传统对 AI 自动化的乐观观点经常引用 ATMs 的例子，它自动化了银行柜员的核心任务，却让柜员的就业人数在三十年间持续上升。这种现象可以由 19 世纪的 Jevons 悖论来解释，即技术降低了现有服务的成本，从而扩大需求并增加整体就业。然而，该悖论仅在技术能提高现有服务模型效率时才成立。一旦技术允许客户透过照片存入支票或自主转帐，自我服务机制便以银行外的客户取代了柜员，使得内部的效率悖论失效。

社会学家 Jonathan Gershuny 于 1978 年首次指出，向自我服务经济的转型将劳动力直接转移到了家庭中。诸如洗衣机、自助结帐和线上旅游预订等历史先例表明，技术的成功并非借由将专业人员的工作自动化，而是透过使消费者能够在没有该工作的情况下生活。 AI 将这种转变延伸到了手工业和精英专业领域，例如房屋所有者使用聊天机器人诊断锅炉，或患者自行解读检测结果。正如 MIT Cryptoeconomics Lab 的 Christian Catalini 所指出，当 AI 将执行成本降至接近零时，人类验证成为了主要的限制因素，而企业则将这一负担转嫁给了客户。

这种从市场回到家庭的劳动力转移具有深远的宏观经济影响。尽管企业记录了更低的成本，且国民帐户登记了生产力的提升，但消费者花费在执行这些自我服务任务上的时间却完全被排除在劳动统计数据和 GDP 衡量之外。这一趋势悄然逆转了数世纪以来将家务劳动转化为有形、专业化市场活动的转变。因此，依赖传统经济指标的政策制定者将对转移到公众身上的真实且未被衡量的劳动力成本保持盲目，因为 AI 驱动的自我服务将继续扩大。

Traditional optimistic views on AI automation often cite the example of ATMs, which automated bank tellers' core tasks but saw teller employment rise for three decades. This phenomenon is explained by the 19th-century Jevons paradox, where a technology makes an existing service cheaper, expanding demand and overall employment. However, this paradox only holds when the technology improves the efficiency of the existing model. Once technologies enabled customers to deposit cheques via photos or transfer funds independently, the self-service mechanism replaced tellers with the customers themselves, rendering the internal efficiency paradox obsolete.

First identified by sociologist Jonathan Gershuny in 1978, the transition toward a self-service economy shifts labor directly to the household. Historical precedents like the washing machine, self-checkouts, and online travel booking demonstrate that technology succeeds not by automating a professional's job, but by enabling consumers to do without it. AI extends this shift to manual trades and elite professions, such as when homeowners diagnose boilers using chatbots or patients decode test results. As Christian Catalini of the MIT Cryptoeconomics Lab notes, when AI reduces execution costs near zero, human verification becomes the primary constraint, which companies offload onto customers.

This labor transfer from the market back to the household has profound macroeconomic implications. Although companies record lower costs and national accounts register productivity gains, the hours consumers spend performing these self-service tasks are completely omitted from labor statistics and GDP measures. This trend quietly reverses the centuries-long shift of domestic labor into visible, specialized market activity. Consequently, policymakers relying on traditional economic indicators will remain blind to the true, unmeasured labor costs shifted onto the public as AI-driven self-service continues to expand.