Widespread adoption of artificial intelligence (AI) has the potential to deliver a major boost to productivity across various industries and economies around the world. In a recent Capital Economics survey, nearly 80% of clients responded that AI would transform the global economy.
How will that occur, and what is the potential size and speed of productivity gains? Could AI systems replace humans in a way that results in mass unemployment, or will they drive employment growth?
The Impact of Technological Revolutions
AI can be considered a general-purpose technology (GPT), which historically has had major economic effects.
For examples of these, let’s take:
- the invention of the internal combustion engine in the late 1800s
- the development of the steam engine in the UK during the 1700s and 1800s that facilitated the Industrial Revolution
- the introduction of electricity in the US in the early 1900s
- The information and communication technology (ICT) revolution in the late 1900s all brought significant changes to economies by boosting productivity.
While it has historically taken decades for GPTs to deliver tangible productivity gains, adoption lags – the length of time for countries to implement a new technology after its invention – have shortened.
Annual productivity in the UK and US from the advent of steam and electricity averaged 0.2-0.3% growth, in part because the gains materialized over a long time period. But improvements have become compressed, along with the adoption lags. Hence, the US saw annual productivity growth of 1.5 percentage points between 1995 and 2005, Capital Economics notes in its report AI, Economies and Markets.
The emergence of previous GPT revolutions has typically been net positive for labor markets. There has been significant short-term friction from technological advances, but workers have tended to transition to new, higher-paying jobs.
This is because GPTs have either increased demand for certain goods or services or created new economic sectors altogether. There is likely to be a direct boost to AI-related jobs, as well as an overall increase in employment and real wages, as productivity growth drives economic aggregate demand throughout the economy.
A recent study by MIT estimated that 60% of workers in the US are employed in occupations that did not exist in 1940. This implies that more than 85% of employment growth over the last 80 years is attributed to the technology-driven creation of new positions, according to an analysis by Goldman Sachs.
Shifts in workflows driven by advances in AI could expose the equivalent of 300 million full-time jobs to automation, Goldman’s research shows. But not all automation will translate into layoffs, as most jobs are more likely to be “complemented rather than substituted by AI”.
What does this mean for productivity driven by AI?
The Ways AI Could Boost Global Productivity
There are several ways that AI applications can transform workforce productivity around the world.
Companies may be able to use fewer resources or do more with existing resources. This could see AI directly replace humans by carrying out tasks more efficiently or by helping humans to become more productive and free up time to spend on other pursuits.
AI could also free up other resources. For instance, the use of shared driverless vehicles could allow land previously used for parking to be repurposed for more productive uses.
AI can increase output per hour worked in three ways:
- Increasing the amount of capital per worker, for example, by using new AI software or enabling more preventative equipment maintenance to slow the depreciation of assets.
- Raising the quality of the workforce.
- Increasing the efficiency with which labor and capital are combined to facilitate better work practices.
AI has the potential to help facilitate new inventions. AI can filter vast amounts of research to suggest the most viable projects and predict the results of real-world experiments. This has the potential to boost innovation and result in breakthroughs across a wide range of fields, from drug discovery to education to transport. It could boost productivity growth permanently and create a virtuous cycle in which AI trains itself to improve innovation.
Quantifying the Impact of AI
While AI offers the potential for major benefits, these are likely to be realized gradually, as in previous technological revolutions, according to analysts at Capital Economics.
“We think that the impact of past technological advances is the best guide to the likely impact of AI in the coming years and, given the nature of AI, the ICT revolution is probably the best one to look at. Data show productivity gains during the ICT revolution reached 1.5% per annum for the US, so that seems like a reasonable guide to what is achievable for a country that is at the forefront of developing and deploying the technology.”
This could bring an end to the malaise in productivity in developed economies over the last 10-20 years.
The gains could be larger, as many applications that emerged during the digital revolution were focused on improving the consumer experience, whereas AI has more business use cases that will directly boost productivity, the analysts note.
Similarly, Goldman Sachs’ research suggests that AI tools could lift productivity growth by 1.5 percentage points over a 10-year period and drive a 7% (or almost $7 trillion) increase in global gross domestic product (GDP).
A more powerful AI development could see upwards of 3 percentage points of productivity growth, with around 1 point of labor displacement.
Barriers to AI-Driven Productivity Growth
AI has the potential to change the game, but realized gains are quite different to theoretical gains. Historically, the most significant gains in productivity gains have occurred at times when technological advancements coincided with political and/or social changes or the development of complementary innovations, Capital Economics notes.
There are several barriers to overcome as companies must identify the best ways to use the technology, raise the funds to pay for implementation and overcome any institutional inertia. They will also need to employ enough skilled workers to implement the technology.
Just as making the most of the Internet requires the development of cloud computing and large databases, there will need to be new technologies developed to get the most out of AI.
To deploy AI, companies need to do more than just install or plug in software. They need accompanying capabilities, such as databases, data management systems, and IT specialists. And they need to make significant organizational and process changes.
The barriers to adopting new technology are internal and within a company’s control and external, which it cannot directly influence. The most important internal barrier will likely be the cost involved, particularly in the near term, while regulatory uncertainty is likely to be the primary external barrier. Governments could impose limits on the ways AI can be deployed, and companies may anticipate facing costs associated with data security and privacy regulation.
All of these factors mean that there may not be an imminent surge in productivity over the next few years, and it may take until the late 2020s or early 2030s for the benefits to take effect.
The potential for growth also comes with the caveat that the areas AI may be able to automate fully would not be perfect substitutes for those that it cannot. This means that as part of the economy becomes increasingly automated, the essential but unautomated, less productive part of the economy would grow as a share of GDP, limiting overall growth.
And just as the ICT revolution primarily boosted productivity in the US but less so in the eurozone, productivity gains from AI are by no means guaranteed and will depend on whether countries have the factors in place to help them use it effectively, according to Capital Economics.
The pace of AI deployment and its impact on workplace productivity does not only depend on technical feasibility — AI will likely deliver a significant boost to productivity if several factors come together, including investment, retraining of the workforce, and a balanced regulatory regime.
To the extent that past GPT-driven revolutions indicate the impact on productivity, AI could deliver an annual increase of 1.5%. This is more likely to materialize in the late 2020s and 2030s rather than over the next few years. And the extent to which countries successfully integrate AI into workflows and benefit from productivity gains will differ significantly.