The AI Paradox: Why Hundreds of Billions in Investment Haven't Moved the Productivity Needle... Yet

The AI Paradox

Artificial intelligence is everywhere. From the apps on our phones to the algorithms that shape our news feeds, AI's influence seems to be rapidly reshaping our world. We are told we are living through a technological revolution on par with the internet or electricity. This has fuelled a massive wave of investment, corporate restructuring, and public fascination.

Yet, a puzzling contradiction lies at the heart of this revolution. If AI is so transformative, why can't we see its impact in the broad productivity statistics that measure economic output? This disconnect echoes the famous 1987 observation by Nobel laureate Robert Solow, who noted that the computer age was "everywhere but in the productivity statistics." Today, we face a similar AI paradox. A deep-dive analysis by Citi Research offers some surprising answers to this puzzle, and this post will distill the most important takeaways.

productivity_and_the_ai_revolution_CCM_1757708994556.pdf

Download PDF • 1.33MB

Listen on Spotify: https://open.spotify.com/episode/6WzjWSqoj1dXN09x8dAWc9?si=dzJsdmaYR865n7ss1pVcrw

1. The AI Boom Hasn't Actually Started (Yet)

Despite recent improvements in U.S. productivity, the Citi analysis finds there is "little evidence to date of a sustained acceleration in productivity" that can be attributed to AI. The recent rebound appears to be more of a cyclical recovery from the sluggish period following the Global Financial Crisis, rather than a new, AI-driven structural trend.

The key concept to understand this delay is the “productivity J-curve.” This describes a predictable gestation period for any transformative technology, characterised by a massive, often profit-draining, upfront investment in new infrastructure, coupled with the slow, difficult work of re-skilling a workforce. Only after this foundational work is done does the economy begin to reap the rewards, sending productivity soaring up the steep side of the "J."

"The basic observation is that the full deployment of transformative new technologies entails a period of gestation... productivity gains are typically not achieved until the supply-side of the economy is re-organised and the new technology is broadly adopted."

For businesses today, this means the immense spending on AI is not a sign of immediate returns, but a necessary and costly ticket to entry for the productivity boom to come.

2. History's Lesson: The 40-Year Wait for Electricity's Payoff

History provides a sobering, yet ultimately optimistic, blueprint for this exact phenomenon. The report draws a direct parallel to the adoption of electricity, using the research of economic historian Paul David. While the incandescent lamp was introduced in 1879, it wasn't until the 1920s - roughly 40 years later - that electrification had a significant, measurable effect on U.S. productivity.

This cautionary tale isn't just about the long timeline; it's about the reasons for the delay. Firms were reluctant to scrap existing, functional capital like steam engines. Many ran inefficient "overlays of new and legacy technologies," and it took decades to build up a pool of human expertise - engineers, electricians, and technicians - needed to fundamentally reorganise factories. Meaningful gains only appeared after electricity crossed a critical 50% adoption threshold in U.S. factories. Other transformative technologies, including the steam engine, the railroad, and the internet itself, followed similarly protracted adoption cycles before their full economic potential was unleashed.

3. But This Time, Adoption is on Fast-Forward

While the pattern of delayed gains is historical, the pace of AI adoption is not. The analysis judges that "the diffusion of AI is proceeding at a historically rapid rate." This is where the current revolution diverges sharply from the past.

The report offers a powerful data point to support this conclusion. When comparing the current AI investment boom to the 1990s internet boom, AI-related investment looks to be running "a notch faster." The critical takeaway here is that while we are still in the early, flat part of the "J-curve," we may be moving along it much more quickly than with any previous technological shift.

4. The Biggest Roadblock Isn't Technology, It's People

The organisational and human challenges of today provide a direct echo of electricity's slow rollout and are the primary reason why we are currently in the flat part of the "J". The primary challenge of the AI revolution isn't just inventing powerful tools; it's successfully integrating them into the complex fabric of our daily work. The research highlights a BCG report finding that firms will need to focus "two-thirds of their effort and resources on people-related capacities," with only one-third dedicated to the technology and algorithms themselves.

This reveals a fundamental tension between ambition and strategy. According to a Gallup poll cited in the report, 44% of workers say their firm is actively trying to integrate new AI tools, but only 22% believe their company has a clear plan or strategy for doing so. This disconnect signals the chaotic, experimental nature of the current adoption phase. It’s no surprise, then, that survey data cited in the report finds that only 5% of GenAI projects are considered "fully scaled and creating meaningful value." The real work - and the real bottleneck - lies in training, ensuring worker acceptance, and fundamentally reorganising business workflows around these new capabilities.

5. The Potential Payoff is Staggering

After accounting for the historical delays and present-day challenges, the analysis shifts to the potential upside - and it is enormous. The academic literature suggests a massive potential gain in productivity once the adoption phase matures.

Synthesising this literature, the report presents a stunning estimate: "a total gain in productivity of 6 to 16%, or a boost to productivity growth averaging ½ to 1½ ppts a year if these gains accrue over a decade." To be clear, a sustained 1.5 percentage point boost to annual productivity growth would be a seismic economic event, fundamentally altering the trajectory of economic growth and wealth creation.

The upper end of that range would mirror the incredible productivity boom seen during the peak of the internet revolution from 1995-2004. This is the economic prize waiting at the other end of the J-curve.

Conclusion: Beyond the Hype

The story of AI's economic impact is not one of failure, but of a predictable - if accelerated - gestation. The current lack of a productivity boom is not a sign that AI is overhyped, but rather that it is following a historical pattern, the "productivity J-curve," at an unprecedented speed. We are in the midst of building the foundation - the investment, the infrastructure, and the human capital - required for the coming transformation.

This leads to a final, critical question. Given that the pace of adoption is faster than any previous technological revolution, the critical question is no longer if the AI productivity boom will arrive, but how prepared our businesses and society will be when it does.

Disclaimer: This article is for informational and educational purposes only and does not constitute financial, investment, or professional advice. Past performance is not indicative of future results. Readers should conduct their own research or consult a qualified professional before making any financial or business decisions.

Alpesh Patel OBE www.campaignforamillion.com