Generative AI spread faster than the PC or the internet did — but does faster adoption actually mean faster productivity gains?

This explores whether the record-fast spread of generative AI means productivity gains are already here or close behind. The short answer from the corpus: adoption speed and productivity gains move on different clocks. The adoption numbers really are striking. Three years after ChatGPT launched, 54.6% of US adults were using generative AI, compared with 19.7% for the PC and 30.1% for the internet at the same age Is generative AI adoption now outpacing the personal computer?. An earlier wave of the same survey has a useful detail, though. Use at work at first only kept pace with the PC. The headline lead came from personal use, which was driven by free access and almost no technical barriers How fast is generative AI adoption compared to earlier technologies?. Adoption is fastest where it costs least and changes least.

So what has that use produced so far? The same researchers estimate that generative AI saves about 1.6% of work hours, which works out to roughly a 1.3% rise in labor productivity Did generative AI actually boost productivity growth?. That's real, but modest next to how fast people took up the tools. The OECD's forward-looking estimate is in a similar range: 0.2 to 1.3 percentage points of extra annual productivity growth across G7 economies. The size depends on adoption speed, but even more on whether a country's economy is concentrated in sectors AI actually affects How much will AI boost productivity across G7 nations?. Adoption speed is one input. What gets adopted, and where, matters as much.

The less obvious lesson is the J-curve. Brynjolfsson, Rock and Syverson argue that general-purpose technologies require large hidden investments: retraining staff, redesigning processes, building custom software. Official statistics don't count these as output, so measured productivity first looks too low and later looks too high once those investments pay off Why does AI productivity appear weak when investment is high?. Census data on US manufacturers suggests the dip isn't only a measurement problem. Firms that adopted industrial AI saw productivity fall about 1.33% at first, with real disruption to inventory, robots and staffing. Gains came later, mostly at younger, well-managed firms Does industrial AI adoption hurt productivity at first?. Agrawal, Gans and Goldfarb explain why. AI makes individual tasks better but also makes decisions more variable, so organizations have to rethink how their parts coordinate before the gains add up Does AI adoption require rethinking how organizations coordinate?.

Fast adoption can also make gains feel bigger than they are. Executives report larger productivity improvements than the measurements show, partly because better operations take time to show up as revenue Do AI productivity gains feel larger than they actually measure?. And the gains aren't spread evenly. Firms that are already heavily exposed to AI replace outside labor with it faster and more cheaply, which points to returns from building internal capability rather than smooth, even diffusion Do firms substitute labor for AI at different rates?.

Here's what you might not have expected. How many people use a tool is a poor guide to how much the economy gets out of it. The gains depend on the slower, costlier work of reorganizing around the tool, and that work happens unevenly across firms, sectors and countries. When large gains do arrive, they may not show up in wages. Anthropic's scenarios suggest that faster AI growth can shift income toward capital while knowledge-worker pay stalls Does AI growth inevitably shift wealth away from workers?. Productivity and broadly shared prosperity are separate questions.