Silicon Valley often presents artificial intelligence as the force that will usher in a new era of abundance. Yet when the chief financial officers of the world’s largest companies are asked how AI is affecting their bottom lines, the answer from 88% of them, according to McKinsey’s 2026 State of AI survey, is: “Not so much.”

This disconnect has appeared before. When the iPhone launched in 2007, commentators predicted that mobile technology would transform nearly every aspect of life. When broadband became widely available, Businessweek proclaimed a “New Economy.” And when personal computers reached almost every desk, Nobel laureate economist Robert Solow famously observed that the computer age could be seen everywhere except in the productivity statistics.

Across more than five decades, successive waves of digital innovation have achieved remarkable consumer adoption while productivity growth has continued to decline. I call this the Innovation Paradox: the faster we innovate, the slower we grow.

Two K-shapes, one problem

Digital technology diffusion follows what I describe as a double K-shaped pattern—a K-shape being a divergence in which one group pulls sharply ahead while the other falls behind. The first divergence is between consumers and enterprises. Smartphones reached 80% of Americans in seven years, while enterprise resource planning systems took 33 years to reach 57% of U.S. companies. Consumer AI usage has climbed to 53% of U.S. adults barely three years after ChatGPT’s launch in late 2022. By contrast, AI in productive enterprise deployment stands at only 10%.

The second K-shape exists within the enterprise sector itself. McKinsey’s 2026 survey found that 88% of organizations use AI in at least one business function, but only 6% qualify as “high performers,” meaning they achieve an EBIT (earnings before interest and taxes) impact of more than 5%. That share is unchanged from 2025.

The 82-percentage-point gap is not primarily a technology deficit. It is an organizational-transformation deficit. The 6% of companies that are succeeding are three times more likely to have redesigned workflows end to end. Only 21% of all AI adopters have done so.

The stock market reflects a similar divergence. Thirty-six AI companies in the S&P 500—the benchmark index of large U.S. public companies—now account for 45% of the index’s market capitalization. Over a three-year period, the headline S&P 500 returned 76%; excluding AI stocks, the return falls to 32%.

What electrification got right

Electrification and telecommunications in the early 20th century reached consumers and enterprises at broadly similar speeds and, crucially, with similar depth. Electric motors forced factories to replace rigid layouts with flexible, continuous-flow production. Telephones sharply reduced coordination costs. Organizational redesign followed, including Alfred Sloan’s multidivisional structure at General Motors and Frederick Taylor’s scientific management.

Diffusion was both deep, reshaping how companies operated, and broad, spreading through vertically linked industrial sectors.

Digital technologies, however, have diffused a fundamentally different economy. Services now account for the dominant share of GDP. A hospital or government agency does not have the same supply-chain linkages as a steel manufacturer. Economist William Baumol identified the underlying constraint in the 1960s, a dynamic since known as “Baumol’s cost disease”: “stagnant” service sectors resist productivity improvements because they depend heavily on human interaction. These Baumol sectors—including healthcare, education, public administration, and construction—represent roughly 50% of advanced-economy GDP and remain stubbornly resistant to technological transformation.

The overhead trap

Why does enterprise adoption stall? Not because the technology is unavailable, but because organizations have accumulated layers of complexity.

I have developed a cross-country metric called the Combined Overhead Ratio. It captures government spending, corporate selling, general, and administrative expenses (SG&A), and regulatory compliance costs. In the United States, the ratio crossed a threshold of 47% of GDP around 2000. Above that threshold, GDP growth has not sustainably exceeded 2.5%—roughly half the rate achieved when overhead remained below 35% in the 1960s.

Government spending has risen to 36% of GDP, while corporate SG&A has doubled since the 1980s. The Competitive Enterprise Institute estimates that U.S. companies spend more than $2 trillion each year on compliance.

My analysis of S&P 500 companies by sector and decade from 1985 through 2025 shows that 73% of those observations fall into the “overhead trap,” defined as rising SG&A alongside declining revenue growth. As Jack Dorsey of Block has argued, corporate hierarchy is an obsolete information-routing protocol. Most companies deploying AI today are putting a faster engine in a horse-drawn carriage. A related discussion describes the limitations of treating AI as merely a faster version of existing systems.

AI is different—but the barriers are not

AI is categorically different from earlier digital waves. Previous technologies primarily made it easier to process, transmit, or display information. AI reasons, decides, and creates. That distinction matters because Baumol sectors are not short of information; they are judgment-intensive. They require diagnosis, evaluation, and personalization—precisely the capabilities AI can augment.

The barriers to diffusion, however, remain familiar. More than 40 years of experience as a corporate executive and private-equity partner has revealed a recurring pattern.

At one consumer company, digitization produced impressive growth in the online channel but failed to scale across the company’s much larger offline business. Independently owned distributors refused to surrender customer data to an integrated platform. In healthcare, introducing digital tools to physicians is relatively straightforward. Establishing end-to-end digitization to improve clinical workflows is far more difficult, entangled in data governance, compliance requirements, and the asymmetric power that medical professionals hold over management.

These are not isolated cases. They represent many of the businesses in which I was personally involved. Writer’s 2026 enterprise AI survey, conducted with Workplace Intelligence among 2,400 global leaders, confirms the pattern at scale. Seventy-nine percent of organizations report challenges with AI adoption, a double-digit increase from 2025. In addition, 54% of C-suite executives say AI adoption is “tearing their company apart.”

Depth and breadth, not pace

The lesson of 50 years of digital innovation is that the speed of adoption matters far less than its depth and breadth. Consumer adoption of ChatGPT was the fastest of any consumer technology in recorded history, yet it has moved no macroeconomic needle.

What matters is whether AI penetrates core operations and leads to workflow redesign—that is, depth—and whether the transformation extends beyond superstar companies into the broad sectors that account for most employment and GDP—that is, breadth. On the survey evidence to date, those are also the numbers to watch: the high-performer share has held at 6% for a second consecutive survey year, and end-to-end workflow redesign stands at just 21% of adopters. Movement in those figures, rather than headline adoption rates, is the indicator that would show whether depth is building.

Historical evidence, from electrification through successive digital waves, points to long adoption timelines: generally more than 10 years and often several decades. AI is advancing faster than any predecessor as a technology. But the organizational redesign needed to absorb it proceeds at the pace of human institutions: slowly, painfully, and against intense resistance.

The Innovation Paradox persists not because technology fails to advance, but because institutions fail to adapt. Society will see a substantial economic impact only when AI is adopted deeply within enterprises and broadly across major economic sectors.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com.