文 | wiwi
去年，Shopify CEO Tobi Lütke 给公司定了一条规矩：团队想增加人手，得先说明为什么不能用 AI 完成相关工作。
这条要求当时引发了不少讨论。有人觉得这是科技公司拥抱 AI 的正常操作，也有人担心，它会成为企业减少招聘的理由。但比起那些关于 AI 取代人类的预测，这项具体的管理要求反而更值得琢磨。它没有说员工不重要，只是把招聘的顺序换了：以前业务忙不过来，可以申请增加人手；现在，要先看看 AI 能不能帮忙解决。
一年多过去，越来越多办公 AI 产品开始朝这个方向发展。从腾讯 WorkBuddy、阿里巴巴悟空，到微软 Copilot、Salesforce Agentforce，科技公司不再满足于让 AI 写文档、做表格、回答问题，而是希望它能操作软件、处理业务，直接完成工作。产品发布时，大家关心的是 AI 又学会了什么本领，但站在企业管理者的角度，另一个问题可能更有吸引力：既然这些事可以交给 AI，原来计划增加的人手，还有必要吗？
这让眼下的办公 AI 热潮显得有些微妙。员工忙着学习 AI，希望自己能变得更有竞争力；科技公司却在向企业证明，过去需要员工完成的事情，如今可以交给软件。大家都在为生产力进步感到兴奋，但对这场变化的期待，并不完全相同。
而且，这次科技公司盯上的可能不只是办公软件市场。它们开始触及企业一笔更大的支出：原本花在员工、外包和业务执行上的钱。
软件公司，开始卖工作了
过去企业买办公软件，大多按账号、人数或功能付费。员工越多，通常需要的软件账号也越多。企业招人、员工干活、软件提供工具，这套模式已经运行了很多年。
AI Agent 让这门生意出现了新的可能。
腾讯 WorkBuddy 企业旗舰版目前仍采用每人每月 198 元的席位收费方式，但 Salesforce 的 Agentforce 已经引入了按动作计费的机制，标准智能体操作收费 0.10 美元。Intercom 的 AI 客服 Fin 则采用按结果计费的方式，部分符合条件的服务结果收费 0.99 美元。
这几款产品的用途和计费口径不同，不能简单比较谁更便宜。不过，收费方式的变化很有意思。软件厂商不再只关心有多少员工在使用产品，还开始关心自己的系统替客户完成了多少任务。
拿客服来说，以前一家企业咨询量增加，往往需要增加客服人员，再购买更多软件账号。现在，部分标准化咨询可以由 AI 直接处理，企业就能根据 AI 处理的业务量，再决定需要配置多少人工客服。
软件公司由此获得了一种新的销售方式：不只是告诉客户，自己的工具能让员工工作得更快，而是可以直接提供一部分原本需要人工完成的服务。
当然，完成一次客服咨询，不等于承担客服岗位的全部职责。复杂投诉、客户关系、突发情况仍然可能需要人工处理。一套能够自动生成报告的系统，也不意味着企业可以直接取消分析岗位。
但企业已经有机会把某些工作的人工成本，与 AI 服务的价格摆在一起比较。
以前，一家公司增加业务，首先想到的是需要配多少人、采购哪些系统。现在，至少在一部分标准化工作中，管理者可以先看看现有团队加上 AI 能做到什么程度，再考虑有没有必要继续招聘。
对软件公司来说，这比单纯出售办公账号有更大的想象空间。企业每年花在员工薪酬、外包和业务运营上的钱，远远不只是软件采购费。如果 AI 能够稳定承担其中一部分工作，科技公司就有机会从这些原本不属于软件采购的支出中获得收入。
过去软件公司希望企业多买几个账号，现在它们开始有机会告诉企业，有些工作不一定要再增加一个人。
这也是为什么办公 AI 产品越来越强调自动执行、多智能体协作和跨系统操作。用户当然希望软件越来越好用，但对企业来说，能够减少多少人工干预，才更容易换算成一笔具体的经济账。
AI 还没完全干好，企业已经开始考虑少招人了
真正让人感到不安的，未必是 AI 已经能够替代多少工作，而是企业可能提前按照它未来的能力做决定。
麦肯锡 2026 年 8 月发布的全球 AI 调查中，有两组数据值得放在一起看：39% 的受访者预计，未来一年 AI 会让所在企业的整体员工数量下降；但回顾过去一年，只有 14% 的受访者表示，AI 已经促成所在企业整体人员规模减少。
更有意思的是，在上一年的调查里，32% 的受访者曾预计接下来一年会出现人员减少，实际报告发生这一变化的比例却远低于此前预期。
这说明，企业对 AI 减少人员需求的想象，至少目前还跑在实际结果前面。
对不少企业而言，这并不奇怪。AI 产品更新太快了，几个月前还需要人工反复操作的任务，现在可能已经能自动完成一部分。管理者很容易产生一种判断：既然 AI 还在快速进步，有些岗位是不是可以先不招，看看半年之后的情况？
