90 percent of executives see no productivity gain from AI — yet the job cuts continue
A survey by the Federal Reserve Bank of Atlanta and an analysis of millions of Glassdoor reviews point the same way: fear of job loss costs exactly the productivity AI is meant to deliver.
Illustration · AI-generated (AI IN LIFE)
At a glance
- Around 90 percent of executives surveyed report no productivity gain from AI so far
- Source: survey by the Federal Reserve Bank of Atlanta, published as an NBER working paper
- Second study: Mark Ma, University of Pittsburgh, analysis of millions of Glassdoor reviews across five years
- Finding: AI-related comments markedly more negative than average; the leading theme is fear of job loss
- Correlation: more announced AI investment went together with more announced AI-related job cuts
Around 90 percent of executives surveyed say artificial intelligence has not yet raised productivity at their company. That is the finding of a survey by the Federal Reserve Bank of Atlanta, published as a National Bureau of Economic Research working paper. The figure is striking because it does not come from sceptics but from the decision-makers themselves — at firms that mostly already use the technology.
The cuts continue regardless. That contradiction sits at the centre of a second study: Mark Ma, professor of business administration at the University of Pittsburgh, analysed millions of employer reviews on the platform Glassdoor along with thousands of company reports over a five-year period.
His finding: comments about artificial intelligence are markedly more negative than the overall tone of reviews. The dominant theme is concern about one's own job. Employees also cite a lack of training, poorly managed rollouts, and doubts about whether the tools are any good.
That produces a mechanism which undercuts the intended effect. Companies announcing more AI investment also announced more AI-related job cuts. The resulting insecurity among remaining staff in turn correlates with lower company productivity. Justifying layoffs with AI partly destroys the conditions under which AI could be productive.
Methodological caution is warranted. These are correlations drawn from survey and review data, not a controlled comparison. Productivity can also be genuinely hard to measure over short periods — with earlier general-purpose technologies, years passed before effects showed up in the statistics. That is the familiar productivity paradox.
A hard core nonetheless remains for the debate. As long as 90 percent of executives report no measurable effect, every layoff justified by AI-driven productivity gains is a claim about the future — not a description of something that has already happened.
FAQ
Does this mean AI delivers nothing?
That conclusion does not follow. The survey captures what executives report as measurable effects — not whether such effects are technically possible. Productivity effects of new general-purpose technologies typically appear in metrics only after considerable delay.
What is the productivity paradox?
The observation that major technological shifts take a long time to appear in productivity statistics. The usual reason is that companies must first rebuild their processes before a new technology takes effect — and that rebuilding itself costs time and resources.
How reliable is Glassdoor data?
These are voluntary, non-representative self-reports that tend towards extremes. Their strength lies in sample size and in change over time: shifts in tone across millions of reviews carry more weight than individual entries.
