Humans limit AI-driven productivity gains
In brief
- AI investments fail to boost productivity outside the tech sector, per Apollo Global Management.
- « Botsitting » (verifying AI results) consumes part of the productivity gains.
- MIT-Wharton and Harvard studies show increased code production without real impact on deliverables.
Apollo Global Management, a major asset manager and AI infrastructure investor, warns that massive AI investments have yet to deliver tangible productivity gains for U.S. companies outside the tech sector. According to its chief economist Torsten Slok, this shortfall stems from the nature of economic activity, where human judgment and contextual factors (responsibility, culture, ethics) remain essential.
Slok’s analysis relies on several indicators: revenue per employee rises only for the S&P 500’s « Magnificent Seven » (Alphabet, Amazon, Apple, Meta, Microsoft, NVIDIA, Tesla), while margins and total factor productivity stagnate elsewhere. He also highlights the concentration of AI spending and its circular nature, as well as the time needed for internal process adaptation.
The « botsitting » phenomenon — verifying AI-generated outputs — exemplifies this bottleneck. Recent studies (MIT-Wharton, Harvard) confirm that despite increased code production or ticket resolution, real productivity gains remain limited due to validation and human integration needs. Consumer attention and the expertise required for these tasks also curb AI’s macroeconomic impact.
What it changes for you
- Finance & BankingProductivity gains from AI remain marginal in the financial sector, where complex processes and human verification limit benefits.
This summary gives you the essentials. The full article, its figures and quotes are on Next.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error