Generative AI saves only 2.8% of hours
The article contrasts generative AI's laboratory speedups of 15% to 55% with its 2.8% real-world savings in Danish workplaces, where gains had no significant effect on earnings. Furthermore, 95% of enterprise AI pilots fail to yield financial returns unless workflows are explicitly restructured to capture and monetize the saved time.
The AI productivity boom is largely an illusion on the balance sheet because companies and individuals fail to actively capture and monetize the minutes saved.
* **The Jagged Frontier trap**: AI's lack of error signaling means users spend significant time debugging confident but incorrect outputs on tasks outside its capabilities.
* **Dilution of gains**: Large task-specific speedups shrink to minor weekly gains once meetings, admin work, and human-in-the-loop validation are included.
* **Leaky productivity**: Saved time only turns into revenue if it is directly tied to billing more, serving more clients, or removing concrete costs.
DISCOVERED
1h ago
2026-07-03
PUBLISHED
5h ago
2026-07-03
RELEVANCE
AUTHOR
ermantrout