从单家公司的经营角度看，暂缓招聘可能只是谨慎控制成本。但如果许多企业都这样考虑，劳动市场就会出现一种不容易察觉的变化。
裁员会留下新闻，招聘计划取消却未必有人知道。一个团队原本打算招五个人，最后只招了两个，没有员工失去现有工作，企业甚至可能仍然在增长。可是对那些本来可能获得工作机会的人来说，剩下三个岗位已经不存在了。
当然，现在不能把招聘减少全部归因于 AI。经济环境、行业竞争、企业利润都会影响用工需求。只是 AI 的出现，确实给企业增加了一个新的选项：暂时不增加人手，等待技术进一步成熟。
这和过去的软件升级不太一样。传统软件更多是帮助现有员工提高效率，而当 AI 开始承担实际任务，企业就有机会在业务尚未扩张之前，重新估算需要多少员工。
Shopify 的规定之所以有代表性，就是因为它把这个思路直接放进了招聘流程。
但这里还有一个容易忽略的问题：AI 的能力需要经过实际业务验证，招聘计划却可以先调整。企业可能还没证明智能体能稳定承担某项工作，就已经开始按照未来可能达到的效率安排人员。
如果 AI 最终做到了，企业自然可以获得收益；如果没做到，剩余员工可能需要承担更多工作，而原本能够培养新人的岗位也已经被压缩。
AI 还没有真正替代一个人，但企业已经可能因为相信它将来能做到，而决定暂时不招这个人。
相比讨论某款产品到底有多聪明，这种企业决策上的变化，可能更早影响普通人的工作机会。
少招几个人，生意就一定更好吗？
现在不少办公 AI 产品喜欢用一个很直观的方式展示价值：原本需要几个人做的工作，现在一个人加上 AI 就能完成。
这样的演示确实有吸引力，尤其对那些人力成本较高、业务流程重复的企业来说。但做过企业项目的人应该清楚，能把任务跑起来，与能够长期稳定交付，中间还有不小的距离。
比如让 AI 生成一份报告，可能只需要几分钟。可如果其中引用的数据有误、分析偏离客户需求，最后仍然需要有人检查、修改，再与客户沟通。有时候生成内容只花了十分钟，后续确认和返工却耗费了几个小时。
代码开发也是如此。AI 可以很快写出功能，但测试、部署、异常处理和系统维护，并不会因为代码生成得快就全部消失。对于需要承担真实业务责任的企业来说，这些环节往往比生成结果本身更加重要。
所以，AI 能够节省多少操作时间，与企业最终能够节省多少钱，不能画等号。
2026 年 5 月，Gartner 发布的一项调查发现，在已经试点或部署自主化技术的受访大型企业中，约 80% 报告过人员减少，但研究没有发现，人减得更多的企业就一定获得了更好的投资回报。
这项调查针对的是年营收至少 10 亿美元、已经试点或部署相关技术的企业，并不代表所有企业的情况。但它至少提醒管理者，减少人手是一回事，真正提高经营收益又是另一回事。
麦肯锡的调查也有类似的发现：80% 的受访者认为 AI 提高了个人工作效率，但只有 37% 表示 AI 对所在企业的息税前利润产生了正向影响。
为什么员工觉得工作变快了，企业却不一定赚到更多钱？
原因并不复杂。一名员工用 AI 把原本两个小时的工作压缩到十分钟，如果结果还需要其他部门审批三天，整个流程的效率未必提高多少。企业为了部署 AI，可能还要支付系统接入、模型调用、维护、安全审计和人工复核等费用。
这些成本如果没有算进去，就很容易高估 AI 带来的收益。
更麻烦的是，有些岗位的价值根本没办法按照任务数量来计算。一个经验丰富的员工，也许每天真正动手操作的时间并不多，但他知道客户为什么不满意、项目出了问题该找谁、哪些需求看起来简单却不能随便答应。这些能力平时不容易出现在绩效报表里，出现问题的时候却很值钱。
如果企业只把员工的工作拆成一项项可以自动化的任务，再根据这些任务占用的时间推算可以减少多少人，最终算出来的未必是真实的人力需求。
对 AI 厂商来说，这同样不是一个简单的问题。把智能体演示得很聪明是一回事，让企业愿意长期付费又是另一回事。前者只需要展示它能完成什么，后者必须回答它完成得有多稳定、出了问题谁负责，以及算上全部成本之后到底值不值。
当企业越来越认真地算这笔账，办公 AI 的竞争也会变得更加现实。产品能处理多少任务当然重要，但能否降低一次合格业务交付的总成本，才是企业最终愿意持续采购的理由。
企业越来越高效，人怎么办？
说到底，企业采用 AI 并没有什么问题。重复劳动能够交给机器，许多原本需要专业人员才能完成的工作，现在普通人也能尝试，这些都是实实在在的进步。
