文 | 山自
硅谷的 CEO 们最近达成了一种浪漫共识：只要算力足够、模型够强，人就是多余的变量。
他们站在财报电话会、行业峰会、媒体专访里，一遍一遍复述同一个叙事：AI 重塑一切，效率解放组织，旧时代的人力包袱可以一键清退。从奥尔特曼谈 AI 介入养育，梅耶尔畅想夫妻 AI 互相协同安排家事，扎克伯格笃定大众会逐步接纳 AI 深度介入生活，这套话术的底色高度统一。技术本身可以解决所有组织、流程、成本难题，人类只是等待被优化的冗余。
但资本市场用十月的行情，狠狠泼了一盆冷水。
一、十月，市场先给 AI 狂热降温
刚刚过去的十月，宏观市场出现了一个反常信号，被贝莱德资深组合经理 Jeffrey Rosenberg 指出，原油价格与美债收益率维持了一整年、高达 80% 的强相关性，突然断裂。
在此之前，交易逻辑简单清晰：油价推升通胀→美联储维持高利率→美债收益率上行，地缘冲突是定价主线。但十月风向变了。 Rosenberg 给出的判断足够扎眼：在市场眼里，地缘政治风险，已经让位于 AI 风险。
Mag 7 集体回调、谷歌单周市值蒸发近五千亿，成为这一轮预期扭转的标志性事件。资本曾经默认，只要持续砸钱买算力、堆大模型，AI 就能源源不断兑现增长；三季度美国 GDP 数据里，信息处理设备投资暴涨 37%，大量资本开支贴着 AI 的故事向上冲。
但财报季拉开之后，市场开始分层定价 AI。 一部分企业交出了可核验的收益：桑坦德银行前三季度 AI 项目落地，创造 1.94 亿欧元业务价值；诺华、渣打银行陆续表示 AI 已经渗透进研发、营销、客户经营，抬升经营杠杆；Meta 的 AI 广告增量、谷歌云收入翻倍，都是实打实的正向案例。
而另一部分，只剩下叙事。 超大规模云厂商的资本开支已经跑赢自身自由现金流。微软时隔近十年，在市场发行 250 亿美元债券加码 AI 投入。债券市场短期愿意买单、认购超额，但 CDS 信用利差持续走高，投资者开始对冲一个现实：未来四年预估近 4 万亿 AI 相关融资需求，债券市场的承载能力不是无限的。
资本不再无脑相信 “AI = 永续增长”。它慢慢看懂一件事：AI 分两种，一种嵌入业务、和人协同创造增量；另一种，只是管理层用来降本、裁人、美化季报的包装纸。后者，就是市场正在定价的 “AI 风险”。
二、光鲜 AI 代理背后，藏着影子人力
在各大发布会和宣传稿里，AI 代理已经无所不能：帮你打电话协商账单、预约上门服务、比价采购、安排家装。OpenAI、Meta 相继入局个人智能代理赛道，描绘一个机器自主办事、解放人类的未来。
可剥开宣传外壳，行业内部是另一套运行规则。 Dazzle AI，也就是玛丽莎・梅耶尔创办的 AI 代理公司，对外毫不避讳这套模式：大模型负责检索信息、给出候选清单，一旦任务进入现实落地环节 —— 打电话沟通、预约施工、确认服务细节，很容易出现常识性错误。他们曾遇到典型故障：AI 代理把屋顶翻新项目里的窗户玻璃和餐桌玻璃混为一谈，试图把桌面玻璃装到窗户上。 遇到这类卡点，系统不会死循环消耗海量 token，而是直接转单给签约外包人员，由真人接手完成沟通。
Duckbill、Wajo AI 这类代理初创公司，干脆把 “人机转接” 做成标准化 API 产品。它们对接一批背景审核过的合同制工人，专门承接 AI 搞不定的环节：电话沟通、线下预约、网站人机验证。创始人梅根・乔伊斯直言，这套逻辑和早年 Uber 模式很像：AI 负责接单展示，人力负责兜底交付。
Meta 的 Muse 代理也曾爆出内部测试争议：对外宣传 AI 自主通话，内部测试阶段大量通话由呼叫中心真人代打。事后 Meta 解释这只是早期产品测试，但这个细节已经戳破行业共识：现阶段绝大多数能 “代办现实事务” 的 AI，本质是一层前端交互皮，真人承包商才是底层底座。
