AI's Productivity Paradox: CEOs See Little Impact Today Despite High Hopes

The AI Productivity Disconnect

The promise of artificial intelligence has been nothing short of revolutionary: a new era of soaring productivity, streamlined operations, and transformed work. Yet, a stark reality is emerging from the C-suite. A comprehensive new study from the National Bureau of Economic Research (NBER), surveying thousands of CEOs and executives across the U.S., UK, Germany, and Australia, reveals a surprising consensus.

Despite an estimated $250 billion in corporate AI investment in 2024, nearly 90% of firms report that AI has had no measurable impact on employment or productivity over the last three years. This data resurrects a famous economic conundrum, forcing a critical question: Is history repeating itself?

Resurrecting Solow's Paradox

In 1987, Nobel laureate economist Robert Solow famously quipped, "You can see the computer age everywhere but in the productivity statistics." He was observing the so-called "productivity paradox" of the information technology age, where massive investment in computers initially failed to deliver expected economic gains.

Today's AI landscape is drawing direct parallels. Apollo Chief Economist Torsten Slok recently echoed Solow, stating, "AI is everywhere except in the incoming macroeconomic data... you don't see AI in the employment data, productivity data, or inflation data." The NBER findings provide the hard numbers behind this sentiment.

The study details that while about two-thirds of executives report using AI, that usage averages only 1.5 hours per week. A significant 25% of respondents do not use AI in the workplace at all. This slow, shallow adoption contrasts sharply with the hype.

Conflicting Signals and Sector-Specific Realities

The data, however, is not monolithic. The picture is nuanced, with significant variations by sector and perspective. While the broad CEO survey shows minimal current impact, other sources point to early, concentrated gains.

• Financial Sector Optimism: A separate EY survey of 240 financial services CEOs found that 60% believe AI investment will maintain or increase headcount in 2026. Only 28% predicted a reduction.
• Internal Productivity Claims: Hedge fund executive Stephen Craver, cited by Business Insider, reported founders describing "mind-blowing" internal productivity gains, with some CEOs claiming they could "triple revenues" without hiring another human.
• Government Analysis: Federal Reserve Governor Michael Barr, in a recent speech, noted evidence that AI adoption is leading more to internal re-allocation of tasks within firms than to layoffs, with many companies planning to retrain existing workforces.

This divergence highlights a key theme: the benefits of AI may be highly concentrated. Slok points out that outside the "Magnificent 7" tech giants, there are "no signs of AI in profit margins or earnings expectations." The gains are real for some, but not yet widespread.

The J-Curve Theory and Lagging Indicators

Economists like Erik Brynjolfsson of Stanford's Digital Economy Lab argue we are witnessing a predictable pattern. They point to the concept of a "J-curve" for general-purpose technologies like AI.

Massive investment and organizational disruption come first, potentially depressing measured productivity. Only later, after systems are integrated and workflows are redesigned, does productivity surge. Brynjolfsson's own analysis suggests U.S. productivity jumped roughly 2.7% in 2025, nearly double the pre-AI average.

He and others see signs of this transition. A Federal Reserve Bank of St. Louis report observed a 1.9% increase in excess cumulative productivity growth since ChatGPT's late-2022 launch. Furthermore, as noted by former Pimco CEO Mohamed El-Erian, the recent decoupling of strong GDP growth from more modest job gains mirrors a phenomenon seen during the 1990s office automation boom.

Employment Impact: Substitution vs. Transformation

The effect on jobs remains a central and complex question. The NBER CEO survey indicates firms expect only a 0.7% cut to employment over the next three years, while forecasting a 1.4% productivity increase.

Governor Barr's speech highlighted research showing AI is starting to adversely affect early-career workers in highly exposed occupations like software development and customer service. This aligns with Brynjolfsson's own past research.

However, the narrative is not solely one of displacement. IBM's Chief Human Resources Officer recently stated the company would triple its number of young hires, fearing that automating entry-level tasks could create a future leadership pipeline crisis. This underscores a longer-term view where AI changes the nature of work rather than simply eliminating it.

Trust, Implementation, and the Road Ahead

Beyond the macroeconomic data, human factors are crucial. A ManpowerGroup survey found that while regular AI use among workers increased 13% in 2025, confidence in the technology's utility plummeted 18%. This "toxic relationship" and trust deficit could hinder effective adoption.

The future of AI's economic impact, as Slok posits, depends less on the technology itself and more on "how generative AI is used and implemented in different sectors." The fierce competition driving down the price of AI tools, unlike the monopoly pricing of early IT, makes widespread adoption more feasible—if organizations can navigate the cultural and operational challenges.

Ultimately, the current moment is one of measured investment and cautious experimentation. The grand promises of an immediate productivity revolution have collided with the messy reality of enterprise integration. Whether AI follows the trajectory of the 1990s IT boom or charts a new path remains the multi-trillion-dollar question for the global economy.