發生了什麼事
Revelio Labs analyzed companies that publicly attributed layoffs to AI, comparing their workforce changes with matched industry peers. It reports that the companies generally expanded AI-related roles while reducing non-AI headcount, but that nearly half did not show a consistent shift toward AI.
Revelio Labs reports that it partnered with System2 to identify companies that publicly attributed layoffs to artificial intelligence in news coverage and then examined their workforce trends using Revelio Labs data. The cohort included technology companies such as Meta and Amazon, as well as companies including General Motors, The Washington Post, UPS, and McKinsey. The source does not provide the cohort’s exact size, selection criteria in full, or a complete list of the companies analyzed, which limits how precisely the findings can be evaluated.
In the two years before the announcements, the companies in the cohort grew AI-related roles at a median rate of more than 11%, according to Revelio Labs, while matched industry peers grew those roles by more than 13%. At the same time, the companies that later announced AI-related layoffs reduced non-AI headcount by more than 3%, while their peers kept non-AI headcount roughly flat. Revelio Labs interprets that combination as evidence of a workforce shift toward AI-related work, even though the companies were not expanding AI roles as quickly as comparable firms.
Revelio Labs also reports that AI-related roles represented nearly twice the share of total employment at the companies in its cohort compared with matched peers, both before the layoff announcements and six months afterward. In the preceding 24 months, the areas with the strongest relative hiring included electrical and information-technology infrastructure, product strategy, research and development, and cybersecurity. The source says non-AI employment shifted toward large-scale operations, physical goods transportation, and legal work—areas where Revelio Labs says artificial intelligence remains less reliable. At the company level, Revelio Labs identifies four patterns: firms shifting toward AI, firms expanding their overall workforces, firms shrinking their workforces, and firms shifting away from AI.
為什麼這很重要
The analysis complicates claims that AI alone drove recent layoffs. It suggests that some companies were reorganizing around AI, while others may have used AI as a rationale for cost cutting, reversing overhiring, or other workforce reductions.
The findings challenge two overly simple accounts of AI-related layoffs. One account treats every announcement as evidence that automation has already replaced large numbers of workers. The other treats every AI explanation as public-relations language for ordinary cost cutting. Revelio Labs reports evidence for both realities: more than half of the firms it studied showed workforce changes consistent with a move toward AI, while a substantial minority reduced their AI workforces or adopted AI more slowly than their industries.
The company-level comparisons are important because aggregate hiring can conceal different strategies. Revelio Labs says Atlassian, Meta, and Coinbase were expanding their overall workforces before announcing layoffs, meaning overhiring may have contributed alongside any AI-related reorganization. By contrast, Block and McKinsey had already reduced headcount during the preceding two years, which the source says is harder to reconcile with a story centered on AI-driven expansion. Revelio Labs also identifies Cloudflare as appearing to shift away from AI before its announcement, while describing the Washington Post, Dropbox, and UPS as examples of companies that did show a clearer movement toward AI-related work.
For workers and the public, the distinction affects how layoffs should be understood and evaluated. A company can cut non-AI roles while hiring specialists who build infrastructure, develop products, or secure AI systems; that is materially different from replacing a specific occupation with a tested automated system. The analysis suggests that investors, employees, and policymakers should ask what work changed, which roles were added, and whether the company’s AI adoption outpaced peers. It does not establish that the remaining workers became more productive, that displaced tasks were fully automated, or that the reported workforce patterns improved products or services.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
What should a useful AI forecast state?
接下來看什麼
The analysis does not directly measure productivity gains or prove that AI caused any particular layoff. Further scrutiny should focus on company-level evidence, the duties eliminated, AI-related hiring after layoffs, and whether stated automation plans produce measurable operational changes.
The most important limitation is causal. Revelio Labs explicitly says it cannot directly observe the internal productivity gains that would have enabled the layoffs. Its workforce measures therefore provide a signal of organizational direction, not proof that AI caused a particular reduction. Companies can hire AI specialists for defensive or experimental reasons, classify roles differently, or cite AI alongside broader financial and strategic pressures. The source also does not independently verify each company’s public explanation or provide the underlying records for every employment change.
Future reporting should examine company-level evidence after the announcements. Relevant indicators include whether firms continue hiring AI engineers and infrastructure specialists, whether they restore or replace eliminated functions, and whether their job postings show sustained adoption of AI tools rather than a temporary recruiting push. It would also be useful to compare announced automation plans with operational results, such as changes in service capacity, product output, error rates, delivery times, or customer support. Those measures could help distinguish genuine process redesign from language aimed primarily at investors or employees.
Revelio Labs bases its AI adoption rate on the percentage of job postings for roles that can build AI systems or leverage AI tools. That is a practical indicator, but it does not show how extensively workers use those tools, how capable the tools are, or whether they replace human labor. Readers should therefore watch for independent research using payroll, task-level, productivity, and employee-outcome data. The source leaves several meaningful unknowns, including the cohort’s exact size, the companies’ geographic coverage, the rules for AI and non-AI roles, and whether the six-month follow-up period captures longer-term workforce effects.