Last week, I found myself staring at my screen at 11 PM, exhausted but unable to stop. I’d been using an AI coding assistant all day, cranking out features faster than ever. My pull request count was up, my code quality seemed solid, but I felt completely drained.
Here’s the thing: I wasn’t working less. I was working more—way more. And I’m not alone.
The Paradox of AI Productivity
There’s a pervasive narrative in tech right now: AI will handle the drudgery, freeing us up for “high-value work.” Draft emails in seconds, debug code instantly, summarize documents automatically. Sounds great, right?
But a fascinating study from Berkeley Haas School of Business tells a different story. Researchers Aruna Ranganathan and Xingqi Maggie Ye tracked 200 employees at a tech company for eight months. They found that AI wasn’t reducing work—it was intensifying it.
I’ve been seeing this play out in my own work, and it matches exactly what they observed.
The “AI Partner” Trap
The study found that AI created a new work rhythm: employees would manage multiple active threads simultaneously. You’re manually writing code while an AI generates an alternative version. Running multiple agents in parallel. Reviving long-deferred tasks because, hey, AI can “handle them” in the background.
This feels incredible at first. You’ve got a “partner” helping you blast through your workload. There’s this palpable sense of momentum—I’m getting so much done!
But here’s what’s actually happening:
- Continual attention switching - You’re constantly context-switching between your own work and AI outputs
- Frequent checking - You’re mentally half-present, always glancing to see if the AI is done
- Growing task count - Because things feel easy, you keep taking on more
- Cognitive overload - You’re juggling way more mental threads than before
My Experience: The “Just One More Prompt” Cycle
I’ve talked to several developers who are losing sleep over this. It’s that feeling that building another feature is just “one more prompt” away. You’re in the flow, the AI is churning out code, and stopping feels like leaving potential on the table.
Here’s what my day used to look like:
- Morning: Plan tasks, write code methodically
- Lunch: Actually take a break
- Afternoon: Continue coding, wrap up by 6 PM
- Evening: Disconnect, recharge
Now it’s more like:
- Morning: Start coding with AI, see a productivity boost
- Lunch: Skip it, the AI is still working
- Afternoon: Launch multiple AI agents for different tasks
- Evening: Check results, refine prompts, launch more agents
- Late night: “Just one more feature…”
The Productivity Illusion
The Berkeley researchers point out something crucial: organizations are struggling to distinguish between genuine productivity gains and unsustainable intensity.
I’ve fallen into this trap myself. My metrics look great—more features shipped, fewer bugs reported. But I’m operating at 120% mental capacity every day. That’s not sustainable.
Think about it this way: If you’re running 5 miles in 20 minutes instead of 30, you’re faster, but you’re also exhausted. Run that pace every day for months, and you’ll burn out.
The Jevons Paradox Strikes Again
This reminds me of an economic concept called the Jevons Paradox: when technological progress increases efficiency with which a resource is used, the total consumption of that resource increases rather than decreases.
AI is the ultimate efficiency tool for cognitive work. And just like with coal in the 19th century, we’re using it to do more, not to work less.
The difference is that coal doesn’t have a breaking point. Our brains do.
What Actually Helps
I don’t think the answer is to abandon AI. These tools are genuinely useful. But we need to be intentional about how we use them.
Here’s what’s been working for me:
1. Set AI Time Boxes
I schedule specific blocks for AI-assisted work (e.g., 9 AM - 11 AM). Outside those blocks, I code manually or focus on tasks that don’t involve AI. This prevents the constant context-switching.
2. Limit Parallel Threads
I used to run 3-4 AI agents simultaneously. Now I run one at a time, max. The productivity boost from parallel work isn’t worth the cognitive cost.
3. Schedule “No AI” Days
One day a week, I don’t use AI at all. It sounds retrograde, but it’s actually incredibly refreshing. I reconnect with the satisfaction of solving problems through pure effort.
4. Track Energy, Not Just Output
I’ve started tracking how I feel at the end of each day on a scale of 1-10. If I’m consistently below 6, I know I’m overusing AI and need to dial back—even if my metrics look great.
5. Create “Done” Criteria
Before starting AI-assisted work, I define what “done” looks like. No expanding scope just because it’s easy.
The Organizational Challenge
The Berkeley study recommends building an “AI practice” that structures how AI is used. This is smart.
Here’s what I’d suggest for teams:
- Document when AI should and shouldn’t be used
- Set expectations around response time (AI doesn’t always mean instant)
- Create guidelines for parallel AI work
- Recognize that “more features” isn’t always better if the cost is burnout
The Hard Truth
I’ve been writing about tech for years now, and I’m usually the one evangelizing new tools. But with AI, I think we need to be more thoughtful.
The promise of AI reducing our workload is compelling. But the reality, at least so far, is that it’s giving us superpowers we don’t know how to use sustainably.
We’ve disrupted decades of intuition about sustainable work practices. It’s going to take discipline to find a new balance.
My Take
I still use AI every day. It helps me write better code, research faster, and iterate more quickly. But I’m much more conscious now about the hidden cost.
The goal isn’t to maximize output—it’s to maximize sustainable output. And if your AI tools are pushing you into unsustainable patterns, you’re not actually being more productive. You’re just working harder.
Maybe it’s time to stop asking “How can I do more with AI?” and start asking “How can AI help me work in a way that doesn’t burn me out?”
That’s the question I’m trying to answer now.
What about you? Have you noticed AI making your work more intense? I’d love to hear your experience.