黑色幽默就在这里：很多 CEO 拿着 “AI 自主代理” 的案例向董事会汇报，论证岗位可以被替代、团队可以优化，却刻意隐去成本结构的变化 —— 正式员工裁掉，不是人力消失了，只是稳定高薪的全职岗位，变成了随时可以关停、没有社保保障的外包影子人力。 人力成本没有消失，只是转移、隐匿，从财报里显眼的薪酬项，变成零散的外部服务采购。
三、那个把两亿成本压到两百万的工程师
这两天，一张截图在脉脉、各类职场社群反复流转，引发几十万打工人共情。
故事的版本流传很广：一家互联网大厂，二十人的技术小组，维护一套高消耗系统，服务器一年烧掉两个亿，这个状态持续了整整五年。组内一名工程师，抱着解决问题的心态，连续数月加班重构底层架构，上线之后，年度成本直接压缩至两百万。 他原本期待奖金、晋升、项目嘉奖，最后等来的是整组裁员通知，自己的名字同样在名单上。
网传故事没有实锤公司主体，所以严谨来说，它属于社群传播的职场寓言。但它之所以引爆共鸣，核心不是数字夸张，而是无数从业者见过同构的现实：打工人算功劳，老板只算人头。
海外有几乎同内核的真实案例。一名美国电商开发，在本职之外利用业余时间，自研自动化脚本，帮公司每年省下数千小时人工工时。他没有等来加薪、绩效奖励，反而被管理层索要全部代码权限。紧接着，HR 面试录用了一名时薪更低的新人；公司拿到代码之后，迅速注销这名开发者权限，单方面解除劳动合同，离职补偿仅两周薪资，附带严苛竞业协议。
还有一个被反复提起的讽刺循环：有算法工程师，亲手搭建了企业用于人员盘点、人效评估、辅助裁员的内部系统。这套算法识别低效岗位、输出优化名单，运行一段时间之后，算法给出的名单里，出现了它的缔造者。他被优化离开，脱下工服去跑即时配送，继续被平台另一套算法管控、考核、催单。
这刚好对应职场里流传的 “空城计” 解读：司马懿早就看穿空城是陷阱，但他不会进城诛杀诸葛亮。对手消失，司马家族制衡曹家的价值也就消失。 放到企业组织里翻译就是：难题存在，对应的预算、编制、岗位才有存续依据。难题被一次性彻底根治，依附这个难题存在的团队，就会立刻变成报表上需要清除的冗余成本。
不是员工能力过剩有罪，是这套短期财报导向的考核，天然惩罚 “一次性根治问题” 的人。
四、AI，正在成为 “清人” 最好的说辞
这一轮裁员潮和前几轮有本质区别。早年裁员，理由是业务收缩、赛道调整、融资遇冷；现在，很多管理层的统一话术是：AI 可以承接工作，岗位不再必要。
AI 的出现，给 “降本增效” 提供了最时髦的叙事外衣。 上市公司管理层面临季度财报压力，人力是最好调节的可变成本。砍掉正式团队，把工作交给大模型 API、外包人力，当期人效指标立刻变好看，能够向华尔街交出满意答卷。而团队长期沉淀的隐性资产：跨业务的信任关系、踩坑积累的边界经验、复杂异常场景的风险兜底能力，这些东西无法录入利润表，自然不在短期决策的考量范围之内。
这里必须区分两种完全不同的 AI 落地思路： 第一种是人机协同：AI 承接标准化、重复、低判断的流水线任务，释放员工精力去做方案博弈、客户沟通、风险决策、创新研发。桑坦德、诺华的落地，走的就是这条路线，AI 是员工的放大器。 第二种是替代幻想：CEO 预设 AI 可以直接顶替岗位，把 AI 当成替代员工的工具，先裁员，再寄希望模型承接原有业务。大量翻车案例都出自这一类。
不少企业踩进同一个坑：仓促裁掉业务骨干之后才发现，模型只能处理标准输入。一旦遇到异常诉求、多方扯皮、业务特殊约定、历史遗留的灰色流程，AI 全部卡壳。 业务不能停，解决方案就是悄悄采购外包、兼职、影子人力补缺口。最后核算总成本，并没有显著下降，只是把稳定的全职用工，替换成了抗风险能力极弱的灵活人力池。组织的容错垫，被抽走了。
很多 AI 产品的需求原点，来自 CEO 圈层自身的生活痛点 —— 预订航班、安排晚宴、代办琐事、规划家庭日程。玛丽莎・梅耶尔畅想夫妻双方的 AI 互相协同打理家事，奥尔特曼公开讨论 AI 辅助育儿，还有人畅想 AI 代替人类完成约会筛选。 这些是顶层少数人的小众痛点，并不是绝大多数业务岗位的真实场景。高管们沉迷 “一切交互都可以被优化、被自动化” 的乌托邦，低估了真实商业里，人作为沟通枢纽、风险判断主体的不可替代性。
