A Stanford Economist Says the AI Productivity Surge Has Arrived. The Data Agrees

TL;DR

Stanford economist Erik Brynjolfsson says the AI productivity payoff is now showing up in national economic data. US productivity growth hit 2.7% in 2025, nearly double the prior decade, while job growth slowed, meaning the economy is producing more with fewer workers.

What Happened

Stanford economist Erik Brynjolfsson published an essay in the Financial Times declaring that the long-awaited AI productivity surge has arrived in the US economic data. US productivity growth hit 2.7% in 2025, nearly double the average of the past decade. At the same time, payroll growth was revised downward by 403,000 jobs whihle real GDP held strong at 3.7% in Q4. Translation? The economy is producing more with fewer workers.

Brynjolfsson, who has spent years researching the lag between technology investment and measurable economic returns, frames this as the productivity "J-curve," or the pattern where general-purpose technologies require years of massive investment in intangible capital (reorganizing processes, retraining workers, developing new business models) before gains appear.

He argues AI is now transitioning from this investment phase into a "harvest phase" where those efforts show up as measurable output. He posted on X that US productivity growth was "nearly double the average of the previous 10 years" and expanded on the implications in a follow-up thread.

Micro-level evidence backs the shift: Brynjolfsson's research found a 16% decline in entry-level hiring within AI-exposed sectors, while workers who used AI to augment their skills saw growing employment. But he also offered a sharp caveat: despite the macro data, most companies are still using AI as a "glorified dictionary." The real productivity gains are being driven by a small cohort of power users compressing weeks of work into hours at the moment, not yet by broad organizational adoption.

The Key Numbers

2.7% — US productivity growth in 2025, nearly double the average of the past decade

403,000 — Jobs by which payroll growth was revised downward, while GDP held strong

3.7% — Real GDP growth in Q4 2025, despite fewer workers

16% — Decline in entry-level hiring within AI-exposed sectors, per Brynjolfsson's research

The macro numbers confirm what power users already know. For anyone deep in AI workflows, this data is confirmation, not news. The economy is producing more with fewer workers because a small group of people and companies are using AI to collapse timelines. Brynjolfsson's data simply proves that the impact has now reached the scale where national economic indicators can detect it.

"My general take is because most companies are still at the starting point," says SmarterX founder and CEO Paul Roetzer on Episode 198 of The Artificial Intelligence Show. "I think the problem is everyone keeps waiting for the data to tell us this. The government officials are waiting to see it in GDP or in productivity or these leading indicators. They're just waiting for proof that it's going to happen."

Past data cannot predict what's coming. Roetzer argues that no amount of economic research, historical analogies, or government data will capture what is about to happen in the next one to three years.

"We just can't see what's coming by looking at what's happened so far," he says. "No matter how many studies we go find, no matter what the Fed does, what these economists research, how many times we look at past general purpose technologies, none of it is going to show us."

The 7-point thought experiment. After spending time with health system executives, banking leaders, and heads of major software companies, Roetzer distilled the future into seven assumptions. He posed them as a thought experiment: assume these are true in your business within one to three years.

1. Everyone has access to a gen AI platform — ChatGPT, Claude, Copilot, Gemini — that can think, reason, understand, and create.

2. Everyone understands what AI is and what it is capable of doing — not just basic chat and answer-engine use cases, but the deeper reasoning and multimodal capabilities that open up a world of business applications.

3. Everyone has received personalized training to help them prioritize use cases and maximize the impact of AI on their jobs.

4. Everyone's personal AI assistants have on-demand access to a comprehensive company knowledge base and clean, real-time data — able to turn data into intelligence, intelligence into insights, and insights into actions.

5. Everyone has on-demand access to AI subject matter experts across every topic inside and outside the organization.

6. Everyone has on-demand access to high-level strategic support and consultation from their AI for any business problem or goal.

7. Everyone has access to AI agents that can complete digital tasks with a high level of reliability and autonomy.

"If these are true, what changes across talent, technology, processes, products and services, and business models?" Roetzer says. "The answer is everything."

