A new Harris Poll survey conducted for INTOO finds that 72% of U.S. employees say artificial intelligence is already used for business purposes somewhere inside their organization. But the way it got there is the real signal for HR: more of that use is informal and employee led than it is sanctioned and governed. The shift underway is not whether AI arrives in the workplace. It is who is deciding how, and the answer, right now, is mostly not HR.
The Adoption Numbers Tell a Governance Story
The INTOO study, fielded by The Harris Poll among 1,085 full or part time employed U.S. adults on June 15 and 16, 2026 (margin of error plus or minus 3.5 percentage points), breaks that 72% headline figure into tiers that matter more than the topline. Only 12% of employees report AI use through a formal, organization wide process. Another 20% describe departmental use through company approved tools. But 22%, the single largest slice, say AI is used informally, without company approval at all. A further 17% say AI is already reshaping workflows and workforce planning. Sixteen percent report no AI use, and 9% are not sure how or whether it is being used.
Read against each other, those numbers describe an adoption curve running ahead of its own governance. Informal, unapproved use (22%) now outpaces fully governed, organization wide programs (12%) by nearly two to one. For a function built around policy, consistency, and risk management, that is the uncomfortable part: the largest category of workplace AI use today is the one nobody in HR or IT explicitly signed off on.
Confidence Follows Exposure, Not a Mandate
Where sentiment splits
Employee attitudes toward AI are not uniformly positive or negative; they split into distinct camps. Thirty four percent say they are curious to learn more, and the same share say they are excited about AI’s potential. Thirty one percent describe themselves as confident using AI tools. Set against that: 20% worry AI could make their job obsolete, another 20% want to advance their AI skills but do not know where to start, 20% worry about making mistakes with the tools, and 10% say they feel intimidated by them.
Exposure changes the picture, but not evenly
The clearest pattern in the data is what happens once an organization actually uses AI. Among employees at companies with active AI use, 43% say they are excited, compared with just 11% at organizations with none. Confidence nearly quadruples, from 10% to 39%. Curiosity rises from 22% to 38%. Direct exposure, in other words, does far more to build comfort with the tools than any messaging campaign could.
What exposure does not do is erase the anxiety. Job obsolescence concern sits at 21% among employees at AI using organizations and 19% among those at organizations without AI, a gap close enough to call negligible. Adopting the technology, even successfully, does not by itself reassure the workforce that its own role is safe.
What This Means for the HR Leader
The practical read for HR is that the 22% informal-use figure functions like a shadow IT problem for people practices. Employees are already running real work, drafting job descriptions, screening resumes, prepping performance reviews, through consumer AI tools that HR did not select, did not train anyone on, and cannot audit. That is not a future risk to plan for. INTOO’s data says it is already the largest single category of AI use in the modern workplace, ahead of anything HR or IT has formally rolled out.
It also reframes what “AI adoption” should mean as an HR initiative. The instinct is to treat it as a tooling decision: pick a vendor, license some seats, announce a rollout. The data argues for treating it first as a governance and skills gap. Twenty percent of employees say outright that they want to build AI skills but do not know where to begin. That is a training and change management problem HR already owns the tools to solve, independent of whichever platform IT eventually licenses. Publications on this beat have already tracked the adjacent risk: as HRTech Edition reported on enterprise AI adoption splitting into two tracks, organizations where HR has a real seat at the table are pulling ahead of those where AI rollout is left entirely to IT. This survey suggests the cost of not having that seat shows up first as ungoverned, informal use, not as a failed formal program.
How to Evaluate the Gap in Your Own Organization
Three steps follow directly from the data. First, run an honest inventory of informal AI use before launching a formal program: ask managers and teams what tools they already reach for, rather than assuming a blank slate. Second, build AI literacy training around the specific workflows employees already use the tools for, since generic “introduction to AI” sessions will not close the 20% who say they want to learn but do not know how to start. Third, do not expect adoption alone to calm displacement anxiety. Because the fear of obsolescence barely moves between AI using and non-AI using organizations, HR communication has to make an explicit case for why the organization is adopting AI and what it means for roles, not just issue guidance on how to use it. Employees are already deciding how AI shows up in their work. The organizations that get ahead of that, rather than discovering it after the fact, are the ones setting the terms of the next phase of adoption instead of reacting to it.
Source: INTOO
