The Exhaustion Problem: AI’s Hidden Productivity Tax

AI makes you more productive. AI also depletes you faster than ever. Both statements are true, and we haven’t figured out how to reconcile them yet.

The Core Insight

Simon Willison surfaces a finding from Berkeley Haas researchers that crystallizes a problem many AI users have felt but struggled to articulate: AI doesn’t reduce work—it intensifies it.

The study of 200 employees at a U.S. tech company (April-December 2025) revealed a pattern that will sound familiar to anyone using coding agents: AI creates “a new rhythm in which workers managed several active threads at once.” Writing code manually while AI generates alternatives. Running multiple agents in parallel. Reviving long-deferred tasks because AI could “handle them in the background.”

The productivity feels real. The exhaustion is definitely real.

Why This Matters

Here’s the trap: AI gives you a “partner” that enables momentum. But that partnership creates:
– Continual attention switching — bouncing between what you’re doing and what AI is generating
– Frequent output checking — validating, reviewing, merging AI work
– Growing open task lists — because AI makes everything possible, everything seems urgent

Willison’s personal experience is vivid: “I’m frequently finding myself with work on two or three projects running parallel. I can get so much done, but after just an hour or two my mental energy for the day feels almost entirely depleted.”

This isn’t ordinary tiredness. It’s a new category of cognitive load that our intuitions about sustainable work haven’t caught up with.

The Dopamine Loop: People are losing sleep because adding “just one more feature with just one more prompt” is irresistible. The friction that naturally limited work—the time to implement, the effort to code—has collapsed. What’s left is pure cognitive bandwidth, and we’re burning through it faster than we realize.

Key Takeaways

• Productivity gains may be unsustainable. Organizations can’t distinguish genuine efficiency improvements from workers running on fumes.
• Decades of work intuitions are disrupted. What “a good day’s work” feels like has fundamentally changed, and we don’t have new calibration yet.
• Structure is essential. The HBR researchers call for organizations to build an “AI practice” that governs how AI is used—not to limit productivity, but to prevent burnout.
• Individual discipline matters. Willison acknowledges it’s going to take “a while and some discipline to find a good new balance.”
• The “partner” framing is a trap. AI enables momentum but extracts cognitive resources you might not notice depleting.

Looking Ahead

We’re in an awkward transitional period. The tools that make us more productive are also the tools that exhaust us faster. And because the exhaustion is cognitive rather than physical, it’s harder to recognize until you’re already burned out.

The researchers suggest organizations need to build frameworks—”AI practices”—that help workers use AI sustainably. But it’s early. Most organizations don’t even understand the problem yet, let alone have solutions.

For individuals, the prescription is probably familiar to anyone who’s navigated other productivity traps: set boundaries before you need them. The unlimited potential of AI assistance is exactly why limits become essential.

The question isn’t whether AI makes you more productive in a sprint. It’s whether the intensity it creates is sustainable over months and years. We don’t know yet. But the burnout reports are starting to accumulate, and that’s a signal worth taking seriously.

Based on analysis of “AI Doesn’t Reduce Work—It Intensifies It” by Simon Willison