AI Efficiency Faces Verification Tax From Output Checks

Artificial intelligence tools have not yet delivered broad productivity gains for many businesses, as companies using AI continue spending significant effort verifying outputs, according to recent research from the Federal Reserve Bank of Atlanta, Stanford University, the Bank of England and the Autonomous Technological Institute of Mexico.

A survey of nearly 6,000 senior business executives found about 90 percent reported that AI had no impact on employment or productivity, while 69 percent said their firms actively used the technology, according to the researchers.

The lack of measurable efficiency gains has also appeared in wider economic data. Torsten Slok, chief economist at Apollo, said reviews of San Francisco Fed data showed that expected AI productivity improvements remain unconfirmed.

"The productivity payoff from A.I. remains a forecast rather than an observation," said Torsten Slok, chief economist at Apollo.

Researchers and economists have pointed to several possible reasons for the gap between expectations and results, including the time needed for companies to adjust workflows and deploy new systems effectively.

A growing challenge is the additional work created by AI systems themselves. Businesses must often review generated content, verify accuracy, assess safety and make corrections before using the results.

This verification workload has been described by researchers and technology observers as a verification tax, botsitting or rework. The process can limit the automation benefits that AI tools are expected to provide.

The survey findings indicate that while AI adoption is widespread among companies, measurable effects on productivity and employment have not yet appeared at the scale predicted by many technology companies.