Study Finds AI Tools Increase Employee Workload and Burnout Despite Efficiency Claims

The pitch was irresistible: deploy artificial intelligence across your workforce and watch productivity soar while employees finally get relief from the grind. Billions have been spent on this promise. But a cascade of new research is delivering a starkly different verdict — AI tools are making knowledge workers busier, more exhausted, and more likely to quit, even as they complete more tasks than ever before.

The findings, drawn from ethnographic studies, large-scale surveys, and randomized controlled trials published between late 2025 and early 2026, paint a picture of a technology that has outrun the organizational structures meant to contain it. The result is not the leisure dividend that futurists predicted, but something researchers at UC Berkeley and Boston Consulting Group have begun calling "AI brain fry" [1][3].

The Berkeley Study: Eight Months Inside an AI-Powered Workplace

The most granular evidence comes from an eight-month ethnographic study by UC Berkeley Haas School of Business researchers Aruna Ranganathan and Xingqi Maggie Ye, published in the Harvard Business Review in February 2026. They embedded themselves in a 200-person U.S. technology company, conducting 40 in-depth interviews across engineering, product, design, research, and operations departments [1][2].

Their central finding upends the conventional narrative: AI does not reduce work. It intensifies it.

The researchers identified three distinct mechanisms of intensification. First, employees expanded the scope of what counted as "their job." Tasks that would previously have been delegated to a colleague or simply left undone suddenly felt achievable with an AI assistant. A product manager began writing code prototypes. A designer started drafting marketing copy. The boundary of each role blurred outward [1].

Second, work seeped into time that had previously served as rest. Because generative AI makes it trivially easy to start a task — fire off a prompt, review an output — employees began working during lunch, before meetings, and in the evenings. The friction that once created natural stopping points vanished [2].

Third, workers began running multiple AI-assisted processes simultaneously, keeping several threads alive at once — reviewing AI-generated code in one tab while an AI agent drafted a document in another. Some ran multiple AI agents at the same time. The researchers described employees who felt "busier, more stretched, or less able to fully disconnect," noting that the cumulative strain of intensification can build even when individual moments feel productive [1][2].

"When we talk about building an 'AI practice,' we mean being intentional about the rhythm and boundaries of AI-enabled work rather than simply accelerating because the technology makes it possible," the researchers wrote [1].

"AI Brain Fry": The Cognitive Cost of Oversight

A separate study, published in the Harvard Business Review in March 2026 by researchers affiliated with Boston Consulting Group, gave the phenomenon a name: "AI brain fry" — defined as "mental fatigue from excessive use or oversight of AI tools beyond one's cognitive capacity" [3][4].

The BCG team surveyed 1,488 full-time U.S. workers and found that 14% reported experiencing AI brain fry. Participants described a "buzzing" sensation, mental fog, difficulty focusing, slower decision-making, and headaches [3].

The numbers behind the phenomenon are striking. Workers who reported excessive AI oversight spent 14% more mental energy in the workplace, were 12% more mentally fatigued, and were 19% more likely to report suffering from information overload compared to peers with lower AI usage. Most alarmingly, self-reported error rates among those who felt they were using too much AI were 39% higher [3][4].

The study identified a critical threshold: when workers deployed one or two AI agents, throughput increased and reported fatigue fell. But adding a third platform began eroding gains. By the time workers used four or more AI tools, self-reported productivity plummeted [3].

The impact was not evenly distributed across industries. AI brain fry was most prevalent among marketing professionals (26%), followed by HR workers (19%) and software engineers (18%) [4].

Source: Boston Consulting Group / Harvard Business Review

Data as of Mar 10, 2026CSV

The Expectations Gap: C-Suite Optimism vs. Employee Reality

The disconnect between executive expectations and worker experience may be the most important finding to emerge from this body of research. An Upwork study found that 96% of C-suite leaders expect AI to increase overall productivity, and 81% acknowledge having increased demands on their workers in the past year [5][6].

Yet the employee perspective tells a different story. According to Upwork's research, 77% of employees using AI say the tools have added to their workload. Nearly half — 47% — report not knowing how to achieve the productivity gains their employers expect. And 71% of full-time employees report being burned out, with 65% struggling with their employer's demands on their productivity [5].

The most paradoxical finding: workers who report the highest productivity gains from AI are also the most burned out. Upwork's 2025 follow-up research found that 88% of workers who described themselves as highly productive with AI also reported experiencing burnout, and they were twice as likely to consider quitting compared to less productive AI users [6].

A February 2026 ResumeTemplates.com survey of over 1,000 workers corroborated these findings. Thirty-one percent of workers said their workload increased after AI was introduced, with some reporting expectations of two to four times as much output. Sixty percent said managers expect or require them to use AI, and 45% said managers explicitly reference AI when assigning additional work. More than a third (37%) reported experiencing "AI fatigue" [7].

"After years of layoffs and reorganizations, many teams are operating with fewer people," said Julia Toothacre, Chief Career Strategist at ResumeTemplates. "AI can help manage that load, but employees should watch for workload creep — when managers assume AI means you can take on more without providing the resources to support it" [7].

