AI is already on the job, and Census is keeping count

Most people probably associate the U.S. Census Bureau with the massive effort to count everyone in the country once every 10 years. But Census researchers are hardly sitting around waiting for the next decennial tally. The agency constantly surveys Americans about how they live, work and even play. The resulting information helps government agencies set policy, support programs and track how conditions across the country are changing.

And lately, that changing world includes a lot of artificial intelligence.

In March, the Census Bureau’s Household Trends and Outlook Pulse Survey asked U.S. workers whether they were using AI on the job and, perhaps more importantly, what they were actually doing with it. The results offer a detailed snapshot of how workers are using AI today.

About 55% of U.S. workers said they had used AI for at least one of 11 work-related tasks included in the survey. The most common applications were fairly ordinary. About 37% used AI to search for information or technical help, 32% used it to write communications, documentation or instructions and another 32% had AI help them generate ideas. About 31% used AI to interpret, translate or summarize information while 27% used it for administrative tasks.

In other words, much of workplace AI adoption does not seem to involve building autonomous agents or developing sophisticated machine learning models. People are using it to help with the kinds of things they already do every day.

Many also said it was saving them time. Among workers who had used AI during the previous week, 31% estimated that it saved them one to two hours. Another 15% said they saved three to four hours and 15% reported saving more than four hours.

AI did not save time for everyone. About 10% said it did not save them any time and 3% said using it actually required additional time to complete their work. So, the reported productivity benefits were not universal.

The Census findings cover U.S. workers generally, not government employees specifically. But they raise an interesting question for agencies trying to expand their own use of AI: As employees become more comfortable with the technology, what does practical adoption look like inside government?

I recently moderated a discussion about government AI adoption with government and industry officials that included Pamela McKnight, chief AI officer for Louisville Metro Government in Kentucky. Louisville has taken exactly that kind of approach, combining centralized governance with efforts to get AI capabilities into the hands of employees throughout the organization.

“There has to be an understanding that business will be done differently now,” McKnight said. “It’s not just about educating people about how to use an AI tool, but also identifying a use case within their area and then actually implementing that AI tool into their workstream.”

Rather than expecting a central AI team to discover every possible use case, Louisville has worked to identify employees who are interested in the technology, give them access to training and tools, and encourage those with both interest and aptitude to become AI leaders within their own areas of government.

Louisville's experience offers a government example to place alongside the broader workplace patterns Census measured. Its approach is to let employees identify potential use cases within their own areas, then pair that operational knowledge with access to AI tools, training and support. In Louisville, however, giving employees room to experiment does not mean creating an AI free-for-all, or deploying AI simply because the technology is available.

“You have to have guardrails in place, and we worked very closely with our cyber teams on those, so that is number one,” McKnight said. “And then our office is an enabler. We come along beside them and support and enable them as partners … we train them, we educate them and we show them how to specifically use those different AI tools.” 

Another implementation question is which AI tool best fits a particular government task. At another discussion I recently moderated, state and local government officials warned that different AI systems have different strengths and weaknesses.

“Make sure you have the right AI as they are not all created equal and will be better at different things,” said Domain Information Security Officer for the State of Tennessee Brendan Taylor. “And even if you have OpenAI or Mythos, Gemini or whatever, make sure you are using their right product for what you are trying to do. Some are better at security or coding or productivity tasks like making reports. Make sure you are matching with the right solution.”

And while government technology leaders are working through those practical questions, the Census Bureau will continue doing what it does best: measuring what is actually happening.

The Household Trends and Outlook Pulse Survey is conducted every two months to provide timely information about changing social and economic conditions. In this case, it captured a workforce already experimenting with AI in very practical ways. Among workers who had used AI for at least one of the tasks Census measured, 24% said they used it every day during the previous week while another 46% had used it on at least one day.

The survey does not say whether employers should encourage more AI use or what tools they should deploy. But it does provide a useful snapshot of where things stand today: what workers are using AI for, how often they are turning to it and, in many cases, how much time they believe it is saving them.

And for government agencies trying to understand how AI is finding its way into everyday work, that may be one more Census count worth watching.

John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys