Measuring AI’s True Productivity Dividend

Writer: Shyan Sahdev

Editor: Samarveer Singh

Introduction

Artificial intelligence (AI) represents one of the most significant technological disruptions since the advent of the personal computer. Businesses worldwide are investing unprecedented sums in AI technologies, with global private investment reaching a record $252.3 billion in 2024, over thirteen times the 2014 level. In fact, Generative AI alone has attracted $33.9 billion of funding in 2024, reflecting an 18.7% jump from the previous year, and organizational adoption has accelerated rapidly, with 78% of companies reporting AI use in 2024 compared to just 55% the year before. These statistics do not just illustrate the soaring importance of AI but also suggest that businesses today view it as a transformative means of enhancing operations. Yet, the answers to the technology’s economic impact remain far and few. Advanced economies face a persistent "productivity puzzle" that continues to confound economists and policymakers alike. Labour productivity growth in many OECD countries has stagnated at barely 0–1% annually since the 2008 financial crisis. This is in stark contrast to the robust productivity gains of previous decades. Can artificial intelligence finally break through this productivity stagnation and reignite meaningful growth in output per worker? And if so, what magnitude of impact can we realistically expect, and over what timeframe?

How does Artificial Intelligence impact Productivity?

Figure 1: The mechanisms through which AI impacts productivity

Artificial intelligence can reshape productivity through three powerful, mutually reinforcing mechanisms, as illustrated. These mechanisms are not to be decomposed – crucially, the effects feed into one another. Productivity gains spur further AI investment, which drives new innovations, consequently creating a feedback loop that compounds growth over time. 

The first channel includes AI transforming human work through both replacement and enhancement. More than substitution, the technology in minutes, which otherwise older systems also augment human capability with real-time insights and intelligent recommendations, enabling workers to achieve far more per hour. In such a setting, early adopters will be gaining a clear productivity edge. They outcompete rivals, they attract capital and talent, spreading gains across markets. Secondly, AI acts as a new form of digital capital. Just as computers, AI systems now deliver scalable output at minimal marginal cost. Consider a chatbot resolving thousands of queries that otherwise would require immense human effort. These deployments, then, replicate the classic effect of capital deepening: more output per worker, higher GDP per hour worked. 

A more variable, third means is that AI accelerates innovation as a research multiplier. It enables scientists to process datasets in hours instead of months, engineers to test thousands of prototypes virtually, and pharmaceutical firms to discover promising compounds far quickly. This leads to dynamic research and development, which fuels total factor productivity.

Numbers to watch, and Doubt

Some forecasts are promising. Goldman Sachs reckons widespread AI adoption could lift U.S. labour‑productivity growth by 1.5 percentage points a year for a decade, matching the Internet‑fuelled boom of the 1990s. France’s AI Commission forecasts 1.3%, comparable to the electricity revolution; the OECD’s more conservative analysis suggests 0.4–0.9 points.

Yet the distance between top‑end optimism and bottom‑line reality is wide. Forecasts hinge on how quickly adoption spreads beyond a handful of AI‑intensive sectors, such as tech, finance, and R&D, into the rest of the economy. Without that diffusion, the gains visible in specific firms will barely ripple through national accounts. For instance, MIT’s Daron Acemoglu suspects that today’s crop of tools is too narrow in scope to shift long‑run growth much at all.

Figure 2: Predicted increase in annual labour productivity growth over a 10-year horizon due to AI (in percentage points)

Adoption: brisk, but patchy

If the potential is clear, so is the pattern: rich, tech‑savvy economies are pulling ahead. In 2024, the United States attracted $109.1 billion of private AI investment, which is almost twelve times China’s $9.3 billion. Greater China’s adoption rate nevertheless jumped by 27 percentage points last year and Europe’s by 23. Unsurprisingly, early inroads are in knowledge‑heavy fields such as software, finance and customer service.

Figure 3: The number of notable AI models by select geographic areas 

Interestingly, even in these frontrunners, truly “AI‑mature” firms are scarce. Stanford’s AI Index finds that only 1% have fully integrated AI into their operations with measurable productivity gains. And the IMF notes that while 60% of jobs in advanced economies could be affected by AI (half positively, via task augmentation), exposure falls to 40% in emerging markets and 26% in low‑income countries. The danger is an AI‑driven widening of the global productivity gap.

From individual tasks to the national accounts

At the micro level, the gains are more tangible. Controlled trials suggest that generative AI can boost task‑level output by 5–25%, particularly for juniors, who benefit most from its “coaching” effect. A novice coder using an AI assistant can match the output of a mid‑career peer; a customer‑service agent with an AI prompt system handles calls faster and with fewer errors. Experienced workers benefit too, cycling through designs, drafts or prototypes at speed.

Conclusion: the dividend, and the race to claim it

The productivity dividend from AI is substantial, but it is not guaranteed to materialise, and it will not arrive fully formed. On cautious assumptions, it could lift growth in output per worker by 0.4-0.9% in advanced economies, and given ideal circumstances, by something closer to 1.5%. The truth, then, will depend less on the technological brilliance of AI models than on the complements in human capital, infrastructure, interoperable systems, and rules that promote diffusion across different geographies and sectors.

If adoption penetrates far beyond the tech frontier, to factory floors, hospital wards and classrooms, the compounding effects could be transformative. If it remains confined to a few data‑rich niches, the rewards will accrue to a small set of firms and countries. With record investment and usage, the decade ahead will decide whether AI becomes an economy‑wide workhorse or an expensive craze. So, the dividend is there to be earned; however, the pace at which it accrues will determine who prospers and who misses out.