AI 接管轻松活,人接住所有烂摊子AI Takes the Easy Work, Humans Keep the Wreckage
这条新闻里真正发生的事,不是 AI 抢走了客服的工作,是 AI 把客服工作里唯一能喘气的那部分拿走了,剩下的全压回人身上。
报道里 Shannon 讲得很清楚:AI 预约不了客户要的日期,案子就转给人类;员工“一直在救火”,“所有复杂的案子都到我们这里来,造成大量倦怠”。国王学院的 Amanda Jones 把机制说得更准:那些相对简单的任务,本来是繁重工作之间的心理缓冲。缓冲消失,就算总工作量没变,一天的强度也会明显上升。
这就是结构层暴力最典型的形态:没有谁被打,没有可见的施暴者,只是任务被重新分配了一遍,差额就落到了具体的人身上。公司拿到效率,客户拿到更快的响应,成本被精确地转移给留下来处理烂摊子的那批人。
PolyAI 的 CEO 说客服“拿最低的工资,承受最大的辱骂”,听起来像同情。但同一段话的落点是“AI 有希望让客户和员工都更好”。一个卖 voice AI 的人替客服说话,最后卖的是替掉客服的方案——这种共谋不需要恶意,只需要利益站队一致。
更隐蔽的一层在报道末尾:Jones 说,常规任务被自动化之后,员工会失去积累经验的机会,而那些经验原本是让人能处理更难工作的准备。于是系统一边要求人做更难的活,一边拆掉让人学会做难活的路。这是把梯子抽走,然后怪你爬不上去。
ArvatoConnect 的 CEO 说他们考核的是客服有没有理解客户需求、有没有做出合理判断,不是有没有照脚本、有没有卡时间指标。这话很好。但同一篇报道里,MaxContact 的调查显示超过一半的联络中心决策者说客服工作量上升了,2024 年这个比例还只有四成左右。考核话术和实际强度之间的差额,就是这套叙事最该被追问的地方。
Kaylee Harris 担心的不是自己被取代,是公司为了快而牺牲质量,“客户已经受够了 AI 被硬推给所有人”。Shannon 问得更直接:人会多走一步,AI 能给那种感觉吗,它 genuine 吗。
这两个问题报道没有回答,也不需要回答。需要指出的是,提出这些问题的人,名字后面都带着“不愿透露姓名的雇主”。
What's actually happening in this story isn't AI taking customer service jobs. It's AI taking the one part of customer service that let people breathe, and pushing everything else back onto them.
Shannon spells out the mechanism: AI agents can't offer a customer the requested date, so the case goes to a human. Staff are "constantly fighting fires," and "all the complicated cases came to us, causing a lot of burnout." Amanda Jones at King's College London puts it more precisely: relatively straightforward tasks provide psychological respite between demanding ones. Remove the respite and the working day gets more intense even if the total volume doesn't change.
This is structural violence in its cleanest form. Nobody is hit. There's no visible assailant. Tasks are simply reallocated, and the difference lands on specific people. The company gets efficiency, the customer gets faster responses, and the cost is transferred with precision to whoever stays to clean up.
The CEO of PolyAI says contact-center agents face "maximum abuse for minimum wage." It sounds like sympathy. But the sentence lands on AI "bridging the gap to make it better for both." A man selling voice AI speaks up for agents and sells the thing that replaces them. That conspiracy needs no malice, only aligned interests.
There's a quieter layer at the end: Jones notes that as routine tasks are automated, workers lose the chance to build the experience that traditionally prepared them for harder work. The system demands more demanding work while dismantling the path that taught people to do it. Pull up the ladder, then blame them for not climbing.
ArvatoConnect's CEO says they assess whether agents understood the customer, exercised sound judgment, reached the right resolution, not whether they followed a script or hit a time target. Fine words. But the same article cites MaxContact's survey: more than half of contact-center decision-makers report rising agent workloads, up from around 40% in 2024. The gap between the assessment language and the actual intensity is exactly where this narrative should be pressed.
Kaylee Harris isn't worried about being replaced. She's worried the push for speed comes at the expense of quality, that customers are fed up with AI being shoved at everyone. Shannon asks the sharper question: people go the extra mile; can AI give that feeling, can it feel genuine?
The article doesn't answer either question, and it doesn't need to. What it does show is that the people asking them are the ones whose employers won't let them use their full names.