一个独立开发者可以借助 AI 做出过去需要团队才能完成的产品，一家小企业也有机会用更低的成本获得设计、分析和研发能力。过去被资金、技术和人员规模限制住的事情，现在有了新的可能。
只是技术降低了工作难度，也可能让一部分工作变得不再稀缺。
以前，企业愿意为某项专业技能支付较高的报酬，是因为掌握它的人不多，培养也需要时间。如今，当越来越多的人可以借助 AI 完成相关任务，企业就可能重新考虑这些工作的价格，甚至重新设计岗位。
对于员工来说，这种变化多少有些矛盾。
大家努力学习 AI，本来是希望自己工作更轻松，也更有竞争力。可一旦所有人都学会了使用 AI，原本的效率优势很可能变成新的工作要求。以前一天完成五项任务算不错，现在借助 AI 完成十项任务，可能才符合团队新的绩效标准。
员工能力确实提高了，但工作负担未必因此减少。
企业也会面临另一种矛盾。既然 AI 能完成越来越多的基础工作，为什么还要花钱培养新人？如果资深员工配合智能体就能完成项目，招聘初级员工似乎没有那么迫切了。
短期看，这样做确实可能节省成本。可一个行业不可能永远依靠现在这批资深员工。经验不是看几篇资料、运行几个 Agent 就能积累出来的。很多专业判断，恰恰来自年轻时做过的那些琐碎工作，以及在一次次失败中积累的教训。
当企业减少基础岗位，却仍然希望招聘有多年经验、能够独立处理复杂问题的人，人才培养的成本就容易被推给其他公司，甚至整个社会。
一家企业少招几个新人，可能不会马上出问题；但如果越来越多企业都不愿意培养新人，几年之后，行业需要的资深人才又从哪里来？
这笔账不会出现在 AI 产品的报价单里，也很难从一家公司的季度财报中看到，却可能影响企业未来的竞争力。
不过，AI 对就业的影响也不只有这一面。企业可以用它减少成本，也可以用它开展过去无力经营的新业务。更低的开发成本、更快的产品交付、更广的服务覆盖，都有机会带来新的需求和工作机会。
问题是，减少成本往往比创造新业务更容易计算。少招两个人，财务部门可以直接看到预算变化；利用 AI 开发一个新市场，却需要承担试错风险，也无法保证马上带来收入。
因此，在经营压力之下，企业很容易先选择前者。
但如果所有公司都只是想着用 AI 把原来的工作交给更少的人，技术带来的增长空间反而可能被浪费。对一家企业来说，真正有价值的 AI 转型，不应该只有人员成本下降，还应该包括产品能否做得更好、客户能否获得更多服务，以及企业有没有能力去做过去做不了的生意。
这也是为什么我觉得，眼下的办公 AI 热潮值得高兴，却还不到只顾着庆祝的时候。
科技公司找到了新的商业机会，企业有了更灵活的生产方式，普通人也获得了更强的工具。但这些好处最终如何分配，并不会随着模型能力升级而自动得到解决。
过去，企业扩大业务，通常需要增加人手。一个行业越繁荣，往往也意味着更多的人能够进入这个行业工作、学习和积累经验。
现在，随着 AI 开始承担具体任务，这种关系可能逐渐松动。企业可以在不增加太多员工的情况下扩大产出，软件公司也有机会从原本属于人工服务的工作中获得收入。
这未必是坏事。人类本来也不应该永远被重复劳动困住。但如果一家企业通过 AI 创造了更多价值，员工却只感受到更高的绩效要求、减少的招聘机会和越来越激烈的竞争，那么技术进步带来的收益，显然还有很大一部分没有转化为普通人的生活改善。
办公 AI 大战最后会诞生多少新产品、多少新的软件公司，现在还不好说。但有一点已经值得企业认真考虑：当机器可以承担更多工作，节省下来的人力究竟应该变成财务报表上的利润，还是成为开拓新业务、培养新人和改善工作条件的机会？
过去，企业需要证明买一款软件是值得的。现在，一些企业开始要求管理者证明，为什么不能先使用 AI，再考虑增加员工。
当软件越来越容易证明自己值得被购买，人却开始需要证明自己值得被雇用，这场办公 AI 大战，恐怕就不只是科技公司之间的生意了。
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, wwi.
Last year, Shopify CEO Tobi Lütke set a rule for the company: the team needs to know why it can't do the job with AI.
