AI Capability Races Ahead of the Business Payoff | American Enterprise Institute

The state of the AI revolution can be confusing. At times, it can seem as if the latest artificial intelligence models are showing huge capability increases. One example is when two OpenAI models broke out of the lab, accessed the internet, and broke into another AI company. Or, as a Wall Street Journal headline put it, “The Day the Bots Broke Loose.”

But then you might come across another headline from the Wall Street Journal: “Big Companies Are Starting to Hire Again, Defying Predictions of AI Wipeout.” And from that piece:

The push to expand head count, at least modestly, is a reversal from the prevailing corporate messaging during much of the AI era. Major employers largely held back on adding people due to economic uncertainties or a belief that artificial intelligence could shoulder more tasks on the job. But some executives say the costs and limitations of AI now demand that more people be added; others want to hire people back following layoffs.

One thing that these two new stories illustrate is the gap between lab capabilities and workplace productivity. It’s a disconnect that shouldn’t be surprising to students of economic history. It takes time for companies to reorganize themselves to fully and productively integrate an important new technology into their workflows and operations.

And that remains a work in progress, even if it might be happening faster than with earlier big technological advances. As Ernie Tedeschi, chief economist at payments processor Stripe, explains in a recent analysis, if that aforementioned integration were happening in a big way, we would be seeing something that we are not currently seeing. 

Tedeschi writes:

We would expect, for example, that strong microproductivity gains would show up in estimates of total factor productivity (TFP). TFP is formally the portion of productivity growth not explained by increasing the amount of capital or labor being deployed in the economy; more informally, TFP is what economists think of as ‘pure’ technological and efficiency growth, with the important caveat that TFP cannot be directly observed, and so must be estimated using economic models. If AI was making workers more productive at the frontier, TFP estimates should be accelerating just like overall labor productivity. But they aren’t: despite the strong headline labor productivity numbers, TFP growth has been near zero over the past year under the estimates of the San Francisco Fed.

(As Tedeschi notes, the Bureau of Labor Statistics puts 2025 TFP growth at 0.8 percent rather than zero — but that’s the same as its own 10-year average, so nothing extraordinary seems to be happening.)

Even though (a) numerous studies show AI can make workers more productive at specific tasks, and (b) overall labor productivity is increasing (due to companies running existing capital hard, especially to meet A, Tedeschi explainsI), we’re not yet seeing (c), a broad, economy-wide AI-driven increase in TFP, or underlying efficiency. And that last one is the whole ballgame—the thing that will determine the future of US living standards and whether all the investment in AI infrastructure will pay off the way many companies and investors expect.

From the conclusion, which shouldn’t be surprising given all the evidence we are still early days:

What the current numbers do not yet support, however, is the claim that this transmission [from micro impacts to macro impacts] is already well underway at the aggregate level: adoption is too thin, the cross-sectional signal is too weak, downstream bottlenecks remain, and the productivity pickup is too attributable to utilization to be confident in that conclusion. Instead, the acceleration we’re seeing so far is largely a result of companies trying to meet the demand for AI capacity by pushing the limits of their existing infrastructure.