The AI Productivity Paradox: Why Getting More Done Is Leaving Us Exhausted

Here’s an uncomfortable truth that nobody warned us about: the productivity boost from AI tools is genuinely exhausting. We’re accomplishing more than ever—and burning out faster too.

The Core Insight

Research from Berkeley Haas School of Business, based on studying 200 employees at a U.S. tech company from April to December 2025, confirms what many developers are feeling intuitively: AI doesn’t reduce work. It intensifies it.

The pattern is familiar: You’re manually writing code while AI generates an alternative version. You’re running multiple agents in parallel. You’re reviving long-deferred tasks because AI can “handle them” in the background. The sense of having a “partner” creates momentum.

But here’s the trap: that momentum masks a continuous switching of attention, frequent checking of AI outputs, and a growing pile of open tasks. You feel productive while your cognitive load quietly spirals.

Simon Willison, one of the most thoughtful observers of AI development tools, captures it perfectly: “I’m frequently finding myself with work on two or three projects running in parallel. I can get so much done, but after just an hour or two my mental energy for the day feels almost entirely depleted.”

Why This Matters

The “one more prompt” addiction is real. Willison reports conversations with people losing sleep because building another feature with “just one more prompt” feels irresistible. The friction that used to pace our work—waiting for builds, context-switching costs, implementation time—has collapsed. Nothing stops us from overworking ourselves.

Decades of productivity intuition are obsolete. We had developed sustainable working practices over careers of experience. How long can you code before taking a break? How many tasks can you juggle? When does context-switching hurt more than it helps? All of that intuition was calibrated for a world without AI amplification.

Organizations can’t tell real gains from unsustainable intensity. When everyone’s output increases but so does burnout, how do you know if you’re actually more productive? The metrics we use to measure work were designed for a different era.

Key Takeaways

• AI creates the illusion of partnership, but that “partner” doesn’t manage your energy or attention
• Parallel workstreams multiply output while depleting cognitive reserves faster
• The friction that used to pace us—implementation time, context-switching costs—is gone
• Existing intuitions about sustainable work don’t transfer to AI-amplified productivity
• Organizations need new frameworks to distinguish genuine productivity from burnout-in-progress

Looking Ahead

The HBR researchers call for organizations to build an “AI practice” that structures how these tools are used. But honestly? We’re each going to need to figure out our own limits.

The old rules—take breaks, limit work hours, avoid constant context-switching—might need to become even more important, not less. The machines can work forever. We can’t.

Finding the new balance will take discipline and some painful lessons. The productivity is real. So is the exhaustion.

Based on analysis of HBR research via Simon Willison’s commentary