Counting the jobs that artificial intelligence creates or eliminates will tell us only part of its economic impact. Just as important is what happens inside those jobs: whether employees spend less time on tedious tasks, develop new skills or gain opportunities to advance. Helping people work faster is valuable, but it is not the same as making their jobs better.

Among U.S. employees who have used AI in their jobs at least a few times in the past year, 63 percent say it helps them complete tasks faster. That is according to new findings from the American Job Quality Study — a partnership between Jobs for the Future, The Families and Workers Fund and Gallup. Forty-six percent said AI helps them produce higher-quality work, and 47 percent said it allows them to spend more time on the most interesting parts of their jobs.

Finishing a task faster creates an opportunity. Whether that opportunity becomes more rewarding work, a chance to learn or simply another assignment depends on the decisions employers make.

The benefits also start from an uneven distribution of use. Forty percent of college graduates use AI at least weekly, compared with 17 percent of employees without a college degree. Managers are more likely to use it regularly than individual contributors. And 52 percent of regular users hold jobs meeting the study’s multidimensional job-quality criteria, compared with 32 percent of employees who never use AI.

That does not mean AI created those better jobs. The same groups that tend to hold quality jobs — including more educated employees and managers — also tend to use AI. AI use is also highly concentrated among remote workers, according to my work. We should not mistake an advantage among users for evidence that handing everyone a chatbot will close the gap.

Nor should we conclude that workers without degrees have little to gain. Among employees already using AI, reported improvements in creativity and work quality are similar across education groups. College graduates are more likely to report working faster, but the educational divide in adoption does not appear across every reported benefit. That is a reason to broaden opportunities to use AI effectively, rather than reserve them for employees already considered technologically sophisticated and familiar with digital technologies.

Many of these differences reflect the work itself. Not every job provides regular access to digital devices or obvious opportunities to use AI directly. A wise response should recognize those differences, rather than treating identical adoption rates across occupations as the goal.

The question, then, is not simply how widely AI is used, but whether it helps businesses produce more — and whether employees share in those gains.

In a working paper I co-authored with Bureau of Economic Analysis economists Tina Highfill and Jon Samuels, we found that industries within states with higher worker-reported AI use in 2025 and 2026 had experienced stronger output growth since 2020. The positive relationship is clearer for output than it is for employment, where the estimates are less precise but still present.

But what remains is what happens inside workplaces: How can organizations redesign work so that both businesses and employees benefit from AI?

Another study I conducted, following more than 30,000 employees through the Gallup Workforce Panel, finds that workers who report a clear organizational AI strategy are substantially more likely to use AI frequently, even after accounting for demographics and other workplace perceptions. Among frequent users, greater trust in leadership is also associated with larger self-reported productivity gains. The evidence points toward communication and trust as complements to the technology, not optional additions after the software is installed.

Thirty-one percent of AI users say their use has led their employer to ask them to take on more responsibilities. That need not be a bad thing: Increased responsibility can open paths to advancement. But it can also mean more work without greater authority or compensation.

Meanwhile, 52 percent of all employees say they have less influence over the adoption of new technology than they would like. Employers should take that seriously. Consulting workers before changing their tasks is a practical way to identify what deserves automation, where human judgment remains necessary and what support employees need.

Employers should begin with paid time to learn and test approved tools on actual job tasks. Managers should explain which uses are permitted, how outputs will be checked, how responsibilities may change, and inspire deliberate experimentation. Training should reach frontline workers, not just the people who volunteer for an AI pilot. Employees should also be able to flag errors or question a proposed use without being labeled resistant to innovation.

Workers are already reporting time savings from AI, but turning those savings into better jobs requires decisions about what employees do next, what they have the opportunity to learn and how their contributions are rewarded. That is the part of AI adoption that employers cannot leave to the software. It must remain a human-driven exercise.

Christos A. Makridis is an associate research professor at Arizona State University and a senior researcher at Gallup.