五、当认真干活变成一件高风险的事
比单次裁员更致命的伤害，是组织文化的不可逆转向。 当 “彻底解决问题” 的人，最终被优化；当交付完美成果的回报是团队解散，组织内部会自发长出一套自保生存法则。它不是某一个员工的选择，而是系统奖惩机制倒逼出来的集体适应。
大家开始学会拆分成果：一个可以一次性闭环的项目，拆成调研、试点、迭代、二期优化，分阶段交付、分阶段汇报。项目永远处于 “进行中”，而不是 “已完成”。 文档只写到 “可以交接基础工作”，核心取舍逻辑、隐性坑点、边界判断，不会一次性归档交付。跨部门协作尽量把自己做成流程节点，而不是孤立螺丝钉，让上下游业务链路都和自己产生绑定。 很多老员工不再交底全部经验，新人只能接触表层工作，核心经验传递链条慢慢断裂。短期人力报表好看，长期来看，业务断层、突发故障、返工事故的概率持续走高，未来会以更高的成本反噬公司。
网上流传的职场自保十条，之所以广泛传播，本质就是一封打工人写给管理层的投诉信。 这套自保逻辑反复强调：不删库、不留恶意后门、所有操作合法；节奏可控、证据留痕、储备外部机会、看懂竞业协议、不把公司当家。它不是教唆投机摸鱼，是普通人在 “认真就吃亏” 的环境里，最低限度的自我保护。
我们也能看到很多反面教训。微盟运维人员删库、算法工程师删除训练模型，最终都以刑事判决收尾。毁掉系统不是博弈筹码，是把自己送进司法程序，这条路绝对走不通。 真正的矛盾根源，从来不是员工想自保，而是组织没有设计一套机制，去承接功臣的价值。解决问题的人承担全部风险，消灭问题的成果反而变成裁员依据；口头表彰满天飞，长期激励、岗位保障、价值兜底全部缺位。
六、资本已经清醒，有些 CEO 还在做梦
十月市场的回调，本质是一场预期的校准。 资本见过 AI 带来的真实增量，也看清了纯叙事的泡沫。微软、谷歌、Meta 持续发债、加大资本投入，说明 AI 长期赛道没有证伪；但 CDS 利差上行、资金从 AI 标的向长端国债避险切换，代表市场不再接受 “只要上 AI，就一定增长” 的简单故事。
AI 能做什么，边界其实已经清晰：它处理文本、归纳信息、生成初稿、标准化统计、批量筛查，效率远超人类。但它无法识别复杂利益博弈、无法承担业务责任、无法在模糊信息里做取舍、无法重建一个涣散团队的信任。 管理病、组织病、短期考核畸形、部门墙、权责错位，这些公司底层顽疾，算力和模型都治不好。
一个管理混乱、目标摇摆、部门互相甩锅的公司，不会因为采购一套大模型，突然变得高效有序。它只会把原有矛盾转移：人裁了，业务兜不住，外包补位；外包不稳定，服务质量下滑，客户流失。最后管理层还会得出一个荒谬结论：AI 还不够强，继续加码算力，继续压缩人力。
这就是当下最讽刺的循环。 有些 CEO 沉迷一个极简幻想：依靠 AI，绕开麻烦的组织治理、漫长的团队磨合、复杂的人才培养，甩掉员工，单独依靠算力改变企业，甚至改变世界。 他们低估了两件事：第一，现阶段 AI 离不开人的兜底；第二，企业的根基永远是协作的人，而不是模型参数。
AI 是很好的效率工具，但不是一剂裁员神药。 健康的数字化转型，是让 AI 托举人；而病态的 AI 狂热，是拿 AI 当借口抛弃人。 资本已经开始重新算账，打工人早就看懂了这套剧本。剩下的悬念，留给那些还沉浸幻想的 CEO：当影子人力越来越多、团队人人自保、业务隐性风险持续堆积的时候，靠裁员和 AI 故事撑起来的漂亮财报，要拿什么去接住反噬。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
♪ Man, ♪
The CEOs of Silicon Valley have recently reached a romantic consensus that people are redundant variables as long as they have enough numbers and models.
