Gallup published polling on July 21, 2026, showing that 52% of US employees use AI in their jobs, up from 27% two years earlier. The figure crossed the halfway mark, with substantial variation in what people use it for.
Fifty-one percent reported writing and editing, forty-nine percent search and research, and thirty-nine percent general assistance or problem-solving. These categories describe common work; they leave the quality and business effect of that work to be evaluated.
Among employees using AI for one or two tasks, 45% reported a positive impact on productivity. Among those using it for seven or more purposes, the figure was 90%. The 45-point difference is an association between breadth of use and self-reported impact. It does not show that assigning more AI tasks causes better productivity.
Define the work behind the usage count
VentureBeat’s enterprise survey that week covered 101 organizations. 71% said a quarter or fewer of their deployed agents performed real multi-step work. That finding raises a question about what organizations count as an agent. It does not establish that the remaining systems produce no measurable business value.
Choose a process, name the person responsible for it, and specify what the AI is expected to contribute. Define the inputs, the output standard, and how people review exceptions. Training and tool spending can then support a concrete assignment.
Gallup found that 47% of organizations had integrated AI tools to improve productivity. Yet 20% of employees surveyed did not know whether their company had adopted AI technology. One in five being unsure about adoption gives leaders a reason to explain which tools are approved and what they are for. It says less about whether those employees know the company’s entire strategy.
Explain the assignment and measure it
Pew Research Center found in a parallel survey that 31% of Americans expected a negative personal impact from AI, while forty percent expected a negative impact on society. Workers under 30 were the most skeptical group. Those views deserve an explanation of how a proposed workflow affects people’s responsibilities and data.
For an initial 90-day test, select a workflow with an observable result. Record the time spent preparing and reviewing output, the errors that require correction, and the exceptions people handle. Keep data-use rules explicit. A tool can be used frequently while leaving a manager with substantial rework.
Gallup’s adoption figures establish that workplace AI use is widespread. Your team’s review needs to establish something more local: what changed in the process, who benefited, and whether the result justifies continuing or expanding it.