That requirement was the subject of much discussion. Some felt that it was the normal operation of technology companies to embrace AI, and others feared that it would be a reason for enterprises to reduce recruitment. But this specific regulatory requirement is more interesting than the predictions about AI replacing humanity. It does not say that the employee is not important, but simply changes the order of recruitment: previously, the business was too busy to apply for additional staff; now, let's see if AI can help.
More than a year ago, more and more of the AI products began to move in this direction. From Tencent WorkBuddy, Alibaba Goku, to Microsoft Copilot, Salesforce Agentforce, the technology company is no longer satisfied with having AI write documents, make tables, answer questions, but rather wants it to be able to operate software, process business and finish its work. When the product publish was produced, the concern was what AI had learned, but another question might be more attractive from the point of view of business managers: is there any need to increase the number of people that had been planned since these things were available to AI?
This makes the current office a little subtle in the heat of AI. Employees are busy learning AI, hoping to become more competitive; technology companies are proving to their firms that what they needed to do in the past can now be given to software. While everyone is excited about productivity progress, the expectations for this change are not identical.
Moreover, this time the technology company may be looking at more than just the office software market. They begin to touch upon a larger expense of the enterprise: money that was originally spent on staff, outsourcing and operational execution.
Software company, start selling jobs.
In the past, businesses purchased office software, mostly by account number, number or function paid. The more employees the more software accounts are usually required. The model has been in operation for many years.
AI Agent made this business possible.
