Productivity has been the quiet engine of living standards for a century. When it stalls, wages stagnate. When it accelerates, everything shifts — pay, prices, the kinds of jobs that exist. A new analysis from Anthropic’s Economic Index suggests the engine might be about to rev.
The researchers examined millions of real interactions with Anthropic’s AI assistant Claude. They wanted to know what AI actually does at work, not what it could theoretically do in a lab. Their central finding: current AI systems can already handle about a quarter of the tasks in the U.S. economy. Roughly 44 to 49 percent of jobs have some portion of their work accessible to these tools.
But here is the concrete stakes number. The report estimates that if widely adopted, current AI could raise labor productivity growth by roughly one to nearly two percentage points per year. That would potentially double recent trends. Doubling productivity growth is not an abstraction. It means the same number of workers produce more value in the same time. That is how economies grow without inflation. That is how real wages rise without employers taking a hit.
The catch is adoption. Most of those capabilities remain unused. There is a large gap between what the technology can do and what people actually ask it to do. The authors suggest the bigger transformation may come not from new breakthroughs but from people gradually using the tools that already exist. That is a different kind of story than the usual AI narrative of sudden disruption. It is slower. It is messier. It is also more likely.
When AI is used, it tends to augment people rather than replace them. That is the pattern in the data so far. The analysis does not predict that pattern will hold forever, but it is what the real-world interactions show. Augmentation means a worker keeps their job but does it differently — faster, with less grunt work, with more time for judgment calls. Replacement means the worker is gone. The difference matters for every person who clocks in.
There is a temptation to read the headline — a quarter of tasks accessible — and brace for job losses. The data suggests a different risk. The risk is that businesses do not adopt the tools, productivity stays flat, and the potential gain evaporates. Or that adoption happens unevenly, widening the gap between firms that figure it out and firms that do not. Or that the augmentation pattern shifts into replacement as the technology improves and companies reorganize work around it.
The analysis examined millions of interactions. That is a large sample. It is also a single source — interactions with one company’s AI assistant. The patterns may hold broadly, or they may reflect the specific users who choose Claude. The authors are careful about that limit. But the size of the data gives weight to their conclusions.
Productivity growth in the U.S. has averaged around one to one and a half percent per year in recent decades. Adding another point or two on top of that is a structural shift. It changes the math on everything from federal debt to retirement savings to the number of nurses a hospital can afford to hire. That is what is at stake in whether businesses actually use the tools they already have.
The study does not say the future is certain. It says the capacity is already here. The open question is whether the workforce will feel the effects as a gentle lift or a hard shove. The answer depends on how fast adoption spreads and whether the augmentation pattern holds. Those are choices, not inevitabilities.
