When an HR team decides whether an AI rollout is working, it often starts from what the people sponsoring it can see, and those people are the heaviest users in the company. I think that is a measurement mistake, and a new survey gives it a number.
The gap in the data
Channel V Media, a public relations agency, published an analysis of a June 2026 Prosper Insights & Analytics survey of 7,675 U.S. adults. In it, 53.2% of executives and business owners use generative AI, against 37.9% of employees. For agentic AI the gap widens to 2.7 times: 17.5% of executives and 6.6% of employees.
The same report shows that executives who use generative AI lean toward writing help, research, creative work and content creation, while the two groups converge on plain internet search. Channel V Media reads that as leaders moving past search while much of the workforce has not followed.
Two cautions apply. Channel V Media sells services that shape how AI engines describe companies, and it selected the questions and wrote the interpretation, while Prosper fielded the survey. The sample is U.S. adults by job role, not the staff of any one employer. I use the numbers as a direction, and I would not apply them to your own headcount.
Why HR should care
Agent rollouts get approved in rooms where most attendees already use the tools. Demos run well there, the sponsor reports good results from personal use, and the pilot gets judged against that experience. The employees who will use the agent every day for scheduling, case handling or candidate screening never appear in the room’s sample.
SAP’s own research makes the same worry visible from the executive side. Its SuccessFactors Future of Work Research Lab reports that 72% of executives believe organizations claiming to transform work with AI are primarily automating existing processes. Siva Sundaresan, general manager and chief product officer for SAP SuccessFactors, wrote in the same post that “automation is not the destination.” If executives themselves doubt that current programs change the work, an executive’s satisfaction with a tool cannot be the pass mark.
The strongest counter-argument
The best objection is that leaders always adopt first. New tools reach the top of an organization early, the gap narrows, and executive enthusiasm is what funds the program. On that view, a gap of fifteen points is a normal stage and not a design flaw.
I accept the timing point and still disagree with the conclusion. A gap that closes by itself does not tell HR where the rollout is stuck. And the same survey shows that executives’ comfort has limits: 76.8% of them still want a live person for banking and 77.4% for healthcare. People who hold that line in their own lives may well want a human in the loop on decisions about pay, performance and hiring. That is my inference, and the survey did not ask about it. It is still a reason to test an agent on the people who will live with its decisions, and not only on the people who approved it.
What I would measure instead
- Usage by role level. Report active use of each AI tool for managers, individual contributors and frontline staff separately, so a high executive number cannot hide a low one.
- The lowest-use group first. Pick pilot teams from the roles with the least current use. If the agent works there, it works.
- Task-level change. Record which steps of a workflow the agent took over and which roles changed as a result. Hours saved describes automation, and the executives in SAP’s survey are asking about something else.
- A human fallback for people decisions. Write down which outcomes an agent may recommend and which need a named person to approve.
I made related arguments in our piece on workers faking their AI fluency and in the opinion on employers free-riding on workers’ AI skills. Both point to the same place: what leaders report about AI and what staff do with it are separate data sets, and HR is the function that can collect the second one.
Before the next sponsor review, ask for the usage report split by role. If the only chart shows the company average, request the cut that includes the employees furthest from the executive suite, and judge the rollout on that.
Source: Channel V Media, The State of AI Adoption in America, 2026