While the flagship version of Tencent WorkBuddy is still based on a seat charge of $198 per person per month, the Agentforce of Salesforce has introduced an action-based billing mechanism with a standard AI agent operating fee of $0.10. Intercom ' s AI client service Fin is based on a result-based fee, with a fee of $0.99 for some of the eligible service results.
These products have different uses and costing calibres and cannot simply be compared to who is cheaper. However, changes in the way charges are charged are interesting. Software manufacturers no longer only care about how many employees are using their products, but also begin to care about how much their systems perform on behalf of clients.
In the case of customer service, the increase in the number of consultations in a previous enterprise often required the addition of passenger staff and the purchase of additional software accounts. Now, some of the standardized consultations can be handled directly by AI, so that an enterprise can determine the number of manual passengers that need to be deployed according to the volume of business handled by AI.
Software companies have thus acquired a new way of selling: not only do they tell customers that their own tools enable employees to work faster, but they can directly provide some of the services that would otherwise need to be performed manually.
Of course, the completion of a customer consultation does not amount to full responsibility for the customer service. Complex complaints, customer relationships and contingencies may still require manual processing. A system that automatically generates reports does not mean that an enterprise can directly remove an analytical job.
However, enterprises have the opportunity to compare the labour costs of certain jobs with the prices of AI services.
Previously, a company increased its business by thinking first of the number of people it needed and of the systems it purchased. Now, in at least part of the standardization exercise, managers can look at the extent to which existing teams can be combined with AI before considering whether there is a need to continue recruitment.
For software companies, this has greater imagination than simply selling office accounts. Businesses spend much more on employee remuneration, outsourcing and operations each year than on software purchases. If AI is able to take on some of these tasks in a stable manner, the technology company will have the opportunity to earn income from these expenditures that would not otherwise have been the subject of software procurement.
In the past, software companies had hoped that the company would buy more accounts, and now they had the opportunity to tell the business that some jobs did not have to add another person.
This is why the AI products are increasingly focusing on automated implementation, multi-AI agent collaboration and cross-system operations. Of course, users would like the software to become more and more useful, but it is easier for enterprises to have convert as a specific economic account by reducing the number of manual interventions.
AI, it's not quite done. The company's starting to think about hiring.
What is really disturbing is not necessarily how much work AI has been able to replace, but rather that companies may decide ahead of schedule according to their future capabilities.
In a global AI survey of publish in August 2026, two sets of data were worth looking at: 39 per cent of respondents expected that AI would cause the overall number of employees in their enterprise to decline in the coming year; but only 14 per cent of respondents recalling the past year said that AI had contributed to a reduction in the overall size of the enterprise.
More interestingly, 32 per cent of respondents in the previous year ' s survey had anticipated a decrease in staff in the following year, while the rate of reported change was much lower than previously anticipated.
This suggests that the business imagination of AI to reduce the demand for personnel is, at least now, ahead of actual results.
This is not surprising for many businesses. The AI product has been updated too quickly and will require manual re-engineering several months ago, and may now be able to do some of this automatically. It is easy for managers to make a judgment: since AI is making rapid progress, can some jobs be taken off the books and see what happens after six months?
From the point of view of the operation of a single company, a suspension of recruitment may be simply a prudent cost control. But if many enterprises consider this, there will be an undetectable change in the labour market.
Retrenchment would leave news, and no one would necessarily know about the cancellation of the recruitment programme. A team that had intended to recruit five people ended up with only two, without any employees losing their jobs, and businesses might even be growing. However, for those who might otherwise have gained jobs, the three remaining jobs no longer existed.
Of course, the reduction in recruitment cannot now be attributed entirely to AI. The economic environment, industry competition and business profits all affect demand for jobs. It is just that the emergence of AI does add a new option to the enterprise: to wait for technology to become more mature without additional manpower.
This is not the same as the previous software upgrade. Traditional software is more about helping existing staff to become more efficient, and when AI begins to take on practical tasks, there is an opportunity for enterprises to re-evaluate the number of staff needed before their business expands.
