The research literature on AI and workplace stress has identified something more specific than general change anxiety. Bao and colleagues' 2025 study in Nature Human Behaviour, covering more than 8,000 workers across twelve countries, found that the psychological impact of AI adoption shows a distinct profile that differs from both general change anxiety and the technostress documented in earlier technology adoption cycles.
The researchers identify what they term algorithmic anxiety: a stress response specifically triggered by the presence of autonomous AI systems in the workplace, characterised by persistent uncertainty about one's relevance, value, and professional identity in an increasingly automated environment.
This distinction matters both clinically and practically. A manager who treats algorithmic anxiety as a variant of general change resistance will apply the wrong interventions, and the interventions will fall short.
What Makes It Different
Traditional technostress, the research dating from Brod's 1984 work and extensively developed by Tarafdar and colleagues, is primarily a competence problem. Workers feel anxious when they lack the skills to use new technology effectively. The solution is established: training, supported practice, and the gradual accumulation of competence. As competence builds, anxiety recedes.
Algorithmic anxiety has a different structure. Workers in Bao et al.'s study reported anxiety not primarily about whether they could learn to use AI tools, but about whether, once they had learned to use them, there would remain a meaningful human role for them to occupy. The anxiety was less about "can I do this?" and more about "will there still be a version of me to do it?"
This reflects something genuinely different in the phenomenology of AI compared to earlier technology adoption. A word processor replaced handwriting but left the cognitive and creative work of composition untouched. AI systems increasingly operate in the cognitive domain itself, producing analysis, recommendations, and draft outputs that feel like the substance of professional work rather than the tools for performing it.
Workers are not questioning their competence to use a tool. They are questioning whether the capabilities they have developed across their careers still constitute a meaningful professional contribution in the presence of systems that replicate many of those capabilities.
The Organisational Triggers
Algorithmic anxiety has specific organisational triggers that managers can both recognise and address.
Opaque evaluation. When workers do not understand how their performance is being assessed relative to AI outputs, or suspect that quantitative AI metrics are quietly superseding qualitative human judgment in reviews, anxiety increases substantially. Clarity about evaluation criteria is a direct management lever.
Perpetual monitoring. AI systems that track productivity in real time create an environment of continuous measurement that research consistently associates with elevated cortisol levels and reduced creative output. The monitoring intended to increase efficiency can directly undermine the higher-order performance it is meant to optimise.
Role ambiguity at the human-AI boundary. When workers are unclear about which tasks they should perform independently, which to delegate to AI systems, and which require human-AI collaboration, the ambiguity itself generates stress, regardless of technical capability to perform any of those tasks. Clear role design at the human-AI boundary is an under-recognised stress reduction intervention.
Social comparison with AI outputs. In professional contexts where AI can produce competent first drafts, comprehensive analyses, and structured recommendations faster than most humans, the natural tendency toward social comparison extends to comparison with machine outputs. Workers who experience their output as consistently below the pace or scope of AI-generated work report significantly lower professional confidence and higher baseline anxiety.
What This Means for Managers
The clinical distinction has direct practical implications for how managers respond.
Providing training addresses the competence dimension of AI anxiety. It does not address the existential dimension. A manager who responds to algorithmic anxiety primarily with skills development will apply the right intervention to the wrong problem.
Addressing algorithmic anxiety specifically requires:
Naming it. Normalising the experience by naming it directly and explaining that the uncertainty being felt is a rational response to genuine ambiguity, not a sign of inadequate resilience.
Making the human contribution explicit. Identifying and articulating the specific dimensions of professional judgment, relational skill, ethical reasoning, and contextual understanding that AI systems do not replicate, and connecting people's work directly to those dimensions.
Designing deliberate human moments. In workflows becoming increasingly automated, creating explicit points of human judgment, connection, and creative contribution is a performance management decision with direct implications for anxiety levels, engagement, and the quality of human oversight of AI outputs.
Treating psychological safety as the primary metric of AI adoption success. Teams with high psychological safety will surface algorithmic anxiety early, when it is manageable. Teams with low psychological safety will conceal it, where it accumulates and eventually manifests as disengagement, attrition, or the shadow AI usage that Cornerstone's 2025 research documented in 80% of workers surveyed, people using AI tools without disclosing this to their managers.
The managers who navigate AI adoption most effectively are not those who move the fastest. They are those who maintain the psychological safety and role clarity that allow their teams to engage with AI honestly, curiously, and without the weight of unaddressed existential doubt.
References
Bao, Y. et al. (2025) 'The impact of AI adoption on employee well-being', Nature Human Behaviour, 9(2), pp. 312-324.
Brod, C. (1984) Technostress: The Human Cost of the Computer Revolution. Reading, MA: Addison-Wesley.
Cornerstone OnDemand (2025) Talent Mobility Report: The Hidden AI Workforce. Santa Monica, CA: Cornerstone.
Tarafdar, M. et al. (2019) 'The technostress trifecta: Techno eustress, techno distress and design', Information Systems Journal, 29(1), pp. 6-42.