The Productivity Trap: Why AI Makes Us Work Harder, Not Smarter

New research reveals that AI adoption isn’t reducing workload — it’s creating a fundamentally new kind of cognitive exhaustion.

The Core Insight

The promise was seductive: AI handles the boring stuff, you focus on what matters. Draft documents? AI. Debug code? AI. Summarize meetings? AI. With all this automation, surely we’d be working less, thinking more, living better.

Reality check: A Berkeley Haas study of 200 employees at a U.S. tech company from April to December 2025 found something disturbing. AI doesn’t reduce work — it intensifies it.

Simon Willison, one of the sharpest observers of the LLM landscape, captured it perfectly: “The productivity boost these things can provide is exhausting.”

The study describes a new work rhythm where employees “managed several active threads at once: manually writing code while AI generated an alternative version, running multiple agents in parallel, or reviving long-deferred tasks because AI could ‘handle them’ in the background.”

Sound familiar? That dopamine hit of parallel productivity is real — and it’s burning people out.

Why This Matters

The “partner” illusion creates cognitive overload. Workers reported feeling like they had a “partner” helping them through their workload. This sense of collaboration enabled momentum. But the reality was “continual switching of attention, frequent checking of AI outputs, and a growing number of open tasks.”

This isn’t partnership — it’s context switching on steroids. Every AI output needs review. Every parallel task needs monitoring. The cognitive load of coordination exceeds the cognitive savings of delegation.

Productivity metrics are lying to us. When everyone’s shipping more features, closing more tickets, generating more content, it looks like a win. But are these genuine productivity gains or unsustainable intensity masked as efficiency?

The study warns that organizations struggle to “distinguish genuine productivity gains from unsustainable intensity.” We’re confusing motion for progress, output for outcome.

Decades of work-life wisdom just became obsolete. Willison hits the nail: “I think we’ve just disrupted decades of existing intuition about sustainable working practices.”

The patterns we developed for managing cognitive work — the Pomodoro technique, time-boxing, single-tasking — assume human-scale parallelism. AI enables superhuman parallelism, but our brains can’t actually process superhuman workloads. We’ve upgraded our throughput without upgrading our architecture.

Key Takeaways

• The “one more prompt” trap is real. People are losing sleep because adding just one more feature feels irresistible. AI lowers the friction of starting tasks but doesn’t lower the cognitive cost of managing them.

• Two hours of AI-augmented work can feel like a full day. Willison reports his “mental energy for the day feels almost entirely depleted” after intense AI collaboration sessions. The productivity is real, but so is the exhaustion.

• Organizations need “AI practice” frameworks. Ad-hoc adoption leads to burnout. Structured approaches to how, when, and how much to use AI are becoming necessary — like nutrition labels for cognitive consumption.

• Individual discipline isn’t enough. This is a collective action problem. If everyone’s running AI-augmented sprints, the baseline expectations shift upward. You can’t opt out without falling behind.

The Uncomfortable Truth

We built tools that make us more productive per hour, then filled those hours with more tasks. The AI productivity dividend got captured by scope creep, not quality of life.

This isn’t new — every productivity tool in history has followed this pattern. Email made communication instant, so we sent more messages. Smartphones made us reachable anywhere, so work followed us home. AI makes everything faster, so everything gets added to the backlog.

But the rate of intensification with AI is unprecedented. Previous tools incrementally increased capacity. AI is a step function. We went from “I can do this task” to “I can do this task plus three others simultaneously” overnight. Our psychological adaptation hasn’t caught up.

Looking Ahead

The researchers call for organizations to build intentional practices around AI use. This probably looks like:

• Explicit limits on parallel AI tasks (like we limit parallel threads in code)
• Mandatory “integration time” to synthesize AI outputs rather than just generating more
• Team-level norms about AI-augmented work hours (like we developed norms around after-hours email)
• Metrics that capture sustainability, not just throughput

The individual advice is familiar but newly urgent: guard your attention, resist the urge to parallelize everything, remember that your brain is the bottleneck no AI can expand.

We’ve created tools that let us work at superhuman speed. Now we need the wisdom to recognize that we’re still human.

Based on analysis of “AI Doesn’t Reduce Work—It Intensifies It” by Aruna Ranganathan and Xingqi Maggie Ye (Harvard Business Review) and commentary by Simon Willison

Tags: #AI #Productivity #Burnout #FutureOfWork #AIAdoption #WorkLifeBalance #CognitivLoad