They stand in financial teleconferences, industry summits, media interviews, and repeat the same narrative over and over again: AI Reshaping Everything, Emancipating Efficiency, the old-life burden of manpower can be removed. From Altman talking about AI involvement in parenting, Meyer thinking about the couple AI working together to arrange family matters, Zuckerberg decided that the public would gradually accept AI deep involvement in life, and that the tone of the speech was very uniform. Technology itself can solve all organizational, process, cost dilemmas, and human beings simply wait to be optimized.
The capital market, however, used its October performance to throw a cold water bowl.
In October, the market went down to AI. Wen.
In the past October, there has been an anomaly in the macro-market, which, according to Jeffrey Rosenberg, a senior portfolio manager in Beled, suddenly broke up with the strong correlation between crude oil prices and the US debt rate of 80 per cent over a year.
Prior to that, the logic of the transaction was simple: the price of oil pushed up the price of inflation and the Fed maintained a high rate of return on US debt, and geo-related conflicts were the main pricing line. But the October wind has changed. Rosenberg gives a good judgment: in the eyes of the market, geopolitical risks have given way to AI risks.
Mag 7 Collective echoes, evaporating the market value of Google for a single week to nearly 500 billion, became a landmark event for this round of expected reversals. Capital has defaulted that the energy supply is growing as long as it continues to be expensive to buy the numbers and piles of LLM; in three quarters of the US GDP, investment in information processing equipment surged by 37 per cent, and a large amount of capital spending went up against the story of AI.
But after the financial crisis started, the market started to price AI. A number of companies have handed over verifiable gains: The first three quarters of the Bank of Santander AI project landed, creating €194 million in business value; Novartis and Slags Bank continued to indicate that AI had infiltrated R & D, marketing, customer operations and boosted its leverage; and Meta ' s increase in advertising in AI and doubling of Google cloud revenues were solid positive cases.
And the other part, only narratives. The capital expenditures of super-large cloud manufacturers have already won their own free cash flow. Microsoft issued $25 billion in bonds plus AI input in the market almost 10 years later. The bond market is willing to buy and subscribe to excess in the short term, but the CDS credit spreads continue to rise, and investors are beginning to hedge against a reality that estimates the financing needs of nearly $4 trillion AI over the next four years, and bond markets are not unlimited in carrying capacity.
Capital is no longer brainless believe in "AI = sustainable growth." It slowly understands one thing: AI is in two categories, one embedded in business, and the other is the packaging paper that management uses to downplay, cut, and beautify the quarterly paper. The latter is the “AI risk” that the market is pricing.
♪ Under the cover of AI Agent, hidden in shadow manpower ♪
In the major publish meetings and brochures, AI agents have been able to do everything: call and negotiate bills, book the door service, purchase the price, arrange the home. Open AI, Meta enters the Bureau's P.I.A. track, depicting a machine that does its own work and liberates the future of humanity.
The publicity shell can be removed, and within the industry is another set of operating rules. Dazzle AI, an AI agent founded by Marissa Meyer, has no privacy in this model: LLM is responsible for retrieval of information, giving a list of candidates, and once the mission enters the reality, it is easy to make common sense mistakes by making phone calls, making appointments for construction, confirming service details. They encountered typical failures: AI agents mixed windows and table glass in the roof renovation project, trying to put desktop glass on the window. When encountered with such a card point, the system would not be recycled token, but would be directly transferred to the contracted contractor, who would take over the communication.
Duck Bill, Wajo AI, a start-up company like this, simply makes a standard API product. They have a series of background-checked contract workers who specialize in an AI-unabled link: telephone communication, an offline appointment, and web site human accreditation. The founder, Megan Joyce, put it bluntly, and this logic and the early Uber model is very similar: AI is responsible for billing displays and manpower for bottom delivery.
The Muse agent in Meta has also blown up internal testing disputes: outreach AI calls, and a large number of calls during the internal testing phase have been made from the call centre. After that, Meta explained that this was only an early product test, but that detail had broken the industry consensus: at this stage, the vast majority of AIs that can be “deposit of reality” are essentially a front-end cross-section, with actual contractors being the bottom seat.
Here is the black sense of humor: many CEOs report to the board of directors with cases of "AI autonomous representation" to argue that jobs can be replaced, teams optimized, but they deliberately take advantage of changes in the cost structure - formal employees are laid off, not lost, but full-time jobs that stabilize high-paying jobs, and become off-the-shelf shadows with no social security coverage. The human costs have not disappeared, but have been diverted, hidden and transformed from prominent pay in financial statements into piecemeal external service procurement.
