The AI Your Organization Doesn’t Know It’s Using
A striking data point was cited in a recent Harvard Business Review article: according to a global study by KPMG and the University of Melbourne fifty-seven percent (57%) of employees report actively hiding their use of AI at work. More than half. In most organizations, that means the majority of AI-assisted output is flowing through invisible channels that is frequently undisclosed, untracked, and unmanaged.
The knee-jerk reaction is to frame this as a compliance problem. And it is, in part. But treating hidden AI use primarily as a governance failure misses the more important issue underneath it.
Why employees hide it
When employees conceal how they work, it’s rarely because they’re doing something wrong. More often, they’re navigating institutional ambiguity. They are unsure whether AI use is sanctioned, afraid of being seen as cutting corners, or simply uncertain how leadership will react. In the absence of clear norms, self-preservation becomes the default posture. The concealment itself is a signal: the organizational environment hasn’t made honesty safe.
This matters because the response to the KPMG finding in many organizations will be to tighten policy, implement usage tracking, require disclosure, and add it to the acceptable use framework. These are reasonable steps. But policy doesn’t rebuild trust. It creates compliance. Those are different things and conflating them is simply a mistake.
What the organization actually loses
The operational cost of shadow AI extends well beyond risk exposure. When employees develop effective AI workflows in isolation, that knowledge stays siloed. The prompt sequence that saves a team member two hours a week never gets shared. The use case that failed and taught a lesson about accuracy or hallucination disappears with the conversation. The organizational learning loop never closes.
In knowledge work, competitive advantage increasingly lives in how people work, not just what they know. AI amplifies this dynamic significantly. Organizations that learn collectively how to use these tools effectively will compound those gains over time. Organizations where AI use is covert will not — because institutional knowledge requires institutional visibility.
This is the deeper problem the 57% figure surfaces. It’s not just that employees are using unsanctioned tools. It’s that organizations are forfeiting the compounding productivity benefits that come from transparent, shared learning. Every hidden workflow is knowledge that could have trained the next person, shaped the next policy, or informed the next investment, but didn’t.
What governance alone can’t do
Governance frameworks are necessary scaffolding. Data classification, approved tool lists, incident reporting protocols: these create the structural conditions for responsible AI use. But they are insufficient conditions for a high-functioning AI culture.
The gap between necessary and sufficient is filled by trust. Employees need to believe that disclosing AI use won’t invite scrutiny, diminish perceived competence, or create career friction. They need to see leadership model transparency rather than simply mandate it. They need formal or informal forums where AI experimentation is discussed openly, including the failures.
Organizations serious about capturing AI’s productivity upside should be asking not just “how do we govern AI use?” but “why don’t employees feel safe telling us how they’re already using it?” The answer to the second question does more to close the gap than any policy update will.
The most productive AI-enabled organizations won't be the ones with the strictest policies — they'll be the ones where people feel comfortable saying "here's how I used AI today, and here's what I learned." That kind of openness doesn't come from a framework. It comes from culture. And culture starts with the signal leadership sends about whether honesty is safe.