Most of this is already real. The uncomfortable part: outside of the data access layer (#4) and reliable autonomous agents (#7), all of this exists today. The people who pay for premium AI tools and understand their full capabilities already live in this world. "Most of that is all true already," Roetzer says. "The people who know how to use gen AI tools, you live in a world where everything I just said is like, well, yeah, of course. That's all obvious."

And yet companies can't clear step one. "Most companies I talk to, especially large enterprises, are still stuck on number one: access to gen AI," says Roetzer. "The ones who have actually provided gen AI access to their teams have rarely moved to two and three. It is shocking how many enterprises are stuck on the most basic and accessible parts of AI adoption transformation."

"I sometimes think like, this can't be reality. And then I go and spend two weeks with the leaders of these companies and you realize like, oh my gosh, it may actually be further behind than I thought it was. And I was already kind of bearish on where the adoption was."

Steps 1 through 3 alone can transform a business. The data infrastructure challenges (#4) are real but often used as an excuse to stall. Roetzer's argument: if your organization just gives people access, helps them understand the full capabilities, and provides personalized training, all doable with zero IT involvement, you can completely transform a business.

"If you just do one through three, you can throw all the research out the door and it'll change your company," he says. "10, 20, 50% productivity gains. If you can't realize those gains, you're doing something wrong."

You don't need engineers to do this. Roetzer's company, SmarterX, itself has no IT department, zero engineers, and zero data scientists. The team is built on strategy, creative, and liberal arts backgrounds. "We are not doing this with engineers who are doing all this grunt work behind the scenes to figure out how to apply all this stuff," Roetzer says. The people driving AI transformation inside the companies SmarterX works with are sitting in marketing, customer success, and HR, not in IT.

SmarterX Take

Brynjolfsson's data is real and meaningful. The harvest phase is not a prediction; it is showing up in national economic indicators right now. But the macro data obscures a critical asymmetry: the productivity gains are being driven by a small minority of power users and forward-leaning companies, while the vast majority of organizations are still failing at step one.

The gap between AI capability and organizational adoption is not closing. It is widening. And every day it widens, the competitive distance grows between the companies that are compressing work and the ones still debating whether to buy licenses. Domain expertise is the knowledge base that most professionals already have. Use it. Talk to the AI. Apply what you know to what it can do. Waiting for your data house to be in perfect order is a stall tactic that costs real ground.

The most important takeaway from Roetzer's thought experiment: the seven assumptions aren't aspirational. Five of them are already true for anyone paying for a premium AI subscription and investing time in learning what these tools can do. The only question is whether your organization will get out of its own way long enough to capture what's already available.

What to Watch

Whether the 2025 productivity data triggers meaningful changes in corporate AI strategy, or just becomes another data point that executives acknowledge without acting on. The macro trend will accelerate as AI coding tools reshape product development cycles and autonomous agents become more reliable. Axios reported that AI is already compressing tech product development timelines significantly.

The companies that move on steps 1 through 3 now (access, understanding, and personalized training) will be positioned to capture value from steps 4 through 7 that Roetzer outlined as the data infrastructure and agent reliability improve over the next 12 to 24 months. The companies that wait for more data before acting will find themselves in the same position a year from now, watching the same productivity charts, still stuck on step one.

Resources

FT: US Productivity Growth and AI — Erik Brynjolfsson Essay → ft.com

Paul Roetzer: 7-Point AI Thought Experiment (LinkedIn) → linkedin.com

Erik Brynjolfsson: US Productivity at 2.7% (X Post) → x.com

Erik Brynjolfsson: Follow-Up Thread on AI and Productivity → x.com

Axios: AI Coding and Product Development → axios.com

Heard on The Artificial Intelligence Show, Episode 198
Paul Roetzer and Mike Kaput break down why the AI productivity surge showing up in national data still doesn't capture what's happening on the ground — and the 7 assumptions that reveal how far behind most companies really are. Listen Now