The project that cost $200 million to $2 million. Division
Over the past two days, an amphibious map has been circulating in the veins and in various occupational communities, prompting hundreds of thousands of battered workers.
The story is widely disseminated: a large Internet factory, a 20-person technical team, maintains a high-consumption system, and servers burn 200 million a year, a state that lasts for five years. One of the engineers in the group, with a problem-solving mentality, has been working over several months to re-engineer the bottom structure and, after going online, the annual cost has been reduced directly to 2 million. He would have expected prizes, promotions, awards for the project and, finally, a whole team of layoffs, whose names were also on the list.
The web-based story does not have the main body of the hammer company, so, strictly speaking, it is part of the world of work that is transmitted by the community. But it triggers a resonance, not at the core of the numbers, but at the same time at the same time at the same time that countless practitioners have seen the concoction: workers are paid for, and employers are heads.
There are real cases of almost identical kernels abroad. A United States electrician has developed, using amateur time outside his functions, self-studying automated scripts to save thousands of hours of manual work per year. Instead of waiting for salary increases and performance awards, he was asked by management for the full code permission. Immediately thereafter, HR interviewed a new person with lower hourly pay; after the company had obtained the code, the developer, permission, was promptly cancelled, the labour contract was unilaterally terminated and the separation indemnity was paid for only two weeks, with stringent competitive agreements.
There is also a chain of sarcasm that has been repeatedly raised: there are algorithmic engineers who have built the enterprise's own internal systems for personnel counts, human impact assessments, and auxiliary layoffs. The algorithm identifies the inefficient jobs and the output optimization list, which, after some time in operation, appears in the list given by the algorithm. He was optimized to leave, take off his uniform and run for immediate distribution, and continue to be subjected to another set of algorithms, tests and reminders from the platform.
This corresponds to the "empty city plan" that has been circulated in the workplace: Sema saw the empty city as a trap, but he would not enter the city to kill Zhugsang. When the rivals disappear, the Sema family's value as a counterweight disappears. Translation in business organizations is that there are problems, and the budget, the staffing, and the job is on a continuing basis. The difficulty is thoroughly addressed once and for all, and the team attached to it becomes immediately redundant costs to be removed from the statement.
It's not overcapacity, it's the test of short-term financial reporting, and it's natural to punish people for "one-off problems."
4-Ai, is becoming the best word for the clean man.
There is a difference in nature between this round and the previous rounds. In the early years, layoffs were justified by the contraction of operations, the re-engineering of tracks and the freezing of financing; now, many managements have a uniform narrative: AI can take over work and the job is no longer necessary.
The emergence of AI has provided the most fashionable narrative for “reducing efficiency gains”. The management of listed companies faces quarterly financial reporting pressures, and manpower is the best adjustable variable. By cutting off the official team and handing over the work to LLM API and outsourcing manpower, the current human impact indicators became immediately better-looking and were able to deliver satisfactory responses to Wall Street. The group's long-term sanctuaries are hidden assets: cross-practice trust relationships, border experience accumulated from pedestals, risk-diplining capabilities from complex anomalies that cannot be recorded in profit statements and are naturally beyond the consideration of short-term decision-making.
Here we have to distinguish between two completely different AI scenarios: AI, which takes on standardized, repetitive, low-judgement waterline tasks, unleashing staff energy for programming, customer communication, risk decision-making, and innovative research and development. Santander, Nova, that's the way. AI is the employee's amplifier. The second is alternative fantasy: the CEO preset AI can directly replace the post, use AI as an alternative to the employee, lay off the staff, and send hope models to take over the business. A large number of overturns have come from this category.
A number of enterprises stepped into the same pit: they were quick to cut down the business backbone before finding out that the model could only process standard input. Once an abnormal claim, multiple ripples, special engagements, and historical grey processes are encountered, AI's whole shell. Business cannot be stopped; the solution is quiet procurement outsourcing, part-time work, and shadow manpower gaps. The total cost of final accounting has not decreased significantly, but has replaced the stable full-time workforce with a flexible pool that is extremely vulnerable to risk. The organization's wrong cushion, it's been taken away.
