The paradox at the core of the AI revolution

Summary

A 2026 McKinsey survey found that 88% of large companies use AI, but only 6% achieve significant financial impact. The report highlights a gap between consumer and enterprise AI adoption, with 53% of U.S. adults using AI tools, while only 10% of enterprises have productive AI deployments. The article argues that AI’s impact on productivity is limited by organizational complexity and resistance to change. It compares AI to past technological waves, noting that digital innovation has not boosted productivity as expected. The author warns that AI’s potential will not be realized without deep organizational transformation.

Our Reading

The numbers tell one story.
Enterprise AI adoption lags far behind consumer use.
Only 6% of companies achieve meaningful AI impact.
Organizational complexity blocks transformation.
The Innovation Paradox persists.

The article reframes AI as a tool that requires deep change, not just tech.
It highlights the gap between hype and reality.
Most companies are not reengineering workflows.
The focus on speed overshadows the need for depth.
Productivity growth remains stubbornly low.

Author: Evan Null

Two K-shapes, one problem

The article outlines a double K-shaped pattern in technology adoption. Consumers embrace new tech quickly, while enterprises lag. For example, smartphone adoption reached 80% in seven years, but enterprise resource planning systems took 33 years. AI is no different—consumer use is high, but enterprise deployment is minimal. The author argues that this gap is not due to lack of technology, but lack of organizational change. Only 10% of enterprises have productive AI, while 53% of U.S. adults use it.

The second K-shape is within enterprises. While 88% of organizations use AI, only 6% are high performers. These top performers are more likely to have redesigned workflows. The stock market reflects this divide, with 36 S&P 500 AI companies making up 45% of the index’s market cap. Over three years, the S&P 500 returned 76%, but without AI stocks, the return dropped to 32%. This shows the growing influence of AI in the market.

The author contrasts this with past technological waves, like electrification and telecommunications. These innovations spread evenly across consumers and enterprises, and they led to deep organizational changes. In contrast, digital technologies have not transformed enterprises as much. Services now dominate the economy, and sectors like healthcare and education are resistant to productivity gains. This is known as the Baumol effect, where service sectors struggle to improve productivity.

The article also introduces the “overhead trap,” a metric showing that when government spending, corporate SG&A, and compliance costs exceed 47% of GDP, economic growth slows. Since 2000, this threshold has been crossed, and GDP growth has not exceeded 2.5%. The author argues that corporate complexity and regulatory burdens are holding back AI adoption. Companies are trying to implement AI without restructuring, leading to inefficiencies.

Despite AI’s potential, the barriers to adoption remain. The author notes that AI is different from past technologies because it can reason and create, not just process information. However, the challenges are the same: resistance to change, data governance, and power imbalances. The article concludes that AI’s impact will only be felt when it is deeply embedded in enterprises and broadly adopted across sectors.

What electrification got right

The article compares AI to past technological waves, like electrification and telecommunications. These innovations spread quickly and led to deep organizational changes. For example, electric motors forced factories to reorganize production, and telephones reduced coordination costs. These changes were both broad and deep, affecting entire industries. In contrast, digital technologies have not led to the same level of transformation. Services now dominate the economy, and sectors like healthcare and education are resistant to change. This is due to the Baumol effect, where service sectors depend on human interaction and struggle to improve productivity.

The author argues that AI is different from past technologies because it can reason and create, not just process information. This makes it more powerful, but also more challenging to implement. The barriers to adoption are not new—companies struggle with data governance, compliance, and resistance to change. The article suggests that AI’s impact will only be realized if it is deeply integrated into core operations and widely adopted across sectors. Without this, productivity growth will remain stagnant.

The author also points out that AI is not just a technology problem, but an organizational one. Companies are trying to implement AI without restructuring, leading to inefficiencies. The article highlights that the most successful companies are those that have redesigned workflows end-to-end. However, only 21% of AI adopters have done this. This shows that the real challenge is not the technology itself, but the ability to adapt and change.

