U.S. Census Bureau finds 55% of workers report using AI on the job, with productivity gains uneven

The U.S. Census Bureau’s Household Trends and Outlook Pulse Survey asked U.S. workers in March whether they had used AI for any of 11 specified work-related tasks. About 55% of respondents reported using AI for at least one task, with the most common applications being information search (37%), writing communications or documentation (32%), idea generation (32%), interpreting or summarizing information (31%), and administrative tasks (27%).

Among workers who had used AI during the previous week, 31% estimated that it saved them one to two hours, 15% said they saved three to four hours, and 15% reported saving more than four hours. Conversely, 10% said AI did not save them any time, and 3% reported that using AI required additional time to complete their work.

The Census findings reflect AI use across U.S. workers broadly, not government employees specifically, but they offer a baseline for understanding how AI is being integrated into routine workplace activities. The survey is conducted every two months to provide timely data on changing social and economic conditions.

In parallel, Louisville Metro Government in Kentucky has pursued an employee-led AI adoption strategy under the guidance of Pamela McKnight, the city’s chief AI officer. The approach emphasizes identifying employees interested in AI, providing them with training and tools, and supporting them as internal champions within their departments.

McKnight emphasized the need for governance and guardrails, stating that AI deployment should not be a free-for-all. She described her office’s role as an enabler that works alongside employees to train them, educate them, and help them use AI tools appropriately within established cybersecurity frameworks.

Government officials also cautioned that not all AI systems are equally suited to every task. Brendan Taylor, Domain Information Security Officer for the State of Tennessee, advised matching specific AI products to the intended use case, noting differences in strengths such as security, coding, or productivity tasks like report generation.