Illustration by Tag Hartman-Simkins/Futurism. Source: Getty Images
Even if AI is – or eventually becomes – an incredible automation tool, will it make workers’ lives easier? That’s the big question explored in an ongoing study by researchers at UC Berkeley’s Haas School of Business. And so far, it’s not looking good for the rank and file.
one in piece for Harvard Business ReviewAruna Ranganathan and Xinqi Maggie Ye of the research team reported that after closely monitoring a technology company with two hundred employees for eight months, they found that AI actually made the work they had to do faster, rather than reducing it.
This “overload”, in which employees take on more work than is sustainable for them, can create a vicious cycle that leads to fatigue, burnout, and reduced work quality.
“You thought that maybe, oh, because you can be more productive with AI, you might save some time, you might do less work,” one employee told the researchers. “But in reality, you don’t work less. You just work the same amount or even more.”
The tech company in the study provided AI tools to its employees, but did not mandate that they use them. Adoption was voluntary. The researchers described how many employees, on their own initiative, first experimented with AI tools out of curiosity, “because AI made ‘doing more’ feel possible, accessible, and in many cases intrinsically rewarding.” This resulted in some employees increasingly taking on tasks they normally would have outsourced, or hiring additional help to cover it, the researchers said.
One consequence is that once the novelty of adopting AI wears off, employees realize they have added more than they can handle. But other effects were visible in the broader workplace, too. For example, engineers had to spend more time correcting AI-generated code passed by their colleagues. AI also gave rise to more multitasking, with some people choosing to write code manually while an AI agent, or even multiple AI agents, cranked out their own version in the background. Instead of focusing on one task, they were constantly switching their attention, giving the feeling that they were “always on task,” the researchers said.
Others realized that AI has slowly managed to infiltrate their free time, with employees prompting their AI tools during lunch breaks, meetings, or just before walking away from their PCs. This blurred the line between work and non-work, the researchers wrote, with some workers describing their downtime as no longer feeling refreshing.
In short, AI tools created a vicious cycle: It “accelerated some tasks, which increased expectations of speed; higher speeds made workers more dependent on AI. The increased dependency increased the scope of effort workers must perform, and the broader scope further increased the volume and density of work.”
The Berkeley Haas team’s findings add to a growing body of evidence that runs counter to the AI industry’s promises that its tools will bring productivity miracles.
Most companies adopting AI saw no meaningful increase in revenue MIT study found. Other research has shown that AI agents often fail at common remote work and office tasks. And at least one study has documented how employees used AI to generate shoddy “workshops” that their coworkers had to fix — in contrast to the engineers forced to correct their vibe-coding colleagues in the Berkeley Haas study — creating resentment and reducing productivity. Employees remain ambivalent about the technology, with a recent survey showing that 40 percent of white-collar workers not in management roles thought AI did not save them time at work.
The Berkeley Haas researchers optimistically suggest that companies should establish strong guidelines and provide structure on how the technology is used. But it is clear that AI can easily produce negative knock-on effects that are difficult to manage, and which we are still unpacking.
More on AI: “Novelist” claims she can produce a new book in 45 minutes using AI, says regular writers would never be able to do