The article concludes that the pace of AI adoption is not as important as its depth and breadth. Consumer adoption of AI has been fast, but it has not led to macroeconomic growth. What matters is whether AI is embedded in core operations and spreads across large sectors. The author warns that without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

The article also notes that AI is not just a corporate issue, but a societal one. The author argues that AI’s impact will only be felt when it is adopted deeply within enterprises and broadly across large economic sectors. This requires not just technological advancement, but institutional change. The article suggests that the real challenge is not the technology itself, but the ability of organizations to adapt and evolve.

The overhead trap

The article introduces the “overhead trap,” a metric that measures the combined impact of government spending, corporate SG&A, and regulatory compliance. In the U.S., this ratio crossed 47% of GDP around 2000, and since then, GDP growth has not exceeded 2.5%. This is significantly lower than the 1960s, when overhead was below 35%. The author argues that this overhead is a major barrier to economic growth and AI adoption. Government spending has risen to 36% of GDP, and corporate SG&A has doubled since the 1980s. Compliance costs alone are estimated at over $2 trillion annually.

The author’s analysis of S&P 500 companies shows that 73% of sector-decade observations fall into the overhead trap. This means that as SG&A increases, revenue growth declines. The author compares corporate hierarchy to an obsolete information-routing protocol, suggesting that many companies are using AI without restructuring. This leads to inefficiencies and limits the potential of AI. The article argues that AI is not just a technology problem, but an organizational one. Companies are putting a faster engine in a horse-drawn carriage, which limits the impact of AI.

The article also highlights that AI is different from past technologies because it can reason and create, not just process information. This makes it more powerful, but also more challenging to implement. The barriers to adoption are the same as before: resistance to change, data governance, and power imbalances. The author notes that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

AI is different—but the barriers are not

The article argues that AI is fundamentally different from past digital waves. Unlike previous technologies that made it easier to process, transmit, or display information, AI can reason, decide, and create. This is significant because Baumol sectors—like healthcare and education—are not information-scarce but judgment-intensive. AI can help with diagnosis, evaluation, and personalization, which are key in these sectors. However, the barriers to AI adoption remain the same as before. The article highlights that AI is not just a technology problem, but an organizational one.

The author points out that AI adoption is not just about technology, but about how companies organize themselves. Many companies are trying to implement AI without restructuring, leading to inefficiencies. The article suggests that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

The article also highlights that AI is not just a technology problem, but an organizational one. Companies are trying to implement AI without restructuring, leading to inefficiencies. The author argues that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

The article also notes that AI is not just a technology problem, but an organizational one. Companies are trying to implement AI without restructuring, leading to inefficiencies. The author argues that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

Depth and breadth, not pace

The article argues that the pace of AI adoption is less important than its depth and breadth. Consumer adoption of ChatGPT was the fastest of any consumer technology, but it has not led to macroeconomic growth. What matters is whether AI is embedded in core operations and spreads across large sectors. The author highlights that historical evidence from electrification to digital waves shows that adoption takes time—often over ten years, sometimes decades. AI may progress faster than past technologies, but the organizational changes required are slow and painful.

The author notes that AI’s potential will only be realized when it is deeply integrated into enterprises and broadly adopted across sectors. This requires not just technological advancement, but institutional change. The article suggests that the real challenge is not the technology itself, but the ability of organizations to adapt and evolve. Without this, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

The article also points out that AI is not just a technology problem, but an organizational one. Companies are trying to implement AI without restructuring, leading to inefficiencies. The author argues that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.

The author also highlights that AI is not just a technology problem, but an organizational one. Companies are trying to implement AI without restructuring, leading to inefficiencies. The article suggests that AI is not just a corporate issue, but a societal one. Without deep organizational transformation, AI will not deliver the promised productivity gains. This is the Innovation Paradox: the faster we innovate, the slower we grow.