Workers using AI are completing tasks faster than ever. Companies are investing billions in AI tools. Yet overall economic productivity remains flat. If this sounds familiar, it's because we've seen it before.
Economist Robert Solow captured the phenomenon in 1987: "You can see the computer age everywhere but in the productivity statistics." Nearly 40 years later, we're watching the same pattern repeat with AI.
The disconnect comes down to two different ways of measuring productivity. Labor productivity tracks output per worker and has shown solid gains recently. Total Factor Productivity (TFP) measures overall economic efficiency and has been struggling since the pandemic. Workers are doing more individually without improving how efficiently the economy functions.
This mirrors what happened during the 1990s computer boom. Companies poured money into IT infrastructure. Individual workers got better tools. But productivity gains didn't show up in economic statistics for years. The benefits were real, but it took time for organizations to figure out how to actually use the new technology at scale.
A Harvard Business Review study found that workers using AI saved time on tasks but just redirected that time into more work. They put in additional hours and faced increased burnout risk without generating proportionally more output. Getting tasks done faster doesn't help if you just get assigned more tasks.
The Atlanta Federal Reserve surveyed executives and found something similar. Leaders perceived larger productivity improvements than measurable indicators like revenue actually showed. People felt more productive. The numbers didn't reflect it.
The problem is integration. Individual workers have access to better technology, but systemic workflows remain unchanged. If one person uses AI to draft reports twice as fast, but everyone else still reviews reports at the same pace, the overall process doesn't speed up. The bottleneck just moves.
Companies are making massive investments based on productivity promises that haven't materialized in aggregate data yet. That doesn't mean the promises are false. The internet eventually delivered productivity gains. But there was a significant lag between adoption and measurable impact.
The San Francisco Federal Reserve noted that determining whether sustained growth has begun "is difficult in real-time and is usually only obvious with the benefit of some hindsight." We might be in the middle of a productivity revolution without knowing it yet. Or we might be watching expensive toys get deployed without fundamental workflow changes.
