AI
Survey: 3% of US Workers Lost a Job to AI; Call Centers Hire
A YouGov survey of 1,250 employed Americans finds 3% lost a job to AI since 2023 and 6% found one because of it. The same day, Apollo's chief economist pointed out that Philippine call-center employment has doubled to 2 million.
A customer-support team lead in Manila would have had a strange four years. Every quarter brought a report predicting her job's disappearance; every quarter her floor added seats. Two pieces of data published by Fortune on August 29 put numbers on that gap between forecast and payroll.
What 1,250 workers said
The first is a survey YouGov ran for the sociologist Jeffrey C. Dixon: 1,250 employed U.S. adults, 25 questions, fielded July 30 to August 4, 2026, not yet peer-reviewed. Asked whether they had lost a job because of AI since 2023, 3% said yes. Asked whether they had gained an AI-related job, 6% said yes; 9% credited AI for a promotion or advancement. Across every question about AI-driven changes, between 88% and 95% answered no. Dixon's summary is blunt: "AI is not having widespread effects on the labor market today."
Fortune lists the caveats and they all lean the same way. The sample excludes people who are currently out of work, so someone displaced by AI appears only if they found another job. Respondents are YouGov panel volunteers. Workers can misattribute a job change to AI, or attribute an AI-caused one to something else. And the survey is a single snapshot, taken in a month when U.S. unemployment stood at 4.1% and AI was the reason companies most often cited in layoff announcements. The interesting number is the distance between those announcements and what workers report.
The Manila test
The second piece is an argument from Apollo's chief economist Torsten Slok. If AI were displacing white-collar work at scale, he says, you would see it first in the Philippines and India, where the most automatable work has been offshored. Instead, Philippine call-center employment nearly doubled between 2016 and 2025 to 2 million people; unemployment there fell from 9% in 2021 to about 5% by July 2026, and in India from about 7% to 6%. This in a sector where Brookings rates 86% of customer-service tasks as having high automation potential.
His explanation is the Jevons paradox: when a task gets cheaper and faster, demand for it rises. Erik Brynjolfsson's 2023 Stanford study of more than 5,000 agents found an AI assistant lifted productivity 14% per hour; the evidence since suggests firms used that to serve more customers rather than to cut agents. Wages help explain why the floor is in Manila rather than Ohio: Filipino agents earn between roughly $243 and $1,948 a month against a U.S. average near $2,866.
What that team lead should actually watch
Not the headline forecasts. Last week we covered why Meta abandoned its plan to replace staff with AI; the plan had underestimated what the work involved, which is the same finding these two datasets reach from the other direction. Dixon's own warning is the one to keep: agentic systems could devalue current expertise in ways a July survey cannot see.
The pattern we see at companies that deploy support chatbots matches the Manila numbers. Headcount rarely falls; agents move from repetitive questions to complaints, returns and sales, where a human is worth more. The real risk in the data is not unemployment but layoffs announced "because of AI" and quietly reversed a few quarters later. The question to ask before automating a process is not how many people it removes, but where the freed hours go.
Written by
Muhammet Fatih Batman
Founder & Editor
Founder of YZ Uzman, with 20+ years of experience in web design and software development.