Rising Productivity Has Not Settled AI’s Role in the Labor Market

Stacey Smith said the latest productivity numbers support the view that output per hour is improving, while leaving open how much of that improvement should be credited to AI. The first-quarter 2026 figure she cited was 2.9% year over year, higher than expected.

Bloomberg showed a chart sourced to the US Bureau of Labor Statistics via FRED titled “US productivity growth has picked up since the pandemic.” The chart plotted the US labor productivity index — output per hour for all nonfarm workers — against a dotted 2010–20 trend line, with the index running above that earlier trend by 2026.

Smith framed the evidence as a “good news, bad news” situation. Productivity is rising. The job market, in her words, is “not great,” and there is “a lot of concern and rightly so” that AI is displacing jobs. Caroline Hyde had put the question directly: whether AI is boosting productivity “not at the expense” of labor. Smith’s answer was that signs of productivity improvement exist, but the labor-market implications are troubling.

That was the measurement problem Ed Ludlow pressed. Productivity has been growing at about 2% since 2020, he noted, but the aggregate number does not reveal whether AI is making a tangible contribution inside each reading. Smith’s answer was cautious: “It is so hard to know where the productivity is coming from.”

The point is not that the productivity data are irrelevant. They are the core evidence in the debate. But the number compresses several possible stories into one measure: workers producing more with AI tools, firms producing with fewer labor hours after layoffs, or some combination.

The wager Smith described is between two US economists: Erik Brynjolfsson of Stanford and Robert Gordon of Northwestern. They bet on whether productivity would grow by an average of 1.8% between 2020 and 2030. According to Smith, recent productivity performance suggests the economy may clear that bar.

But the wager’s benchmark is narrower than the argument surrounding it. If productivity exceeds 1.8%, that would say the measure rose faster than the agreed threshold. It would not settle why productivity rose.

Bloomberg Businessweek’s on-screen excerpt framed the same tension around AI investment. Companies are investing heavily in AI, the excerpt said, but productivity growth will likely remain modest until they figure out how to retool workflows to make the best use of the technology. The visual also cited a Massachusetts Institute of Technology study from the previous year finding that, despite huge investments in AI, 95% of businesses reported no measurable return on investment.

That detail separates AI adoption from realized productivity. The source presented AI spending as conditional rather than self-executing: companies must redesign workflows around the technology before the payoff shows up broadly.

Stacey Smith said Robert Gordon offered a more skeptical interpretation of the productivity acceleration: layoffs may be part of the explanation. The mechanism is straightforward. Productivity, as she described it, is “the amount our economy produces divided by hours worked.” If labor hours fall because of layoffs, the denominator shrinks. If output does not fall by the same amount, output per hour rises.

That is why the same productivity figure can carry different labor-market meanings. A higher output-per-hour number could reflect workers using AI to produce more in the same time. It could also reflect firms producing roughly the same amount with fewer labor hours.

Smith’s account leaves the bet in an unsettled position. Productivity may be strong enough to beat the 1.8% average threshold. But Gordon’s layoff explanation means a win for productivity would not automatically be a clean win for the optimistic AI story. It could be evidence of a more productive economy, a leaner labor market, or both.