1][2] • AI-assisted work creates secondary labor demands: code review cycles expand as engineers must validate AI-generated or AI-assisted outputs, including informal coaching of colleagues using 'vibe-coding' approaches[2] • The acceleration cycle operates as: faster task completion → raised speed expectations → increased AI reliance → expanded task scope → higher work density and quantity[5]

Organizations face a critical inflection point: without deliberate governance frameworks ('AI practice' standards and norms), workplace AI adoption will drive unsustainable intensity, talent attrition, and quality degradation despite apparent productivity metrics[1][4]. The convergence of job displacement risk (10.4 million positions by 2030) with burnout acceleration suggests a bifurcated labor market where remaining workers absorb expanded responsibilities while experiencing diminished autonomy and recovery time. Companies must implement work sequencing, coherent project phasing, and intentional pauses to preserve cognitive capacity and decision quality; failure to do so risks masking organizational dysfunction as productivity gains[1].