Many of the AI products are demand-driven and come from the CEO community ' s own life pains - booking flights, arranging dinners, doing errands, planning family agendas. Marisa Mayer wants both spouses AI to work together on family matters, Altman to talk openly about AI Auxiliary Childcare, and people to think AI to do dating screening for humans. These are the small pains of the top minority and are not the real scene of the vast majority of operational positions. The utopia, where executives are obsessed with “all interactions can be optimized and automated”, underestimates the real business, where people are irreplaceable as hubs of communication, subject of risk judgement.
When working hard, it becomes a high-risk thing.
A more lethal injury than a single reduction of staff is the irreversible direction of the organizational culture. When people “finished the problem”, they were eventually optimized; when the payoff for delivering the perfect result was the dissolution of the team, a self-sustaining survival code emerged spontaneously within the organization. It is not an employee ' s choice, but a collective adaptation driven by systematic reward and punishment mechanisms.
You are beginning to learn from the results of split: a one-time closed-ring project that breaks down into research, pilot, iterative, two-stage optimization, phased delivery and reporting. The project is always “ongoing” rather than “completed”. The document only reads " May hand over basic work ", core trade-offs logic, hidden pits, boundary judgements, and no one-time archiving delivery. Cross-sectoral collaboration tries to make itself a process node rather than an isolated screw that binds up and down the business chain to itself. Many older employees no longer share the full range of experience, new recruits only have access to surface work, and the core chain of experience is slowly broken. Short-term manpower statements are good and, in the long run, business faults, sudden failures, back-to-work accidents continue to rise, and the future will be counterproductive at higher cost.
The 10 rules of the online workplace are widely disseminated by means of a letter of complaint addressed to management by battering workers. This self-preservation logic has repeatedly emphasized: the absence of red tape, the absence of malicious backdoors, the legitimacy of all operations; the controllability of rhythms, evidence marks, the availability of external reserves, the understanding of competitive agreements, and the lack of corporate ownership. It's not an abettement of speculation, it's a minimum of self-protection for ordinary people in an environment of “seriously losing”.
We can also see many negative lessons. Micro-Alliance relief staff detachment and algorithm engineers remove training models, which eventually end with criminal sentences. Destroying the system is not a game chip, it is sending itself to judicial proceedings, and this road will never work. The real source of conflict was never the desire of the staff to protect themselves, but rather the failure of the organization to devise a mechanism to take on the value of its achievements. Those who solve the problem take full risks, and the results of the elimination become the basis for lay-offs; verbal recognition is full, with long-term incentives, job security and a total absence of value.
VI. Capital is sober. Some CEOs are still dreaming.
In October, the market turned back was essentially an expected calibration. Capital has seen the real increase brought by AI, and has seen the pure narrative bubble. Microsoft, Google, Meta's continued debt and capital investment suggest that AI's long-term track is not perjured; however, the CDS spreads, the funds swapped from AI to long-end national debt, represent a simple story that markets no longer accept “a certain increase as long as it goes on”.
What AI can do is clear: it processes text, summarizes information, generates first drafts, standardizes statistics, screens in bulk, and is far more efficient than humans. It is, however, unable to identify complex interests, assume operational responsibilities, make trade-offs in vague information and rebuild the trust of a disintegrated team. The management of diseases, tissue diseases, short-term malformations, sectoral walls, miscalculation of authority and responsibility are poorly calculated and modelled.
A company that is in a state of mismanagement, with a wave of targets, and with a scrambling of sectors will not suddenly become efficient and orderly by procuring a set of LLMs. It would only shift the existing contradictions: downsizing, unwieldy operations and outsourcing; instability in outsourcing, declining quality of services and loss of clients. Finally, management would have reached the absurd conclusion that AI was not strong enough and would continue to add weight and reduce manpower.
This is the most ironic cycle of the moment. Some CEOs are fascinated by a very simple fantasy: relying on AI, bypassing troublesome organizational governance, long team grinding, complex talent development, dislodging staff, relying alone on arithmetic to change business, or even the world. They underestimated two things: first, at this stage, AI can't get out of the way; second, the foundation of an enterprise is always the person who works together, not the model parameters.
AI is a good efficiency tool, but it's not a dose of cutting. A healthy digital transition is about giving AI credit; a pathological AI fanaticism is about taking AI as an excuse to abandon people. Capital had begun to recalculate, and workers had understood the script. The rest of the suspense is left to the fantasies of CEOs: when shadows are growing, teams are protecting themselves, and business risks continue to accumulate, what are the good financials to feed on layoffs and AI stories?
Auto-fetched and translated by AI Atlas. All rights belong to the original author.