Research finds AI drives productivity in customer service by up to 34%, with similar results in software development, business consulting, sales; AI adoption still low among US companies, but big investments suggest transformative potential in next decade

March 21, 2025 (Marketplace) –

Productivity in the U.S. has been on the upswing recently after a long period of slower growth. It's calculated roughly by dividing gross domestic product by labor hours. If we can get more economic growth out of the same amount of labor, that usually means wages rise — without causing inflation.

Most of the big jumps in living standards we've seen throughout history have been driven by jumps in productivity: the introduction of the steam engine, electricity, computers — these all brought productivity booms. And a lot of economists are hoping that what we're seeing right now might be the start of another one, driven by artificial intelligence.

Productivity rose by 2.7% last year, according to the Labor Department . That's well above the 1 to 1.5% we've averaged since the early aughts, and approaching the levels we saw during the last boom in the 1990s.

Two economists made a friendly bet a few years ago about how much productivity will grow this decade. Erik Brynjolfsson and Robert Gordon registered their wager on the prediction website Long Bets. The stakes are $400 .

"But I think it's more of a reputational bet," said Brynjolfsson, who's a professor at Stanford , where he directs the Digital Economy Lab . He's bullish on productivity.

His own research on customer service agents who started using an AI large language model showed their productivity improved as much as 34%. And other recent studies on software development, business consulting, writing and sales showed similar results.

"These are the biggest gains I've ever seen," said Brynjolfsson. "I mean, to see double-digit gains just within a few months is almost unheard of."

In the other corner: Robert Gordon , professor at Northwestern and author of "The Rise and Fall of American Growth."

"It's not going to be a revolution," Gordon said of AI. "It's not going to blow out human nature. I think that's all greatly overblown."

He believes AI will raise productivity, but not in the transformative way that electricity or computers did. He points to the last 20 years or so of modest gains, despite high tech innovations ranging from smart phones to the app economy.

"Those things really are minor compared to the difference it made to have trucks instead of horses, to have airplanes instead of trains," said Gordon, who doubts that the share of tasks AI can currently automate in the workplace will be enough to significantly move the needle in the broader economy.

For now, Brynjolfsson has pulled to the lead, though it's not clear yet if the bump in productivity we're seeing is evidence of the start of a durable AI boom.

Joseph Briggs has been searching for signs. He's a senior economist for Goldman Sachs Global Investment Research .

"If we look at an economy-wide basis, the impacts still look quite negligible," he said.

Goldman Sachs has forecast that once generative AI has been widely adopted, it could raise productivity in the U.S. by 15% over 10 years.

Briggs is watching for signals from the labor market that would indicate AI adoption is slowing job growth. He sees limited evidence in certain industries like computer programming, customer service, management consulting and legal services, where AI has been more aggressively adopted.

But Briggs thinks it will take a couple of years for AI to show up in federal productivity data. Goldman Sachs estimates only about 6% of U.S. companies are using the technology.

"You need to restructure workflows, you have to have workers comfortable using it, and all of these things take time," Briggs said.

He does, however, see plenty of evidence that companies are making big investments in AI, especially in industries like law.

Annie Datesh is the chief innovation officer for Silicon Valley law firm Wilson Sonsini , which works with a lot of tech startups. The firm developed an AI tool to help review agreements for cloud services, which many tech companies frequently require.

"You'd be negotiating them a lot. It would be high volume, but a standard form, and suddenly automation and AI start to make sense," she said.

Wilson Sonsini uses AI, trained on its own best practices, to customize the agreements. Then a human lawyer reviews and finalizes them.

The tech should eventually help the firm serve more clients, faster. But it's not quite there yet.

"Well, think about when you hire a new person, right?" said Datesh. "They're always going to kind of slow you down at first, but then they get trained up, and then they can take over tasks."

Right now, she said, the AI is still kind of green. "And so you're still monitoring it and supervising it like it's that new person, which will theoretically slow you down."

It's a pattern you often see when a powerful new technology diffuses through the economy said Stanford's Brynjolfsson — at first it looks like nothing is happening, until suddenly productivity takes off in what he calls a J-curve.

"With the steam engine, with electricity, with the internal combustion engine, when we measured that, it often took literally decades," he said. "It took about 30 years for electricity to really have its full payoff."

Brynjolfsson expects AI to deliver productivity benefits much faster, since it doesn't require specialized hardware or skills to use. He's got until 2029 to win the bet.

Roberts at Northwestern said he's still skeptical, but he'd be glad to be proven wrong.

"I'm happy for the economy if AI surprises me, because don't we all want more growth?" he said.

Though: whether a productivity boom from AI will necessarily be a win for most workers? That they're not so sure about.