AI Speeds Work—And Swells It

A field study of a 200-person tech company shows that generative AI doesn’t reduce work; it intensifies it through task expansion, blurred boundaries, and more multitasking. While initial productivity appears to rise, hidden oversight work and cognitive load grow, risking burnout and poorer decisions. The remedy is an intentional “AI practice” with pauses, sequencing, and human grounding to keep productivity sustainable.

Key Points

Generative AI intensifies work by accelerating pace, widening job scope, and extending working hours—even without managerial pressure.
Task expansion shifts responsibilities across roles and creates hidden oversight work, especially for engineers reviewing AI-assisted output.
AI blurs work–life boundaries as micro-prompts during breaks and off-hours erode natural recovery time and make work feel always-on.
Parallel use of AI (multiple agents/threads) increases multitasking and context switching, raising cognitive load and pressure for speed.
Leaders should institute an “AI practice” with intentional pauses, sequencing of workflows, and human grounding to make productivity gains sustainable.

Sentiment

The community overwhelmingly agrees with the article. Most commenters validate from personal experience that AI intensifies rather than reduces work, framing it through well-known economic concepts like Jevons Paradox and the Red Queen's Race. The pessimistic labor economics perspective dominates, with widespread skepticism that productivity gains will benefit workers. Even those reporting positive AI experiences often acknowledge working more hours. The tone is resigned rather than hostile — commenters see this as a predictable consequence of how productivity tools operate under capitalism.

In Agreement

Jevons Paradox applies to AI-assisted labor: efficiency gains lead to more work as expectations and scope expand, not less
AI removes the cognitive buffer of routine work, leaving only high-intensity problem-solving tasks that accelerate burnout
Workers are trapped in a prisoner's dilemma — not using AI puts you at a disadvantage, but using it intensifies your work
Productivity gains flow to capital owners, not workers — nobody is getting raises or working fewer hours because of AI
AI enables task expansion beyond one's expertise, creating dangerous dependency cycles and increased impostor syndrome
Monitoring AI agents is more cognitively fatiguing than doing work manually, similar to Level 3 autonomous driving
Speed of accountability is the real bottleneck — responsible developers can only produce code as fast as they can verify it
Companies will not voluntarily reduce hours or set reasonable AI usage norms — they will demand more output

Opposed

AI genuinely unlocks new capabilities for solo developers and those in niche domains, enabling previously impossible work
Offloading rote coding to AI frees experienced developers to focus on higher-level architecture and design thinking
The study uses an old-world lens — the steady state is labor replacement, and remaining humans being overloaded is transitional
Effective LLM use is a rare skill that does translate to better outcomes for those who master it
AI models are rapidly improving with capabilities like visual reasoning that create step-change improvements prior studies cannot account for