The productivity statistics from AI adoption so far are surprisingly weak and I think the reasons why matter a lot

I am an economist. I follow the productivity data carefully. The case that AI is already transforming economic productivity is, at the aggregate level, not well supported by the numbers yet. We are not seeing it in GDP figures, in hours worked, in output per worker at the economy-wide level in the way the technology's impact on individual tasks would predict.

This is not unique to AI. The same pattern appeared with computers in the 1970s and 1980s, what economists called the productivity paradox, where transformative technology took decades to show up in aggregate productivity statistics.

The reasons matter for how we think about what is coming. The most likely explanation is not that the technology is overhyped but that adoption, organisational change, and complementary investments take much longer than the technology itself. If that is right the productivity impact is real but delayed and we are in the early part of a very long curve.

I want to discuss this with people who have thought carefully about it rather than either dismissing the technology or assuming the headline claims about productivity are already verified.