Shopify ' s provisions are representative because they put the idea directly into the recruitment process.
But there is also an easy question to ignore: AI capacity needs to be validated in practice, while recruitment plans can be adjusted first. The enterprise may have begun to organize its personnel according to the efficiency that may be achieved in the future, without proving that AI agent is able to take on a job steadily.
If AI had finally done so, the enterprise would naturally have received benefits; if not, the remaining staff might have to take on more jobs, and jobs that would have produced new recruits had been reduced.
AI has not actually replaced a person, but it is possible that an enterprise may decide not to hire a person for the time being because it believes it can do so in the future.
Such changes in corporate decision-making may affect the job opportunities of ordinary people earlier than discussing how smart a product is.
You think business would be better if there were fewer people?
Now there's a lot of office AI products that like to show value in a very intuitive way: jobs that would have to be done by a few people, and now one person can do it with AI.
Such a demonstration is indeed attractive, especially for enterprises with high human costs and duplicate business processes. But those who have done business projects should be clear that there is a distance between running missions and being able to deliver steadily over time.
For example, let AI produce a report that could take just a few minutes. However, if the data cited are incorrect and the analysis deviates from the customer ' s needs, there is still a need to check, modify and communicate with the client. Sometimes it takes only 10 minutes to generate content and hours to follow up and return to work.
The same applies to code development. AI can write functionality quickly, but testing, deployment, anomaly processing and system maintenance do not disappear because the code is generated quickly. These links tend to be more important to enterprises that have real business responsibilities than to generate the results themselves.
So, how much time can AI save on operations, and how much money can an enterprise eventually save, and not equalize.
In May 2026, a survey by Gartner publish found that about 80 per cent of the large enterprises interviewed that had piloted or deployed autonomous technologies reported a decrease in personnel, but the study did not find that a better return on investment would be achieved if more businesses were lost.
The survey was aimed at enterprises that received at least $1 billion annually, that had piloted or deployed the relevant technology, and did not represent the situation of all enterprises. However, it at least reminds managers that it is one thing to reduce staff and another to actually increase business returns.
A similar finding in McKinsey's survey: 80% of respondents found AI to have improved individual productivity, but only 37% indicated that AI had had a positive impact on pre-tax profits in their businesses.
Why do employees feel that work is getting faster and businesses don't necessarily make more money?
The reasons are not complex. One staff member used AI to compress the original two hours of work to 10 minutes, and if the result required another three days of approval by other departments, the efficiency of the entire process would not necessarily be much improved. In order to deploy AI, enterprises may also have to pay for systems access, model calls, maintenance, security audits and manual reviews.
If these costs are not accounted for, it is easy to overestimate the benefits of AI.
Even more problematic is the fact that the value of some posts cannot be calculated according to the number of tasks. An experienced staff member may not have much time to actually do business on a daily basis, but he knows why clients are not satisfied, who should be asked for project problems and what needs seem simple but cannot be readily accepted. These capabilities are not readily present in performance statements and are valuable when problems arise.
If an enterprise breaks down the work of an employee into a task that can be automated, then the number of people that can be reduced depending on the time taken by these tasks is calculated, which does not necessarily mean real manpower needs.
This is also not a simple issue for AI manufacturers. It's one thing to show that AI agent is smart, and it's another thing to make businesses want paid for the long term. The former simply had to show what it could accomplish, and the latter had to answer how stable it had been, who was responsible for the problems and what was worth the full cost.
The competition for office AI will become even more realistic as firms work more and more seriously. While it is of course important how many tasks a product can handle, whether the total cost of a eligible business delivery can be reduced is the reason why an enterprise is ultimately willing to buy on a continuous basis.
Business is becoming more efficient. What about people?
After all, there's nothing wrong with companies using AI. The fact that duplication of work can be handed over to machines, that many of the work that would otherwise be done by professionals can now be tried by ordinary people is a real advance.
