Half of American workers now use artificial intelligence in their jobs, and most of them use it for a very short list of things. Gallup's second-quarter 2026 workforce study (opens in new tab), based on responses from 22,573 U.S. employees, found that 52% of workers use AI in their role and 30% use it a few times a week or more. When asked what they actually use it for, employees pointed mostly to three things: writing and editing (51%), search or research (49%), and general assistance or problem-solving (39%). Everything else trailed well behind, with 16% using AI for coding assistance, 16% for automating a process, and 16% for task, scheduling or project management. Most teams have found one or two places where a chatbot helps and have not gone looking for a third.
Access may help employees get started, but the next stage of artificial intelligence in business will likely depend on helping them apply AI more specifically, consistently and practically in the work they do.
— Andy Kemp, Gallup
Breadth is where the measurable return shows up. Gallup sorted AI users by how many distinct purposes they applied the tools to, and the spread was wide. Among employees using AI for one or two purposes, 45% said it had a somewhat or extremely positive effect on their productivity. That figure climbed to 66% at three or four purposes, 78% at five or six, and 90% among those using AI for seven or more. The same pattern shows up when the data is cut by task instead of by count. Employees using AI for coding assistance or process automation were the most likely to report gains, at 77% each, followed by slide deck creation at 76% and data analytics at 75%. The two most common uses sat at the bottom of that ranking: 68% for writing and editing, 65% for search and research. The applications employees reach for first are the ones least associated with a productivity payoff, and the ones tied to the biggest gains are used by roughly one in six. The contrast between frequent and occasional users points in the same direction. Employees who use AI a few times a week or more are nearly three times as likely as occasional users to apply it to coding (22% versus 8%) and to process automation (21% versus 8%), and more than twice as likely to use it for scheduling or project management (21% versus 9%). Frequent users are not simply doing the common tasks more often. They have also found jobs for the tool that connect to a specific function of their work. Gallup is careful about causality here, and the caution is warranted. Employees who already find AI useful are more inclined to keep finding new uses for it, and some roles offer more openings than others. Still, the distance between 45% and 90% is hard to explain by selection alone.
of employees who use AI for seven or more purposes say it has improved their productivity, compared with 45% of those using it for one or two. Gallup, Q2 2026 Workforce Study →
For a small business, this changes what an AI rollout is for. Buying licenses and announcing that the tools are available produces the narrow-use pattern by default, because drafting help and search are the two applications a person discovers without being taught. The uses tied to the strongest gains, such as automating a recurring process, building a report that regenerates itself, or assembling a deck from material that already exists, are the ones somebody has to name out loud, since spotting which part of a given job is a candidate takes deliberate thought. That is a management job rather than a purchasing one, and in a company without a dedicated technology function, nobody is holding it. The same study surfaced a clarity problem alongside it. While 47% of employees say their organization has integrated AI tools, 33% say it has not, and 20% do not know either way. A fifth of the workforce cannot say whether AI is officially part of how their company operates. Gallup's April 2026 report (opens in new tab) found a related ceiling: among employees at organizations that have adopted AI, only 8% strongly agree that AI has changed how work gets done there. The gains are landing on individual tasks and stopping short of the systems around them. The useful first move looks more like a workflow review than a technology decision. Which recurring tasks in a role take the longest, and which of those are shaped like something a tool could take a first pass at? Owners who have already worked out where their own hours go have most of what they need to run the same review for a team. In a one-on-one, the useful question is narrow: what did you do last week that you would rather not do by hand again? A short list of answers from four or five people usually surfaces more candidates than any vendor demo will. None of that argues for handing over more than makes sense. Work that costs something when it goes wrong still needs a person accountable for checking the output, and widening the range of tasks a team gives to AI raises the value of drawing that line deliberately. What the breadth finding suggests is that the ceiling most teams are hitting is set less by the tools themselves than by how few jobs anyone has thought to give them.
