Most organizations approach artificial intelligence adoption as a technology challenge.
The conversation has largely focused on model accuracy, data security, governance and risk. Those concerns matter. But our research suggests another obstacle may be emerging inside organizations: Employees may trust AI itself more than they trust one another’s use of it.
In a recent Blanchard survey of leaders and individual contributors, nearly 43 percent of respondents reported observing undesirable AI-related workplace behaviors, ranging from subtle judgment of colleagues who use AI to reliance on AI-generated content without adequate verification. About 24 percent said these behaviors have become normalized in their workplaces, while only 18 percent acknowledged engaging in them themselves.
Respondents were therefore roughly 2.4 times more likely to report seeing these behaviors in others than to acknowledge engaging in them personally. The gap is revealing.
Employees consistently recognize AI-related friction around them far more often than they identify themselves as contributors to it. The result is a growing trust gap—not between people and technology, but among colleagues attempting to navigate a rapidly changing way of working.
This finding points to an overlooked reality of AI adoption: Even when the technology works as intended, the informal norms that develop around its use can become a significant barrier.
When leaders encourage AI adoption but rarely model its use, when employees use AI but avoid acknowledging it, or when people use AI to critique others rather than collaborate with them, ambiguity grows around what is acceptable, expected and safe.
AI is not creating these tensions. It is exposing them.
How AI shadow culture takes hold
What is quietly emerging in many organizations is an AI shadow culture: A workplace dynamic in which people use, judge, hide or avoid AI in ways that remain largely unspoken.
The phenomenon is easy to miss because it is rarely addressed in formal policies or guidelines. Instead, it emerges through everyday interactions: The colleague who quietly uses AI but never mentions it; the manager who encourages experimentation but never models it; the team member who questions the legitimacy of AI-assisted work while privately using AI themselves; the leader who endorses AI publicly but rarely or never uses it.
AI shadow culture develops when employees are uncertain about the informal rules surrounding AI use. They may understand the technology, yet remain unclear about what is acceptable, respected, rewarded or safe.
The consequence is not primarily technological. It is cultural.
AI is increasingly embedded in everyday work, yet conversations about how it is being used often remain limited, inconsistent or avoided altogether. Employees are left to interpret expectations on their own, creating mistrust and missed opportunities for collective learning.
Workplace norms shape AI adoption
Many organizations have formal policies governing AI use. Far fewer have established shared norms.
Policies answer questions of compliance. Norms answer questions about culture:
- Can I openly discuss how I used AI?
- Will my work be judged differently if I disclose it?
- Is using AI viewed as resourceful or as a shortcut?
- Can I challenge AI-generated output without appearing resistant to change?
When organizations fail to establish clear norms around transparency, accountability and collaboration, employees create their own norms. These norms emerge through small interactions, casual comments and observed behaviors rather than formal guidance.
Over time, these informal norms generate the shadow culture that shapes AI adoption.
5 workplace archetypes that undermine trust
Across our research, five recurring workplace archetypes emerged. They are not fixed identities. Most people move in and out of these patterns depending on circumstances. Yet each reflects a behavior that can either accelerate or undermine trust.
The judgmental observer
The judgmental observer signals—often subtly—that AI-assisted work is less legitimate than work produced without AI assistance. The behavior rarely takes the form of direct criticism. More often it emerges through skepticism, dismissive comments or social cues that imply using AI reflects a lack of expertise or effort.
For some employees, the attitude may reflect fears about job displacement. For others, it reflects ethical concerns or a belief that AI-assisted work lacks authenticity. Regardless of its origin, the effect can be similar: Employees become less transparent about how they work.
Forty-seven percent of respondents reported observing colleagues judge or diminish others for using AI, making it the most frequently observed of the five behaviors measured in the survey.
When employees anticipate judgment, AI use does not disappear. It becomes less visible. Organizations lose opportunities for shared learning, experimentation, and norm-setting.
A better alternative: Curiosity
Replace assumptions with curiosity. Ask how AI was used, what role it played and what ideas the individual contributed. Focus on the quality of the work rather than the mere presence of the tool.
2. The competitive user
The competitive user uses AI to critique, revise or improve upon a colleague’s work without meaningful collaboration. The intention is often efficiency. The effect can be diminished trust.
When employees run a colleague’s idea through AI and return an improved version without discussion, they inadvertently communicate that speed matters more than partnership. Over time, people become less willing to share unfinished thinking, reducing collaboration precisely when it is most needed. Forty-two percent of respondents reported observing this behavior.
AI can improve ideas. But when it replaces conversation rather than supporting it, it can erode trust.
A better alternative: Collaboration
Use AI as a collaborative tool rather than a competitive advantage. Before significantly revising another person’s work, involve them in the process and make your intentions explicit.
3. The overconfident adopter
The overconfident adopter mistakes efficiency for accuracy.
As generative AI becomes more sophisticated, polished results