An independent developer can use AI to produce products that used to require a team to complete, and a small enterprise has the opportunity to acquire design, analysis and R & D capabilities at a lower cost. There are new possibilities for what has been constrained by financial, technical and human size.
It is just that technology reduces the difficulty of work and may also make part of it less scarce.
Previously, enterprises were willing to pay higher wages for a particular professional skill because of the small number of people with it and the time required to develop it. Today, when more and more people can use AI to do the job, firms may reconsider the price of these jobs and even redesign jobs.
For staff, this change is somewhat contradictory.
People were trying to learn AI, hoping to work easier and more competitive. But once everyone learns to use AI, the original efficiency advantage is likely to become a new job requirement. Five assignments a day were good before, and 10 missions were now done through AI, possibly meeting the new team performance standards.
It is true that staff competencies have increased, but the workload does not necessarily decrease.
Businesses also face another contradiction. If AI can do more and more basic work, why pay for new talent? If senior staff were to cooperate with AI agent to complete the project, it would seem less urgent to recruit junior staff.
In the short term, this may indeed result in cost savings. But an industry cannot depend forever on this group of senior employees. Experience doesn't come from looking at a few pieces of information, running a few Agents. Many professional judgements come from the petty work done at a young age and from the lessons accumulated in one failure.
When enterprises reduce their basic jobs but still wish to recruit people who have many years of experience and are able to deal with complex problems independently, the cost of talent development can easily be pushed to other companies and even to society at large.
A small number of new recruits in an enterprise may not be immediately problematic; but if more and more firms are reluctant to prepare new recruits, where will the seniors needed by the industry come from a few years later?
The account would not appear in the quotation for the AI product, and it would be difficult to see it from a company ' s quarterly financial statements, but could affect the future competitiveness of the enterprise.
However, AI ' s impact on employment is not unique. It can be used by enterprises to reduce costs or to carry out new business that was previously unaffordable. Lower development costs, faster product delivery and wider coverage of services offer opportunities to generate new demands and jobs.
The problem is that reducing costs is often easier to calculate than creating new business. With two fewer members, the financial sector can see direct budgetary changes; using AI to develop a new market, it has to take the risk of error and does not guarantee immediate revenue.
As a result, it is easy for enterprises to choose the former first under operational pressure.
But if all companies are just trying to use AI to give their original jobs to fewer people, technology can be wasted on the growth space that technology brings. A truly valuable AI transformation for an enterprise should not only be a reduction in human costs, but should also include whether products can be better done, whether clients can access more services, and whether the enterprise has the capacity to do business that it could not do in the past.
And that's why I think that the current office of AI is happy, but not just to celebrate.
Science and technology companies have found new business opportunities, enterprises have become more flexible in their production methods and ordinary people have been given stronger tools. However, how these benefits are ultimately distributed will not be automatically addressed as model capabilities are upgraded.
In the past, the expansion of business usually required additional staff. The more prosperous an industry, the more often it means that more people are able to enter the profession, learn and gain experience.
Now, as AI begins to take on specific tasks, this relationship may gradually loosen. Enterprises can expand their output without adding too many employees, and software companies have the opportunity to earn income from jobs that are otherwise manual.
This is not necessarily bad. Human beings should never have been trapped by repeated labour. But if an enterprise creates more value through AI, and employees feel only higher performance requirements, reduced recruitment opportunities and increased competition, it is clear that much of the benefits of technological progress have not translated into improvements in the lives of ordinary people.
It's hard to tell how many new products, how many new software companies will be born at the end of the war. But there is a point that deserves serious consideration: when machines can take on more work, should the human savings be translated into profits in the financial statements, or should they be an opportunity to start new businesses, train new people and improve working conditions?
In the past, enterprises needed to prove that it was worth buying a software. Now, some enterprises are beginning to require managers to prove why AI cannot be used before additional staff are considered.
When software became easier to prove worth being bought, people began to have to prove worth being hired, and the AI war was probably more than just business between technology companies.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.