To identify the job roles most susceptible to AI disruption, researchers are integrating large language model (LLM) functionalities with actual usage data, enabling them to monitor “observed exposure.”
The effect of AI on human employment has recently become a prevalent discussion, sparking numerous predictions, reports, and warnings.
Nevertheless, two recent studies suggest that AI has not yet significantly transformed the job market, and conventional assessment methods may not precisely reflect its true influence on labor.
A report on job cut announcements by the employment analytics firm Challenger, Gray & Christmas indicates that AI has been responsible for 12,304 job displacements in 2026 to date, accounting for only 8% of total job reductions.
Furthermore, Anthropic researchers have unveiled an innovative method for evaluating AI’s influence on employment, contending that a substantial disparity persists between the potential of large language models (LLMs) and their practical application.
Through this analytical framework, they concluded that “AI is still far from achieving its full theoretical potential, with its actual implementation covering only a fraction of what is possible.”
AI’s Role in Job Reductions Remains Minor
Since 2023, Challenger, Gray & Christmas has monitored AI’s influence on employment. To date, AI has been mentioned in 91,753 job cut reports, accounting for approximately 3% of all proposed layoffs. In 2025, AI was the cited cause for 54,836 layoffs, making up 5% of the year’s total job reductions.
In February, AI was specifically attributed to 4,680 job eliminations, which constituted approximately 10% of all cuts during that month.
Notably, the technology sector has experienced significant shifts, with some firms reducing human staff to adopt AI. For instance, Block, a payments and financial services firm, recently decreased its workforce by 50% as CEO Jack Dorsey advocated for an “intelligence-native” operational framework.
Alarmingly, the tech industry reported 11,039 job cuts in February. By 2026, major tech companies have shed 33,330 positions, marking over a 50% increase in sector job losses compared to the previous year’s corresponding period. Nevertheless, these reductions are not exclusively attributable to AI.
“The tech industry is currently navigating various challenges,” Challenger observes. “Although AI garners significant attention,” other factors contributing to the situation include international regulations, a decline in digital advertising linked to tariffs and economic instability, and elevated operational expenses.
Computer Programmers and Customer Service Representatives Face Highest Risk
Concurrently, Anthropic is assessing the risk of AI-driven job displacement using an innovative “observed exposure” technique, acknowledging that previous methodologies warrant “humility.” This approach integrates the theoretical potential of large language models (LLMs) with actual usage statistics, prioritizing automated and work-specific applications over what the company categorizes as “augmentative” uses.
According to this framework, no “consistent rise” in joblessness among highly-exposed employees has occurred since late 2022, as reported by Anthropic researchers Maxim Massenkoff and Peter McCrory. They do, however, observe “indications” of a decelerated hiring rate for younger individuals in certain job sectors.
Anthropic’s methodology integrates information from the O*NET database, which links tasks to numerous US job roles, alongside task-specific exposure evaluations determining if LLMs can complete a task at least twice as quickly as a human, and usage statistics from the Anthropic Economic Index.
The study examined work-specific interactions within Claude’s data streams, differentiating between tasks executed entirely by AI autonomously and those that supplemented human workflows. A job was deemed more exposed if its related tasks exhibited “substantial utilization” within the Anthropic Economic Index, were performed in professional settings, and constituted a considerable portion of the overall job function. The frequency of these tasks appearing in automated usage patterns or API deployments for a particular job was also considered.
Utilizing this methodology, the occupations most susceptible to AI-driven displacement encompass:
- Computer programmers, who can use AI to do 75% of their jobs (this is in line with other data showing that Claude is being used “extensively” for coding).
- Customer service representatives, who can use AI for 70% of their work (the researchers point out that their main tasks are increasingly seen in first-party API traffic).
- Data entry keyers (67%)
- Market research analysts and marketing specialists (65%)
- Sales representatives in wholesale and manufacturing (excluding technical and scientific products) (63%)
- Software quality assurance analysts and testers (52%)
- Information security analysts (49%)
- Computer user support specialists (47%)
This methodology highlights several compelling observations, as pointed out by Jason Andersen, VP and principal analyst at Moor Insights & Strategy.
“Actual application does not necessarily align with theoretical capacity,” he commented. “Individuals are still in the process of understanding AI’s full potential and associated risks.” The approach is noteworthy because it corroborates observations made by him and other analysts concerning tasks and job functions, and it offers a straightforward framework for making assessments.
“Such methodologies are likely to endure,” Andersen affirmed.
Massenkoff and McCrory recognize that while Anthropic’s methodology may not encompass “every avenue” through which AI could transform the job market, it offers a way to “more dependably pinpoint economic upheaval compared to retrospective analyses.”
“The effects of AI might become unequivocally clear,” they state. “This framework proves particularly valuable when impacts are uncertain, potentially aiding in the early identification of at-risk jobs before displacement becomes evident.”
AI Necessitates Substantial Sector-Specific Transformation
Consistent with this study, Andersen noted that he observes AI implementation primarily enhancing, rather than eliminating, full job roles. While AI is redesigning certain tasks, human involvement remains paramount. “Automation focused on tasks yields a gradual benefit, boosting employee productivity and expanding their capabilities.”
However, a crucial element missing from both analyses is the broader implication for how work will be conducted in the AI era. To effectively leverage this nascent technology, both workflows and job functions must evolve. Until these adaptations are addressed sector by sector, businesses will largely remain “at a standstill,” according to Andersen.
He highlighted that this situation could disproportionately affect younger job seekers for an indefinite period. Concurrently, current employees might oppose workflow modifications unless such changes are “substantial and structured to acknowledge experience and specialized knowledge.”
Presently, AI is often perceived as a tool to delegate tasks typically performed by less experienced personnel, a trend Andersen identifies as problematic. “It is essential to reconfigure tasks and roles to achieve equilibrium,” he asserted. The positive aspect, he added, is that businesses will be motivated to undertake this, particularly as workplace demographics in developed nations shift with an increasing number of white-collar retirements.
“It is unsustainable for [organizations] to solely employ high-cost, highly experienced staff,” Andersen explained. “A long-term equilibrium is necessary.”
