Generative AI and the coming productivity boom

ARTICLE | September 07, 2023

Authored by RSM US LLP

After years of hype, generative artificial intelligence is leading us to the exact place proponents had predicted: the next frontier in productivity.

The journey has not been smooth. After a period of subdued productivity growth in the decade leading up to the pandemic was a significant factor that fueled skepticism about whether AI could fulfill its promise to enhance productivity.

On top of that, most of the AI usage during that period revolved around the first pillar of AI—machine learning—while the second pillar, generative AI, was at its earliest stage. Access to AI tools was costly in terms of both financial and human capital.

But that cost is plunging now, and it will have an outsize impact on the economy.

Scalability and ease of use are two of the most important factors necessary to propel the economy toward the next productivity frontier, both of which had been lacking.

Those constraints restricted the adoption of AI, resulting in relatively modest increases in productivity within firms, as evidenced by only a handful of available empirical research studies during the period before generative AI.

But the recent surge in public and academic interest in generative AI has provided researchers with a chance to assess its impact on productivity, even if primarily through experimental settings.

For instance, in a 2022 study of GitHub Copilot, generative AI increased coders’ productivity by 126%, while customer support agents improved their effectiveness by 14%.

All of the experiments with generative AI mentioned in the figure above have close to 100% AI adoption rate for the treatment group, it is important to expand the comparison with pre-generative AI research findings by dividing the increase in productivity by the share of the workforce that are exposed to AI.

If we use the base case from a recent study by Goldman Sachs, in which 25% of the workforce was exposed to AI, the increases in productivity growth remain significant compared to the pre-generative AI studies.

Can we replicate the New Economy?

The term New Economy was coined to describe the second half of the 1990s and the early 2000s, when the American economy witnessed robust economic growth with low inflation, driven by an apparent acceleration of productivity growth. Two catalysts were the internet and the widespread adoption of personal computers.

Accounting for shifts in labor and capital, the total factor productivity level—a proxy for technology-enhanced productivity—rose by an average of 1.6% annually from 1995 to 2005, significantly higher than the 0.5% growth rate from 2006 to 2022.

The parallels between generative AI and the transformative effects of personal computers and the internet are not surprising. In a recent statement, Satya Nadella, CEO of Microsoft, a leader in AI development, likened AI to "a tidal wave as significant as the internet."

We think the impact of AI, especially generative AI and the next pillar, artificial general intelligence, may exceed that of the internet—or anything we have ever seen. 

The capabilities of the internet and most of the first pillar of AI are limited by existing knowledge that human users are able to comprehend.

But for generative AI and beyond, a machine’s ability to create new knowledge and learn from it in a matter of seconds implies exponential growth with little to no limit.

According to the 2023 AI Index Report, Nvidia and Google have harnessed AI to enhance existing AI and chip systems, while scientists have used AI to aid hydrogen fusion and generate new antibodies.

In addition, the long-term demographic changes in the United States that are leading us toward an insufficient supply of labor in the next decades should only guarantee a continuing stream of investment in AI and automation for years to come. For many businesses, that investment is more a necessity than a luxury.

In other words, the significant boost to productivity from AI will be critical to the U.S. economy as it navigates the complexities of an aging population and high inflation.

That said, we have to acknowledge the potential headwinds that could slow down the development and adoption of AI—like geopolitical tensions that could affect the supply chain of semiconductors, the climate and environmental costs of running AI, and the negative social impact of labor displacement.

According to a recent report by IBM, the AI adoption rate in the U.S. remained at 25% in 2022, below the global adoption rate of 34%, and far from the 50% threshold that often marks the beginning of a productivity boom spurred by technology advances.

The takeaway

It is too early to know whether we can replicate the magic of the New Economy period. But the aspiration that AI represents is worth not only watching closely but also being a part of.