Customer service, secretary, and sales positions have shrunk for two consecutive years: the latest U.S. Bureau of Labor Statistics data shows AI is replacing these 17 occupations

According to data from the U.S. Bureau of Labor Statistics (BLS), 18 occupations classified as "highly exposed to AI" account for approximately 10 million jobs in total. Between 2024 and 2025, with the exception of healthcare, 17 of these occupations have declined by 1.6% for the second consecutive year. (Background: Cerebras's $95 billion market cap debut — "Trump's eldest son's fund" participated in both funding rounds) (Background: Anthropic report: In the 2028 battle for AI supremacy, if the U.S. fails to maintain its compute advantage, it risks being overtaken by China) Within the year, more than 130,000 customer service representatives have vanished from the U.S. job market, with AI quietly filling the seats they once occupied. The U.S. Bureau of Labor Statistics (BLS) has flagged 18 occupational categories as "highly exposed to AI" — meaning the tasks performed in these jobs can reasonably be partially or fully replaced by current AI systems. According to the BLS annual Occupational Employment and Wage Statistics, these 18 "highly AI-exposed" occupations together account for about 10 million jobs. From May 2024 to May 2025, employment in this group fell by 0.2%, while overall U.S. employment grew by 0.8% during the same period — a clear divergence has emerged. Bloomberg's reporting further breaks it down: if you exclude the special case of "medical secretaries and assistants" (an occupation that, due to the healthcare industry's own high growth, is bucking the trend with increased hiring), employment in the remaining 17 occupations declined by 1.6% between 2024 and 2025 — and this is the second consecutive year of contraction at the same magnitude. Looking specifically at the 2024-to-2025 numbers, the three hardest-hit occupations are: - Customer service representatives: down 130,180 jobs, a decline of 4.8% - Secretaries and administrative assistants (excluding medical, legal, and executive): down 31,030 jobs, a decline of 1.8% - Wholesale and manufacturing sales representatives (excluding technical and scientific): down 28,670 jobs, a decline of 2.3% Extending the timeline further — from May 2022 (the last complete data point before ChatGPT's debut) to today — the occupations with the most severe cumulative contraction are, in order: credit authorizers, checkers, and clerks down 26.2%; broadcast announcers and radio DJs down 20.8%; and sales engineers down 13.2%. To understand why this contraction is concentrated in occupations like customer service, secretarial work, and auditing, it helps to first distinguish between two concepts. "AI-replaceable" jobs refer to work where AI can directly perform the core tasks of the role: for example, answering customer calls, processing refund requests, replying to standard questions, reviewing credit files, and entering data into forms. The common features of such jobs are: predictable inputs, standardizable outputs, and documentable workflows. AI doesn't need to understand context — it just needs sufficient training data. "AI-augmented" jobs are different. Here AI plays the role of a tool that helps humans work faster and better: doctors using AI to read pathology slides, lawyers using AI to quickly compile case precedents, engineers using AI to generate test code. The core judgment in these professions still rests in human hands; AI merely accelerates the work. An earlier Goldman Sachs report from April 2026 offered a more granular snapshot: AI is net-eliminating about 16,000 U.S. jobs per month and replacing roughly 25,000 positions every four weeks, while AI-augmented roles have added back about 9,000 over the same period. This isn't a sudden collapse — it's a structural substitution unfolding slowly, measured in months. Bloomberg's reporting highlights a structural paradox: this wave of AI substitution is concentrated precisely in the types of occupations traditionally regarded as "career entry points." Customer service, data entry, billing processing, basic financial analysis, entry-level software development… these positions have historically been where fresh graduates build their first work experience. The Goldman Sachs report likewise notes that Gen Z is the demographic most deeply affected by this wave, because they enter the workforce en masse precisely through these entry-level jobs where AI substitution is most efficient. This creates a situation unseen in past decades of technological change: previous waves of automation typically started eating away at employment from mid-to-high-level repetitive technical jobs, leaving the entry-level tier as a stepping stone for newcomers. But this time, the entry-level tier is the one shrinking first. From an industry perspective, the most affected sectors include call centers, content moderation, legal support administration, accounting, and early-stage software development support roles. What these areas have in common is their heavy reliance on document processing, rule-based judgment, and standardized communication — precisely the capabilities at which current LLMs (large language models, i.e., AI systems that process and generate text) excel. There are currently two schools of thought on AI replacing human work: one holds that this trend will only accelerate; the other believes that sharply rising productivity will give birth to many new jobs. What do you think?