When Artificial Intelligence tools entered mainstream workplace use, many expected a straightforward outcome: workers would save time and enjoy lighter workloads. Research from ActivTrak analyzing digital activity across more than 10,000 workers tells a different story. Early AI adopters are working harder, not easier. Their time on email, messaging, and chat applications more than doubled after adopting AI. Use of business software rose by 94 percent. The pattern suggests that as AI handles routine cognitive tasks, workers are not stepping back-they are accelerating.
This inversion of the expected labor-saving outcome mirrors a historical pattern. When air travel became faster, people did not work fewer hours; they took more trips. When automobiles reduced commute time, suburbs sprawled and commutes grew longer. The technology itself saves time, but human behavior reallocates that saved time toward increased activity. Workers who adopted AI are squeezing in work bursts during evenings, weekends, and waiting rooms, and multitasking more intensely as they supervise multiple AI systems simultaneously.
The shift has psychological costs. ActivTrak researchers documented a 9 percent decline in time spent on focused, uninterrupted work. Workers and their managers now have raised expectations about daily output. The result is a mental state researchers have begun calling “AI brain fry”-a constant cognitive load that feels crowded and frazzled, even as individual tasks become technically easier to execute.
Cognitive Effort Becomes The New Currency
The UC Berkeley Haas School of Business identified a deeper mechanism. When coding and engineering tasks become easier through AI assistance, workers do not hand those tasks off. Instead, they reclaim work they had previously outsourced to specialized contractors or other departments. AI does not eliminate difficult cognitive work-it lowers the friction of attempting it. This reshuffles the division of labor within organizations and places continuous learning and skill adaptation on individual workers.
In manufacturing, the dynamic plays out differently. Automakers are deploying humanoid robots and advanced cobots designed to handle monotonous, ergonomically demanding, or dangerous tasks. BMW has signaled plans to use AEON humanoid robots for work that poses physical risk or repetitive strain. General Motors has intensified automation at its Factory Zero plant in Detroit, deploying collaborative robots that work directly alongside human employees. The automation targets the kind of labor that does not require judgment or learning-the work that should theoretically be most replaceable. Yet labor dynamics in automotive plants show contested outcomes: unions resist certain automation deployments while acknowledging that some manual reduction frees workers for other roles.
What differentiates workers in an AI-augmented economy is not raw intelligence but relationship to cognitive effort itself. Psychologists describe this as “need for cognition”-the degree to which individuals find mental exertion intrinsically rewarding. Those with a high need for cognition enjoy difficult problems and dense reading. Those at the other end, called cognitive misers, prefer tasks requiring minimal mental engagement. The middle segment will expend effort when motivated by external stakes. In a labor market where AI handles routine cognition, the cognitive miser’s advantage erodes. Effort itself becomes scarce and valuable.
Surveillance And Consent In Educational AI
The expansion of AI systems into schools reveals a separate but related problem: consent and equity. The NAACP Legal Defense and Educational Fund submitted testimony to the New York City Council opposing the expansion of AI-powered surveillance and policing in public schools. The LDF warned that facial recognition, predictive policing software, and automated decision-making tools carry significant racial bias risks. Black students are more than four times as likely as white students to attend schools with the highest surveillance levels. Expanding AI monitoring could intensify racial disparities in school discipline and widen achievement gaps.
The LDF’s concern centers on a structural inequality: students have no choice in whether their data is collected or how AI systems process it. Unlike workplace AI adoption, which involves voluntary or negotiated implementation, school surveillance is imposed on minors and their families. The LDF urged NYC to redirect resources toward restorative justice, mental health services, and educational support rather than monitoring infrastructure. The distinction matters because it highlights a limit to the cognitive-effort framework: not all populations experience AI as a choice about how to work harder. Some experience it as a system of control with unequal consequences.
AI governance requires clear authority structures and escalation procedures to prevent systems from operating without meaningful oversight. This principle applies to schools as forcefully as to federal agencies. When humans are present in AI governance but lack defined power to stop or redirect the system, the human presence becomes a compliance layer rather than genuine control.
The Shape Of Work Is Changing Without Blueprint
The broader implication is that AI adoption is not simply intensifying existing work patterns. It is restructuring what work means and who can sustain it. Workers with appetite for continuous learning and cognitive challenge find new leverage. Those expecting routine tasks or lighter schedules face pressure to do more. Organizations deploying AI in schools without clear equity frameworks risk automating discrimination. Manufacturing plants can reduce physical strain, but only if labor displacement is managed through negotiation and retraining, not directive.
The research on AI adoption so far offers no evidence that intelligence abundance produces leisure or reduced obligation. Instead, it produces intensity, specialization, and sharper divisions between workers who thrive on cognitive engagement and those who do not. That outcome is not inevitable. But it requires deliberate choices about how organizations structure AI integration, which workers gain access to cognitive leverage, and which decisions AI systems are permitted to make without human override. The technology itself is not the constraint. Human choice is.
