When employees avoid AI, hide their use of it, or resist adoption, the signal points to the organizational conditions around them, not to individual resistance.
Mercer’s 2026 research found that forty percent of employees fear AI will cost them their job, up from twenty-eight percent two years earlier. In the same study, sixty-two percent felt their leaders were underestimating the emotional and psychological impact of AI on their teams. ResearchGate’s research on workplace AI anxiety identifies a consistent behavioral split: employees who perceive AI as a threat move toward avoidance, while those who perceive it as a challenge move toward engagement. The perception that determines which direction an employee goes is shaped far more by organizational conditions than by individual personality.
The Human Performance Intelligence framework categorizes AI anxiety as a workforce-level signal, measurable across the four pillars of the framework and addressable through changes to organizational conditions rather than individual interventions. Understanding what the signal is measuring is the necessary first step.
What the Research Identifies as the Drivers
The technostress literature identifies two primary drivers of AI anxiety in workplace settings. The first is FOBO: fear of being obsolete. The second is the cognitive load that AI-enabled work environments place on employees who have not been adequately prepared for them.
FOBO is a structural response to genuine uncertainty. When organizations deploy AI without communicating clearly what will change for specific roles, employees fill the gap with their own projections. In the absence of honest information, those projections tend toward the negative. The fifty-seven percent of employees who report hiding their AI use and attributing AI-generated work entirely to themselves are not expressing a character flaw. They are responding rationally to an environment where AI use feels risky rather than supported.
Cognitive load operates differently. AI-enabled work environments increase the mental demand on employees in ways that are not always visible. More decisions arrive faster. Outputs need to be reviewed critically rather than accepted at face value. The pace of work increases without a corresponding increase in the structural support required to sustain it. HBR’s documentation of AI brain fry captures the endpoint of this trajectory: the mental fatigue that builds in employees who spend sustained periods overseeing AI outputs they cannot fully trust. Anxiety and fatigue compound each other in environments where both are present and neither is addressed.
How the HPI Framework Maps the Signal
The Human Performance Intelligence framework maps AI anxiety across its four diagnostic dimensions, each of which contributes to the picture in a distinct way.
Within the human performance pillar, AI anxiety shows up as cognitive load, attentional depletion, and degraded decision quality. Employees experiencing high AI anxiety spend cognitive resources on worry and vigilance that are unavailable for the actual work. Over time, this produces the performance decline that looks, from the outside, like resistance or disengagement.
Within the workplace wellbeing pillar, anxiety manifests as technostress, reduced psychological safety, and the erosion of recovery boundaries. Research consistently shows that employees who do not feel psychologically safe are less likely to experiment with AI, less likely to flag errors in AI outputs, and less likely to share learning with colleagues. The wellbeing conditions of the team determine whether anxiety stays contained or compounds.
Within the people management pillar, the signal points to leadership behavior and communication quality. Teams whose managers have clear answers about what AI means for their roles, and who model genuine engagement rather than projected confidence, show significantly lower anxiety levels. The management layer is both a primary driver of anxiety and the primary lever for reducing it.
Within the AI-enabled work pillar, the signal maps to workflow design. Where task boundaries are unclear, where review processes have not been adapted to the pace of AI-assisted production, and where governance around AI use is absent, anxiety is consistently higher. The structural conditions of how work is organized either support or undermine an employee’s ability to develop a stable, confident relationship with AI tools.
The Adoption Clusters That Anxiety Creates
One of the more consequential effects of unaddressed AI anxiety is the adoption clustering it produces inside organizations. Research from ResearchGate and the broader AI adoption literature identifies a consistent pattern: early adopters develop competence and confidence while anxious employees fall further behind, and the gap between those two groups grows over time without deliberate intervention.
This matters for organizational performance because the value of AI in team-based work depends on shared capability, not individual capability. A team where half the members are skilled AI users and half are avoidant produces fragmented workflows, uneven quality, and coordination friction that absorbs the productivity gains the skilled users generate. The performance ceiling for the organization is set by the distribution of capability across the team, not by the capability of its strongest individual performers.
Research also shows that adoption clustering correlates with demographic factors including generation, gender, and role type, creating equity implications that organizations are only beginning to account for. The HPI framework treats these patterns as organizational design problems with organizational design solutions, not as population-level differences in adaptability.
What Reduces Anxiety at the Organizational Level
The HPI framework’s diagnostic approach to AI anxiety points toward three categories of intervention, each grounded in the research on what actually changes the signal.
The first is communication specificity. General reassurance about AI does not reduce anxiety. Specific, role-level clarity about what will change, what support is available, and what the new expectations look like does. Employees who receive concrete answers to concrete questions about their situation report significantly lower anxiety than those receiving organizational messaging about opportunity and growth.
The second is involvement in design. Employees who participate in defining how AI will be used in their context develop a sense of agency over the transition that reframes it from something being done to them to something they are shaping. Peer-to-peer knowledge sharing, where early adopters work alongside more anxious colleagues rather than being separated into different competence tiers, has been shown across multiple studies to reduce anxiety and accelerate adoption more effectively than formal training alone.
The third is management preparation. The anxiety gap between teams in the same organization most often traces back to whether the manager had the tools and confidence to lead the conversation. Equipping managers to address AI anxiety directly, including the willingness to acknowledge their own uncertainty, is among the highest-leverage interventions available to an organization navigating AI adoption.