You were promised that AI would give you your time back.
Draft emails in seconds, summarize documents instantly, automate the tedious work so you can focus on what matters. The pitch was seductive, and millions of workers believed it. Companies invested billions in AI tools expecting efficiency gains that would transform the bottom line.
But something strange happened on the way to the productivity paradise. Workers who adopted AI tools didn't start leaving the office earlier. They didn't find themselves with more free time. Instead, they found themselves doing more work than ever before.
A landmark study just published in Harvard Business Review confirms what many workers have been feeling: AI doesn't reduce your workload. It intensifies it.
"You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less," one employee told the researchers. "But then really, you don't work less. You just work the same amount or even more."
This isn't an isolated finding. A 2024 Upwork study of 2,500 workers found that 77% of employees using AI said these tools had actually increased their workload. Nearly half reported they didn't know how to achieve the productivity gains their employers expected. The gap between the C-suite's expectations and workers' reality has become a chasm.
The Workload Creep Nobody Saw Coming
The Berkeley researchers identified a pattern they call "workload creep." It's subtle, almost invisible at first, but it fundamentally changes the nature of work.
Here's how it happens. AI accelerates certain tasks, which raises expectations for speed. Higher speed makes workers more reliant on AI. Increased reliance widens the scope of what workers attempt. A wider scope further expands the quantity and density of work. It becomes a self-reinforcing cycle with no natural stopping point.
The study documented three specific mechanisms driving this intensification.
- Task expansion. Product managers started writing code. Researchers took on engineering work. Roles that once had clear boundaries blurred as workers used AI to handle the initial stages of tasks that previously sat outside their job description. If AI can help you do something, the logic goes, why not add it to your plate?
- Temporal expansion. Work started bleeding into non-work hours. The simplicity of prompting AI made the boundaries between work and personal time feel artificial. Employees found themselves experimenting with tools during breaks, on weekends, over dinner. The excitement of new technology masked what was actually happening: unpaid labor disguised as curiosity.
- Multiplication of active threads. AI introduced a new rhythm where workers managed several tasks simultaneously. Writing code manually while AI generated an alternative version. Running multiple agents in parallel. Reviving long-deferred projects because AI could "handle them" in the background. Workers felt they had a partner that could help them move through their workload faster.
But this sense of partnership came with hidden costs. The reality was continual switching of attention, frequent checking of AI outputs, and a growing number of open tasks competing for mental space. What felt like momentum was actually cognitive fragmentation.
The Supervision Tax Nobody Accounts For
There's a fundamental misunderstanding at the heart of AI adoption: the assumption that automation inherently reduces human effort.
The Berkeley study found the opposite. When organizations deploy AI tools, the immediate effect is often an increase in the volume of work that needs human attention. Employees spend more time managing, reviewing, and correcting AI outputs than they saved by delegating tasks to the technology in the first place.
This creates what might be called a "supervision tax." Every AI-generated draft needs review. Every automated output requires verification. Every delegated task demands quality control. The work doesn't disappear. It transforms into a different kind of work that's often more cognitively demanding than the original task.
This dynamic is especially insidious because it's largely invisible. Traditional productivity metrics might show that more work is getting done. Managers see output increasing. What they don't see is the human cost: the cognitive load of constant context-switching, the stress of managing multiple parallel workstreams, the exhaustion of always being "on" because AI never needs a break.
When Efficiency Becomes Exploitation
The gap between executive expectations and employee experience has become one of the defining tensions of the AI era.
But what looks like productivity from the top looks very different from the ground. Executives see more output. Workers feel more pressure. The efficiency gains accrue to the organization while the costs accrue to individuals.
This dynamic is particularly concerning because it often happens without explicit direction. The Berkeley study found that employees intensified their own work voluntarily. Nobody told them to work during evenings or take on tasks outside their role. They did it because AI made it possible, and because the implicit expectations of their workplace made it feel necessary.
The researchers describe a troubling cycle: AI acceleration raises expectations, higher expectations increase AI reliance, greater reliance expands task scope, expanded scope intensifies work. Workers get trapped in an escalating commitment to productivity that erodes the very quality of life technology was supposed to improve.
