Early task loss data
Recent survey reports indicate about 20% of U.S. full‑time workers say AI has already replaced some of their job tasks, while roughly one in 12 workers report using AI agents for work activities (thecooldown.com). These are self‑reported figures from multiple polling efforts and have been discussed as early signals of task displacement rather than precise employment tallies (futurism.com).
One in five United States full-time workers now say artificial intelligence has taken over tasks they used to do themselves. (nbcnews.com) That figure comes from an Epoch AI and Ipsos survey of 2,021 United States adults fielded March 3-5, 2026, using Ipsos KnowledgePanel, a probability-based online panel weighted to Census benchmarks. The same survey found half of adults had used artificial intelligence in the previous week for work or personal tasks. (epoch.ai) The survey also found 15% of full-time workers said artificial intelligence created new tasks they would not have done otherwise. Among adults who used artificial intelligence in the past week, nearly half used it two to five days a week. (nbcnews.com) A separate Pew Research Center survey, released February 25, 2025, found 16% of United States workers said at least some of their work was currently done with artificial intelligence. Pew surveyed 5,273 employed adults on October 7-13, 2024, through its American Trends Panel. (pewresearch.org) Pew found 52% of workers said they were worried about future artificial intelligence use in the workplace, while 36% said they felt hopeful. It also found 32% thought artificial intelligence would lead to fewer job opportunities for them in the long run. (pewresearch.org) These numbers track task change, not payroll cuts. Epoch AI’s Caroline Falkman Olsson told NBC News that the survey showed both automation and augmentation, and that more granular research is needed to identify which tasks are changing. (nbcnews.com) The United States Bureau of Labor Statistics has taken a slower view of how new tools show up in employment data. In a February 2025 Monthly Labor Review article, the agency said technological change usually registers gradually in historical data and that projection work is not designed to capture extremely rapid change. (bls.gov) That leaves a gap between what workers say is happening inside jobs and what official labor statistics can confirm across the economy. For now, the clearest early signal is not mass unemployment data, but workers reporting that pieces of their jobs are already being handed to software. (epoch.ai)
