AI is spreading through American workplaces, but the average adoption figure hides a sharp divide between occupations. Gallup surveyed more than 23,000 full- and part-time US workers and found usage concentrated in roles built around digital information, writing, and analysis.
More than three quarters of IT workers report using AI at least a few times a year. The figure is just under 60 percent in finance and professional services, while roughly one third of retail workers report comparable use.
The gap shows that workplace AI is currently shaped as much by job design and organisational communication as by employee enthusiasm.
Current Tools Fit Desk Work Better
The most common uses remain consolidating information, searching, and generating ideas. More than 60 percent of AI users interact with chatbots, while writing and editing follow behind. Coding assistants and data-science tools are less common but popular among frequent users.
These tasks align with the strengths of language models: transforming digital information that is already available to the worker. Technology, finance, and professional services naturally contain more of that work than jobs centred on physical activity or face-to-face service.
Healthcare and manufacturing use AI more than retail in Gallup’s data, yet regulated environments can slow adoption even when digital workflows exist. Privacy, safety, accountability, and legacy systems affect whether an available capability becomes an approved daily tool.
This means leaders should not treat unequal use as a simple training deficit. A store employee, technician, or care worker needs tools embedded in the equipment and decisions of the role, not a generic chatbot designed around office documents.
Many Workers Do Not Know Their Employer’s Position
Nearly one quarter of respondents were unsure whether their organisation had adopted AI. Just over one third said their employer had implemented it, while 40 percent said it had not.
Gallup notes that earlier surveys did not provide a “don’t know” option, which encouraged guessing and made the apparent rise between 2024 and 2025 harder to interpret. Once uncertainty could be expressed, a large information gap became visible.
Non-managers, part-time employees, and hands-on workers were more likely to be unaware. The further someone sits from technology decisions, the less likely they are to know what is approved or available.
That ambiguity creates both underuse and shadow use. Some employees may avoid valuable tools because they assume AI is prohibited; others may use public services without understanding data rules. A clear policy, list of approved tools, examples, and support route is an inexpensive intervention.
Occasional Use Is Not Embedded Adoption
Around 45 percent of workers say they use AI a few times per year, but only about 10 percent use it daily. Access has expanded faster than integration into routine work.
Daily use should not become the objective for every role. Frequency is valuable only when the tool improves an outcome. Organisations need to identify repeatable tasks where assistance saves time, raises quality, or removes friction, then measure whether those changes appear after review and correction.
Frequent users are more likely to adopt advanced tools, suggesting that capability compounds through experience. Structured practice, role-specific examples, and peer support can help occasional users move beyond one-off experimentation without forcing adoption where the fit is weak.
The survey also warns against announcing a universal transformation based on knowledge-worker experience. Most American employees are not using AI every day, and many are uncertain whether their employer uses it at all. Benefits and exposure will vary with occupation, industry, authority, and access.
Leaders have a straightforward starting point: tell employees what the organisation is doing. The harder next step is to redesign tools and workflows for people outside the office. Workplace AI will remain fragmented until it becomes useful in the actual environment where each role makes decisions—not merely available in another browser tab.