Companies thrilled by the prospect of more work for less pay should be careful. The Harvard Business Review study warns that initial productivity surges can give way to cognitive fatigue, weakened decision-making, and eventually turnover. Workers who realize their workload has grown while they were busy experimenting with ChatGPT may not stay around to see what happens next.
The Burnout Crisis Nobody's Talking About
The connection between AI adoption and burnout isn't hypothetical. It's showing up in survey after survey.
These numbers predate the Berkeley study, but they provide context for its findings. Workers were already stretched thin. AI was supposed to provide relief. Instead, it added new forms of pressure.
The Berkeley researchers emphasize that work expansion might look productive in the short term. More gets done, metrics improve. Executives pat themselves on the back for successful technology adoption. But this apparent success masks a building crisis.
Cognitive fatigue degrades the quality of human judgment, the very thing that AI cannot replace. Workers who are constantly switching between tasks, monitoring AI outputs, and managing parallel workstreams don't have the mental resources for the deep thinking that creates real value. They're busy, but they're not doing their best work.
Eventually, something breaks. The researchers warn that burnout leads to turnover, and turnover is expensive. The productivity gains from AI can be wiped out by the costs of replacing exhausted workers who've had enough.
What Companies Should Actually Do
The researchers call for organizations to develop what they term an "AI practice": intentional norms around when and how to use AI that might stem some of this workload creep.
This isn't about limiting AI use. It's about using AI thoughtfully rather than reflexively. The distinction matters because the current default, deploying AI tools and expecting productivity gains without considering human impact, isn't working.
Several practical recommendations emerge from the research.
- Organizations need to account for the supervisory and editorial labor that AI generates. This work is real, it takes time, and it should be built into workload planning. Pretending that AI simply reduces human effort leads to unrealistic expectations and invisible overwork.
- Companies should resist the temptation to use efficiency gains as justification for headcount reduction or workload expansion. When AI saves an hour, that hour shouldn't automatically fill with new tasks. Some of it should genuinely return to the worker.
- Training should go beyond how to use AI tools. Workers also need guidance on how to manage the cognitive and emotional demands of working alongside AI. This includes recognizing the signs of workload creep and developing boundaries that protect sustainable performance.
- The researchers suggest structural interventions: deliberate pauses before major decisions, sequencing work to reduce context-switching, and protecting time for actual human connection. These aren't soft suggestions. They're necessary counterweights to AI's tendency to accelerate and intensify everything.
- Organizations need honest conversations about what AI adoption is actually doing to their people. The metrics that matter aren't just output and efficiency. They include wellbeing, sustainability, and the quality of human judgment that no AI can replicate.
The Productivity Paradox We Need to Face
Technology has always promised to reduce work, and technology has always found new ways to fill the time it saves. The printing press was supposed to reduce the labor of copying manuscripts. The washing machine was supposed to free hours of domestic work. Email was supposed to make communication more efficient.
In each case, the technology delivered on its promise while simultaneously creating new forms of work. We can produce more documents, maintain cleaner homes, and send more messages than ever before. We're also busier than ever before.
AI is following the same pattern, but at accelerated speed and scale. The efficiency gains are real. So is the intensification. Both things are true simultaneously, and both need to be acknowledged.
The Berkeley study doesn't argue that AI is bad or that companies should avoid it. The researchers simply document what's actually happening when organizations embrace these tools enthusiastically without thinking carefully about human impact. The technology works. The problem is what we're doing with the gains.
For individual workers, the study offers a kind of validation. If you've felt busier since you started using AI tools, you're not imagining it. If your to-do list has grown even as individual tasks take less time, that's a documented pattern. If you've found yourself working more hours despite having more powerful tools, you're not alone.
The question isn't whether AI will continue to transform work. It will. The question is whether we'll let that transformation happen to us, or whether we'll shape it deliberately toward outcomes that serve human flourishing rather than just organizational output.
Right now, based on the evidence, we're not doing a great job. But knowing the pattern is the first step toward changing it.
Related Articles
Mark
Dani
Dani