The Productivity Paradox Returns

The aggregate data is perhaps the most sobering element of the story. A study published by the National Bureau of Economic Research examined 6,000 CEOs and executives from firms in the United States, United Kingdom, Germany, and Australia. Despite AI adoption rising from 61% in early 2025 to 71% by January 2026, the vast majority reported little measurable impact on their operations [8].

Economists have begun invoking Robert Solow's famous 1987 observation: "You can see the computer age everywhere but in the productivity statistics." The same paradox — investment in transformative technology yielding no visible aggregate productivity gains — appears to be repeating itself with AI [8].

Source: Bureau of Labor Statistics (CES0500000003)

Data as of Mar 12, 2026CSV

A key bottleneck has been identified. Workday's 2026 research found that 37–40% of time supposedly saved by AI gets consumed by reviewing, correcting, and verifying AI-generated output [9]. In software development, the pattern is even more pronounced. While developers on high-AI-adoption teams complete 21% more tasks and merge 98% more pull requests, pull request review time increases by 91%, creating a bottleneck that erodes the speed gains [9].

The most controlled evidence comes from a randomized trial by METR, a research organization, which studied 16 experienced open-source developers completing 246 tasks. Developers predicted AI would reduce task completion time by 24%. The actual result: AI increased completion time by 19%. The researchers attributed the slowdown to time spent prompting, reviewing AI-generated suggestions, and integrating outputs with complex codebases [10].

The Human Cost: Burnout by the Numbers

The burnout crisis extends well beyond technology companies. Upwork's data shows that burnout rates are higher among women (74% vs. 68% for men) and dramatically higher among younger workers, with 83% of Gen Z workers reporting burnout compared to 73% of millennials [5]. One in three employees say they will likely quit within six months due to burnout or being overworked [5].

The BCG study linked excessive AI use to concrete organizational risks: increased employee errors, decision fatigue, and greater intention to quit. Workers experiencing AI brain fry were not merely uncomfortable — they were producing lower-quality work and planning their exit [3][4].

These findings arrive during a period of elevated labor market uncertainty. The U.S. unemployment rate stood at 4.4% in February 2026, while average hourly earnings for private-sector workers rose to $37.32, reflecting continued upward pressure on wages even as companies deploy AI tools aimed at doing more with fewer people [11].

Why the Promise Falls Short

Several structural factors help explain why AI's workplace benefits have not materialized as expected.

The verification bottleneck. AI can generate output at extraordinary speed, but someone must still check it. As the volume of AI-generated material increases, the cognitive burden of quality assurance falls on the same workers who were supposed to be liberated. The BCG study's finding of 39% higher error rates among heavy AI users suggests this oversight is failing [3].

Scope creep by default. The Berkeley study documented how AI tools expand the perceived scope of every role. When an AI assistant makes a previously impossible task merely difficult, employees absorb it. The total volume of work grows even if individual tasks take less time [1].

Inadequate training and unclear norms. The ResumeTemplates survey found that nearly half of AI-using employees do not know how to achieve the productivity gains expected of them. Organizations are deploying powerful tools without establishing practices around their use [7].

The always-on effect. Because AI tools work asynchronously and are available around the clock, they erode the boundaries between work and rest. The Berkeley researchers documented employees sending prompts during evenings and weekends — not because they were required to, but because the technology made it frictionless [1][2].

What Organizations Can Do

The research does not argue that AI is inherently harmful. The Berkeley team notes that intensification can feel positive in individual moments. The BCG study found genuine productivity gains at low levels of AI usage. The problem is not the technology itself but the absence of institutional guardrails.

Ranganathan and Ye propose the concept of an "AI practice" — deliberate organizational policies that govern the rhythm and boundaries of AI-enabled work. This might include limits on the number of AI tools employees are expected to manage simultaneously, protected time blocks free from AI-assisted multitasking, and explicit norms against after-hours AI use [1].

The BCG researchers recommend that organizations monitor AI tool proliferation carefully, given the evidence that productivity declines sharply when workers use four or more tools. They advocate for measuring not just output but cognitive load, error rates, and employee wellbeing [3].

At a broader level, the findings suggest that companies need to resist the temptation to treat AI-driven efficiency gains as a reason to reduce headcount or increase individual workloads. The Upwork data shows that the workers absorbing the most AI-amplified work are the ones closest to leaving [5][6].

A Technology Without a Practice

The emerging research consensus is not anti-AI. It is a warning that the technology has been adopted faster than the management practices needed to deploy it sustainably. Billions of dollars in enterprise AI investment are producing a workforce that is simultaneously more capable and more fragile — able to handle an unprecedented volume of tasks while burning through the cognitive and emotional reserves needed to sustain that pace.

The parallel to Solow's paradox is instructive. When personal computers arrived in the 1980s, it took more than a decade for organizations to restructure workflows in ways that actually captured productivity gains [8]. AI may follow the same trajectory — but only if the current wave of burnout does not drive away the workers who are supposed to lead that transformation.

The data makes one thing clear: the tools designed to make work easier are, for now, making it harder. Whether that changes depends less on the next model release and more on whether organizations can build the human infrastructure — the norms, the boundaries, the practices — to match the technology they have already deployed.