Tech companies say AI is taking jobs. The story is more complicated.

Over the past few months, several large techtech: Technology, especially as applied to a particular industry; a technology company or sector — 科技 companies have announced big staff cutsstaff cuts: reductions in the number of workers a company keeps — 裁员. They say the reason is artificialartificial: Made or produced by human beings rather than occurring naturally — 人造的；人工的 intelligence. Atlassian, Block and Amazon have all said they will lay offlay off: to end workers' jobs because the company no longer needs them — 解雇 thousands of workers because they are using AI more.

The message from these companies is simple. AI can now do work that people used to do, so fewer workers are needed. However, the evidence tells a more careful story.

Part of the AI story is true

There are real changes in some parts of the job market, but the size of the change is often exaggerated. Research from Anthropic, published this month, shows that AI can do many work tasks. Yet most of those tasks are still done mainly by people, not by AI tools.

Some jobs face more pressure than others. Computer programmers are at the top of the list, followed by customer service workers and data entry staff. Eveneven: Exactly, just, fully — 正好 in these jobs, however, AI is still used in a limitedlimited: Restricted in size, amount, or extent. — 有限的；受限制的 way.

A 2025 Goldman Sachs report estimatedestimate: To calculate roughly, often from imperfect data — 估计 that if companies used AI for everything it can currently do, about 2.5% of US jobs would be at risk. That is not a small number. However, the report also notes that workers in AI-exposedAI-exposed: (of jobs or industries) likely to be affected by AI tools — 易受AI冲击的 jobs are not yet more likely to lose their jobs or earn less than other workers.

The report does see some early stressstress: Discomfort — 不适，不舒服 in certain industries. Marketing, graphicgraphic: Relating to visual art or design; giving vividly explicit detail. — 图形的；生动的 design, office work and call centres are growing more slowly than before. In the US tech sectorsector: A distinct area or part of something, especially of a country's economy or society — 部门；行业；领域, young workers in their 20s in AI-exposed jobs saw unemployment rise by almost 3% in early 2025. These are real signals, but they are limited to specific sectorssector: A distinct area or part of something, especially of a country's economy or society — 部门；行业；领域. They are not the wide loss of jobs that company announcements often suggest.

So what else might be going on?

The other reasons companies are letting people go

The timingtiming: The choice, judgement, or control of when something should be done. — 时机；时间选择 of these layoffslayoffs: the act of dismissing workers because the company has less work or money — 裁员 is interesting. Many tech firmsfirm: A business or company, especially a small one. — 公司；商行 hired too many people during the post-pandemic boomboom: A period of rapid economic growth or increase in activity; a deep resonant sound. — 繁荣；景气；隆隆声post-pandemic boom: the rise in business activity that came after the COVID-19 pandemic — 疫情后繁荣期, when online services were in high demanddemand: An insistent request; the desire of consumers for a particular product or service. — 需求；要求. Investorsinvestor: A person or organisation that puts money into financial schemes or businesses with the expectation of profit. — 投资者 are also pushing companies to improve profit marginsmargin: The edge or border of something; the amount by which something is won; profit margin. — 边缘；差额；利润率profit margins: the share of money a company keeps as profit after paying its costs — 利润率. These pressures exist alongsidealongside: Next to or together with; at the same time as. — 在…旁边；与…一起 the real progressprogress: To move forward or develop towards a better or more advanced state. — 进展；进步 in AI, but they are rarely mentioned together.

There is also a strong financial reason for companies to look like they are using AI. Since ChatGPT launchedlaunch: To start or set in motion something new; to send a spacecraft into the air. — 发起；推出；发射 in 2022, AI-related stocksstock: A supply of goods kept for sale; shares in a company; an amount available for use. — 库存；股票；储备 have driven about 75% of the gains in the S&P 500. A workforceworkforce: The people engaged in or available for work in a country or area. — 劳动力；全体员工 reductionreduction: The act or process of making something smaller or less in amount, degree, or size. — 减少；降低；缩减workforce reduction: a cut in the number of people a company employs — 裁员 linked to AI sounds better to investors than a simple cost cut.

Meta is a good example. The company is reportedlyreportedly: According to reports or rumors; supposedly — 据报道 planning to lay off as much as 20% of its staff. At the same time, it has promised to spend US$600 billion on data centresdata centres: large buildings full of computer servers, often used to run AI — 数据中心 and top AI researchers. The workers losing their jobs are not really being replaced by AI today. They may be paying for the AI betbet: A wager, an agreement between two parties that a stake (usually money) will be paid by the loser to the winner (the winner being the one who correctly forecast the outcome of an event) — 赌注AI bet: a large, risky investment a company makes, hoping AI will pay off in the future — AI豪赌 that Meta is making on the future.

The bigger picture

A recent PwC report suggests that jobs are still growing in most industries that use AI, just more slowly. Wageswages: [Algebra] Daily or weekly pay — 薪水 (日薪/周薪) are also rising faster in those industries, and workers with AI skills earn about 56% more on average.

The big picturebig picture: the overall situation, when small details are set aside — 总体形势, then, may not be mass displacementdisplacement: The act of displacing, or the state of being displaced; a putting out of place — 移置；流离失所；使脱离原位的行为或状态mass displacement: the loss of jobs across many industries at the same time — 大规模失业. It may be a workplaceworkplace: A place where people work, such as an office or factory. — 工作场所；职场 where AI tools change what people do, rather than replace them completely.