AI gets credit for a productivity surge — but could the economy be improving for other reasons entirely?

This explores whether the recent productivity acceleration really belongs to AI, or whether something else could explain it. A direct caveat first: this collection is about AI, so it has almost nothing on rival explanations such as post-pandemic economic shifts, new business formation, or energy and capital costs. What it does have is useful in a different way. It shows how much of the acceleration AI can plausibly claim, and so how much is left for other explanations.

Start with the arithmetic. One estimate puts generative AI's time savings at about 1.6% of work hours, which works out to roughly a 1.3% productivity boost. The authors present that as consistent with the 1.89-point acceleration in US productivity since ChatGPT's release Did generative AI actually boost productivity growth?. Read it the other way, though, and part of that acceleration is still unexplained even on a generous reading. A more skeptical model puts AI's total ten-year gain at no more than 0.66% How much will AI actually boost productivity over ten years?. If that model is right, AI is a small contributor, and most of the recent jump has to come from somewhere else. The gap between these two estimates is where the question really lives.

The second factor isn't a rival cause at all. It's measurement. When companies adopt a new general-purpose technology, they spend heavily on things statistics don't count well: retraining, redesigned processes, internal software. That hidden investment first makes productivity look worse than it is, then better than it is once those assets start paying off. This pattern is called the productivity J-curve Why does AI productivity appear weak when investment is high?. Part of a measured surge could just be the upswing of earlier hidden spending, and that spending wasn't only on AI. Brynjolfsson reads the 2025 pattern of slower hiring alongside growing GDP as AI finally showing up in the data. The same note points out that other economists dispute this, because AI leaves no clear signature in jobs, productivity, or earnings outside the big tech firms Is AI productivity finally showing up in economic data?. Slower hiring raises output per worker no matter what's causing it.

Third, the collection suggests that where gains appear depends on what firms reorganize, not just which tools they adopt. Executives report bigger AI gains than the numbers show, and the effects cluster in high-skill services and finance, where workers are moving between roles rather than being cut Do AI productivity gains feel larger than they actually measure?. AI's real returns may depend on redesigning how decisions are coordinated across a company Does AI adoption require rethinking how organizations coordinate?. Country-level forecasts also show that gains depend on which industries an economy is built around How much will AI boost productivity across G7 nations?. So a productivity jump could reflect workers moving into higher-output sectors, or firms restructuring, with AI as only one ingredient.

The idea worth taking away: "is it AI or something else?" is partly a question about accounting and timing, not only about causes. Hidden investment, shifts in hiring, and differences between industries can all produce a productivity surge that looks like AI. If you want the non-AI explanations themselves, you'll need sources outside this collection.